NON-PROFIT HOSPITALS AND COMMUNITY HEALTH: HOSPITAL STRATEGY UNDER THE AFFORDABLE CARE ACT’S ENHANCED COMMUNITY BENEFIT REGULATIONS A DISSERTATION SUBMITTED TO THE FACULTY OF THE UNIVERSITY OF MINNESOTA BY HENRY STABLER IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DR. KATIE WHITE (CO-ADVISER) DR. TIMOTHY BEEBE (CO-ADVISER) AUGUST 2022 Copyrighted by Henry Stabler 2022 All rights reserved i Acknowledgements To say that this doctoral process was a difficult, arduous journey would be an understatement to beat out all other understatements. Yet, now that I have reached the end of this journey, ready to start a new one, I look back with complete and total satisfaction that I would not have changed a thing and that I did my absolute best. It feels pretty good. There are so many people I now need to acknowledge and thank for helping me along the way. To my family, I owe you everything: my wife Becca whose love, trust, and devotion I continually work to earn; my son Everett, whose humor and reverence for his family has kept me going when things were most difficult; and my daughter Cora, whose happy personality has been a welcome oasis and whose health today is the thing I am most grateful for above all else. My mom and dad, who will forever be my heroes; my brothers, Tom and Griff, who are my best friends and always will be; my in-laws, Felton and Susan, who have been an invaluable source of support; and the rest of my extended family, who have provided the love and encouragement we’ve needed to get through these past seven years. I love you all so much. To my dissertation committee, starting with my advisor Dr. Katie White: you were the first person to hear my out-there, wild idea for a dissertation and always believed in me and in the importance of this research. I have relied on you to get this dissertation out into the world and I am indebted to you for your never-ending support and counsel. My co-advisor, Dr. Tim Beebe: you took on the task of advising midway through my dissertation, despite the enormous responsibilities associated with running our division and then the entire School of Public Health, and still contributed so much to ii making my dissertation better. My former co-advisor, Dr. Jim Begun: you recognized the value of my work from the very start and not only validated the enormous effort that I put into it, but helped me see how valuable the potential research contributions of my dissertation could be. Dr. Dori Cross: I have been able to rely on you to provide clear- headed advice and direction, and you have always been accommodating and fair. I’ve learned so much from your guidance and am incredibly grateful. Dr. JP Leider: your deep understanding of my research and what I have tried to accomplish have guided me from the very beginning of this process. You have consistently looked out for me and my research by holding me to clear standards and connecting me to peer researchers. I am very thankful for all of your support and am looking forward to continuing under your mentorship. Dr. Valerie Yeager: I cannot tell you how grateful and honored I am to have you on my committee as my external member. Your contributions have been incredibly validating. To you all: thank you for your patience and understanding throughout; I’ve been so lucky to have such a supportive committee. To my two research assistants, Denise Quintanilla and Alyson Sundby: you’ve both helped me with some of the less glamorous tasks associated with putting together this dissertation, but your help has been so vital. Thank you both for your invaluable assistance in getting this dissertation off the ground. To the UMN Consortium on Law & Values: thank you for providing the funding I needed to get started. I would have gone nowhere fast without your financial support. To my friend and former colleague Gary Cattabriga: more than anyone, you instilled in me a love for discovery via data, and for that I’m grateful. iii To my friends Dr. Richard Bamattre and Inȇs Sacchetti: you have helped me so much behind the scenes over these past few years, and I’m so appreciative for your friendship and support. To my fellow doctoral students: I have been blessed to matriculate at one of the most supportive doctoral programs out there, and you have been such a big part of that support - especially my fellow student parents. Thank you so much. To UMN faculty and staff: I could spend pages detailing the many, many ways you have guided me over these past years. There are several people who have been especially helpful and who I need to publicly thank, so that they know how much I appreciate them: Dr. Joe Gaugler, Dr. Zach Baker, Dr. Lauren Mitchell, Dr. Sarah Gollust, Dr. Lynn Blewett, Ms. Maureen Andrews, and Ms. Sarah Trachet. There are also a few people that are no longer here physically that I also need to thank: Dr. Ron Deprez, the person who first inspired an interest in community health needs assessments; Dr. Douglas Wholey, for ensuring I wouldn’t leave the University of Minnesota without a healthy respect for organizational theory and its relevance in health care; and finally, Dr. Dan Schnobrich, a good friend who provided needed help at the very beginning of the dissertation process. Every day that I spent working on this dissertation was a day I hoped to honor your memories. I do not know what I have done to deserve all of this love and support, but I am eternally grateful for every last bit of it. iv Abstract To better address the root causes of poor population health, US policymakers have advanced a range of policies meant to encourage greater cooperation between hospitals and local stakeholder organizations, centered around improving population health through investments addressing “upstream” determinants of health. This dissertation utilizes public reporting associated with one such policy – expanded community benefit regulations for nonprofit hospitals that require them to periodically assess the health of their communities and then develop a plan for addressing key health issues – to examine how hospitals’ investments in their community’s health and their relationships with other organizations have assembled since 2010. Content analysis was used to compile information from a sample of reports issued by hospitals to develop a rich database of hospital strategy characteristics. The resulting data was first used to group the networks formed between nonprofit hospitals and stakeholder organizations in response to the requirement to assess community health. The second component of this dissertation utilized a community orientation framework to measure different aspects of the investments made by hospitals to address prioritized local health issues. After analyzing the underlying structure of these different aspects using confirmatory factor analysis, the resulting factor scores were then used to develop a composite measure of hospital community orientation. This composite measure was then analyzed in conjunction with the first component of the dissertation to determine if highly integrated, diverse, and decentralized networks address community health issues differently relative to other hospital-based networks. The implications of the findings, particularly as it pertains to policy, are then discussed. v Table of Contents List of Tables & Figures ................................................................................................................ vii Chapter 1 - Introduction ................................................................................................................... 1 Research Aims ............................................................................................................................. 5 Chapter 2 - Literature Review and Conceptual Model .................................................................... 1 Community Orientation in Hospitals ........................................................................................... 1 Hospital-Level Effects ................................................................................................................. 3 Community-Level Effects ............................................................................................................ 4 Conceptual Model ........................................................................................................................ 6 Chapter 3 Methodology ................................................................................................................. 13 Data Sources & Study Population .............................................................................................. 13 Stakeholder Classification & Role in CHNA / Implementation Strategy Process ..................... 16 Implementation Strategy Characteristics ................................................................................... 21 Hospital / Community Characteristics ....................................................................................... 26 Research Aim 1: Development of CHNA Network Typology .................................................. 27 Research Aim 2: Development of a Community Orientation Index Measure ........................... 30 Interrater Reliability ................................................................................................................... 30 Chapter 4 - Developing a Typology of Network Structures Associated with Conducting a Community Health Needs Assessment .......................................................................................... 32 Examination of Three Network Measures ................................................................................. 33 Cluster Analysis ......................................................................................................................... 40 Transitions Between Clusters .................................................................................................... 48 Discussion of Results ................................................................................................................. 49 Conclusions ................................................................................................................................ 53 Chapter 5 - Development of Community Orientation Composite Measure .................................. 56 Examination of Five Community Orientation Measures ........................................................... 56 Exploratory and Confirmatory Factor Analysis ......................................................................... 60 Confirmatory Factor Analysis Results ....................................................................................... 67 Measurement Invariance ............................................................................................................ 70 Cluster Analysis ......................................................................................................................... 72 Factor Analysis and Cluster Analysis Results ........................................................................... 74 Cluster Characteristics ............................................................................................................... 77 Change in Cluster Grouping ...................................................................................................... 79 vi Examination of CFA Results ..................................................................................................... 80 Change in Strategy Among Hospitals .................................................................................... 80 Use of the Community Involvement in Planning Measure .................................................... 88 Establishing A Composite Index Measure ................................................................................. 93 Analysis of Community Orientation Composite Measure ......................................................... 94 Conclusions ................................................................................................................................ 96 Chapter 6 - Conclusion ................................................................................................................ 100 Discussion of Findings ............................................................................................................. 100 Implications of Findings .......................................................................................................... 106 Direct Research & Practice Implications ............................................................................. 106 Implications for Future Community Benefit Policy............................................................. 110 Future Research Directions .................................................................................................. 113 Study Limitations ..................................................................................................................... 115 References .................................................................................................................................... 118 Appendix A: Detailed Codebook ................................................................................................. 126 vii List of Tables & Figures Table 3.1 Nonprofit hospital sample characteristics (%) ............................................................... 15 Table 3.2 Variables compiled from CHNA Reports and Implementation Strategies .................... 19 Table 3.3 Implementation strategy measures ................................................................................. 23 Table 3.4 Key Secondary Data Sources and Measures Used Across Research Aims ................... 27 Table 4.1 Measures of Integration Across Three CHNA Reporting Cycles .................................. 36 Table 4.2 Measures of Centralization Across Three CHNA Reporting Cycles ............................. 38 Table 4.3 Measures of Homophilization Across Three CHNA Reporting Cycles ........................ 40 Table 4.4 Mean Integration, Centralization & Homophilization Values for Five CHNA Network Configurations ............................................................................................................................... 42 Table 4.5 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 1 ...................................................................................... 44 Table 4.6 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 2 ...................................................................................... 44 Table 4.7 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 3 ...................................................................................... 45 Table 4.8 Transition Probabilities from CHNA Cycle 1 to CHNA Cycle 2 .................................. 48 Table 4.9 Transition Probabilities from CHNA Cycle 2 to CHNA Cycle 3 .................................. 48 Table 5.1 Summary Statistics of 5 Community Orientation Measures Over 3 CHNA Cycles ...... 58 Table 5.2 Diagnostic Tests for Conducting Confirmatory Factor Analysis ................................... 61 Table 5.3 Pairwise Correlation Matrix for First CHNA Reporting Cycle ..................................... 63 Table 5.4 Pairwise Correlation Matrix for Second CHNA Reporting Cycle ................................. 63 Table 5.5 Pairwise Correlation Matrix for Third CHNA Reporting Cycle .................................... 64 Table 5.6 Standardized Factor Loadings for Community Orientation Measures Using Confirmatory Factor Analysis ........................................................................................................ 66 Table 5.7 Goodness of Fit Measures for Single-Factor Community Orientation Model ............... 69 Table 5.8 Invariance Testing Across Three Time Periods ............................................................ 72 Table 5.9 Description of Three Cluster Model Overall and Across Three CHNA Cycles ............ 76 Table 5.10 Cluster Membership Over 3 CHNA Cycles ................................................................. 78 Table 5.11 Standardized Factor Loadings for Four Community Orientation Measures Using Confirmatory Factor Analysis ........................................................................................................ 91 Table 5.12 Goodness of Fit Measures for Single-Factor Four-Measure Community Orientation Model ............................................................................................................................................. 91 Table 5.13 Invariance Testing Across Three Time Periods for Four-Measure Single Factor Model ....................................................................................................................................................... 92 Table 5.14 Mean Community Orientation Overall and by Hospital- and Community Characteristics ................................................................................................................................ 95 viii Figure 2.1 Conceptual Model of Hospital Compliance with IRS Expanded Community Benefit Regulations .................................................................................................................................... 10 Figure 4.1 Box Plot Distribution of Three CHNA Network Measures .......................................... 34 Figure 4.2 Five Cluster-Model Membership Frequency over 3 CHNA Reporting Periods .......... 43 Figure 5.1 Box Plot Distribution of 5 Community Orientation Measures Over 3 CHNA Cycles . 57 Figure 5.2 Five Measure Single-Factor Model for Community Orientation ................................. 65 Figure 5.3 Transitions Between Community Orientation Groupings ............................................ 79 1 Chapter 1 - Introduction According to a 2013 Institute of Medicine (IOM) report on international population health comparisons, the U.S. population has consistently poorer outcomes across the lifespan continuum relative to peer high-income countries.1 While there is no single answer for why the U.S. population is so unhealthy as compared with other peer countries, significant contributors to these poor outcomes include the range of behavioral, environmental, economic, and social factors that are empirically linked to poor health outcomes - the social determinants of health.1,2 The U.S. also has fractious, overstretched, and underfunded public health and social service systems compared to other countries, which are largely separated from the U.S. clinical health care system despite efforts to improve coordination across these systems. Better resourced public health and social service organizations that are more interconnected with the health care system are much better equipped to address different determinants of health, and thus lead to better population health outcomes and potentially lower health care spending.3,4 To address these shortcomings, policymakers have advanced policies meant to encourage collective action and investments in programming that address the social determinants that shape population health.5–7 The 2010 Affordable Care Act (ACA) included policies meant to reward hospitals for adopting community-focused, collaborative programs and services that lead to reduced utilization of health care services and better health outcomes. Among these ACA policies is a provision that expands the community benefit requirements of non-profit hospitals (NPHs). NPHs, classified as 2 501(c)3 organizations, must provide an annual measure of “community benefit” approximate with the value of their tax exemption. Typically, this includes charity care for low-income patients, deficits from caring for Medicaid patients, and financial support for health programming and sponsored health services.8 Under the expanded community benefit requirements, NPHs must conduct a community health needs assessment (CHNA) every three years to identify priority health issues facing their communities, and must develop an implementation strategy that specifies their planned programming for addressing these health issues.9 During the process, NPHs are also required to reach out to certain stakeholders (medically underserved populations in their communities, and state/local health agencies), and are encouraged to collaborate with multisector community organizations in developing effective implementation strategies for addressing the identified health issues.9 They also are required to evaluate their chosen interventions and report their findings in subsequent CHNAs.9 Hospitals and health systems are well-positioned to collaborate with and/or lead efforts at addressing the health issues that afflict the communities they serve. Hospitals are eminently familiar with the health issues of their local patients and are generally trusted by community stakeholders10; they also manage a rich dataset of patient information that can be tapped towards designing community-level interventions.6,7 They also are generally better-resourced than other community organizations, and could potentially add significantly to advancing public health activities in the communities they serve.8,11 While encouraging collaboration between hospitals and other stakeholders has been a focus of public policy, there is limited empirical evidence of collaboration having 3 a sustained, significant impact on population health.12 Mays and colleagues found that expansions of public health systems’ network of multisectoral partners led to decreases in influenza, diabetes, and cardiovascular disease-related mortality over sixteen years.13 Other studies that have looked specifically at whether local health departments that report high levels of collaboration with NPHs are associated with improved health outcomes have found little to no significant health effects.14 Evaluations of different multisectoral collaboration efforts to improve population health outcomes have also illustrated some success, but in intermediate measures, such as improved coordination between settings (e.g., hospital to community),15,16 suggesting that health impacts may take longer to materialize.16 Unfortunately, we lack a clear understanding of what motivates and sustains these partnerships, nor do we know what specific mechanisms are necessary to produce lasting significant improvement in population health outcomes. Though our empirical knowledge regarding if and how multisector collaboration might lead to more effective and seamless delivery of social and health services is limited, many policymakers continue to push for such collaboration – particularly between NPHs and their local health departments. This extends to the motivation behind the expanded community benefit regulations under the ACA, which requires NPHs to consult with public health practitioners as part of the CHNA process.9 Some states have advanced their own community benefit requirements for NPHs to dovetail with the federal CHNA requirements, in exchange for state tax exemptions. New York, for example, goes so far as to require NPHs to work with their local health departments on their CHNA and then jointly to select at least two health priorities from the New York Prevention Agenda to address collaboratively through their implementation strategies.17 4 This requirement has resulted in New York-based NPHs spending, on average, $393,000 to $786,000 more per year on population health initiatives than other U.S. NPHs.18 It may be that more collaboration between NPHs and other organizations can lead to greater sense of accountability to their community on the part of NPHs (i.e., make them more community-oriented) and thus lead to higher investment in community health from NPHs. The public reporting of CHNAs and implementation strategies offers an opportunity to assess the extent to which community benefit policy has engendered multisectoral collective action among NPHs and their community partners; and whether interventions chosen by NPHs are “community-oriented” in terms of what is implemented, which stakeholders are involved, and which populations are targeted for intervening. More broadly, researchers and policymakers also need a much better understanding of what works when it comes to advancing population health. The required community benefit reporting offers an opportunity to assess how NPHs view population health management, and where they see their role in advancing community health. It also elucidates the optimal community and regulatory environment that helps facilitate better population health management among NPHs. A better understanding of these considerations is necessary to determining if and how NPHs can build an effective, equitable culture of health through public policy. 5 Research Aims This study utilizes available CHNAs and implementation strategies to examine hospitals’ community linkages and the effects of those linkages more closely on health outcomes. I pursue two research aims in the study:  Research Aim 1: Using data collected on the range of CHNA-related activities undertaken by a sample of NPHs and their partners, I classify each NPH’s network of collaborators into homogenous groups using cluster analysis. These homogenous groups are defined using measures of organizational integration (the number of organizations participating in the CHNA process) and centralization (the number of CHNA-related activities undertaken by only NPHs) as developed by Bazzoli, et al.19 A third strategic measure, network homophily (i.e., how similar network collaborators are to NPHs) is also used to organize network groupings.  Research Aim 2: Adopting a framework developed by the Public Health Institute (PHI), I develop an index score to describe the level of community orientation of each NPH’s implementation strategy. The PHI framework provides five core principles that should characterize NPHs’ strategies for addressing priority health issues. These principles include: (1) an emphasis on serving communities with disproportionate unmet health needs; (2) emphasis on primary prevention; (3) development of links between clinical services and community-based services/activities; (4) advancement of community asset building; and (5) inclusion of community stakeholders in the planning of implementation strategies.20 Each sampled NPH’s implementation strategy is scored based on 6 adherence to these five principles and aggregated into a summary PHI index score. I then investigate an explanatory model for the level of hospitals’ community orientation. This study provides a unique and significant contribution to the literature by (1) compiling and organizing information on how NPHs have partnered with key stakeholders and how they are investing in their community; (2) clarifying how NPHs view community health management, and where they see their role in advancing community health; and (3) elucidating the optimal community and policy environment that helps facilitate better community health management among NPHs. The resulting data and analyses provide an exceptionally detailed look at how hospitals have approached community health management in response to a change in community benefit regulations. The remainder of this dissertation is organized as follows. In Chapter 2, I provide a review of the literature on determinants of community orientation in hospital decision- making, followed by a discussion of the conceptual model that was used to guide this research. In Chapter 3, I discuss the research methodology, including my chosen approach to primary data collection, use of secondary data, and a summary of the analyses used for each research aim. The next two chapters (Chapters 4 and 5) then cover each research aim in detail, including results. Finally, I discuss the implications of my findings in Chapter 6. 1 Chapter 2 - Literature Review and Conceptual Model As defined by Proenca and colleagues, the community orientation of hospitals is the extent to which hospitals rely on the “generation, dissemination, and use of community intelligence” in the formation of strategies to manage community health needs (p 1013).21 Hospitals vary in how they expend resources to address pressing health issues, with many opting to invest more time and money in community-oriented strategies. Other, less community-oriented hospitals, may see expansions of acute care services as a better means of meeting their missions, to the detriment of community partnerships and investments in upstream interventions. This chapter begins with an examination of the research into the factors that influence hospital decision-making as it pertains to community-oriented investments (both community partnerships and financial/material resources), followed by a discussion of the conceptual model that informs my study’s research aims. Community Orientation in Hospitals Hospitals have long faced a litany of changing pressures and expectations, and many have responded to these changes by becoming more invested in community health. Understanding the external and internal incentives that might lead hospitals to adopt such strategies has been a focus for researchers, who have typically relied on either an institutional or a resource-dependence theoretical framework to explain hospital 2 behavior.21 Institutional theory suggests that hospital behavior is predicated on conforming to the expectations and norms of the environment in which hospitals operate. By conforming to their environments, hospitals gain legitimacy, which in turn comes with public support that provides the necessary resources to survive.22 Community- oriented hospitals, under this framework, reflect a need to placate those with the power to confer legitimacy, such as payers, accreditation organizations, and local and state governments. Resource-dependence theory suggests that hospitals act in ways meant to reduce their dependence on those stakeholders who control the resources they need to survive.23 Thus, those hospitals that adopt a community-orientation strategy may be trying to manage their dependence on managed care organizations, payers, and legislators for whom prevention and lowering of health care utilization is a priority.21 Many organizational theorists suggest these two theories are complementary, and that they outline the range of possible behavioral responses from NPHs including the fragmentary, conflicting, variable set of incentives and changing market environments.21,24 The type of community partnerships being formed under the ACA’s enhanced community benefit regulations are unique, in that, unlike traditional community partnerships which are largely voluntary, the ACA requires NPHs to work with certain community stakeholders during the CHNA process to inform how they identify priority health issues.9 In responding to the regulations, some NPHs will choose to satisfice by limiting their interactions with other organizations to what is required and/or limiting their investments in addressing their community’s health priorities, recognizing that straying too far from a standard acute care model is not worth the investment risk. Others 3 may recognize the benefits of collective action and seek to engage a wide array of partners and/or invest in upstream community health programming. Hospital-Level Effects As reported by Proenca and colleagues, certain hospital characteristics are also associated with higher inclinations to invest in and participate in collective action to address population health 21 Among these characteristics, larger NPHs, as measured by bed size, have been found to be positively associated with NPH community orientation, as has network- and system-affiliation.25 Payer mix has also been found to impact NPHs involvement in community health programming. Greater hospital reliance on managed care and HMOs are positively associated with the degree of community orientation,25 while higher reliance on Medicaid is associated with a lower degree of community orientation.26 Less clear is the effect of the financial performance of hospitals on community orientation, i.e., whether better financed hospitals are more likely to be community- oriented than their less financially secure counterparts. Given the increasingly complex system of reimbursement in the US healthcare system that relies on both fee-for-service and value-based reimbursement, financially successful hospitals are possibly more likely to be those that can manage the internal and external stakeholders responsible for hindering or assisting in positive financial performance. Previous research on this topic, however, has been limited. Jennings, et al., for example, found that, while some measures of financial well-being, such as total margin, were positively associated with community orientation, others such as return on assets, were inconclusive.27 4 If we consider community orientation a form of corporate social responsibility (CSR), then the literature is clearer. In general, research has found a positive relationship between corporate actions meant to enhance social welfare and organizational financial success.28,29 An organization’s efforts to engage in CSR results in increased trust among relevant stakeholders, which results in decreases in transaction costs and increases the likelihood of positive financial profitability.30 From this perspective, hospital proclivity to engage in collective action with local stakeholders to address pressing population health-level health issues may be associated with positive financial performance. Community-Level Effects It is also important to consider the community environment in which hospitals operate. Communities provide hospitals with patients of varying levels of socioeconomic status and morbidity, which affects a hospital’s revenue; communities can also enforce expected social norms, which may also influence a hospital’s conduct. Researchers have identified a range of community-level economic and environmental factors that positively affect hospitals’ involvement in community health and health promotion. These factors include increases in median income, decreases in unemployment, lower percent of households living in poverty, and lower proportions of populations with those 65 and older.26,31,32 NPHs in rural areas also lag behind their urban peer NPHs in their level of involvement in community health; while rural areas tend to be more socially compact and less populous, they also tend to have less resources to advance community health.33 Relatedly, the presence of organizations and governmental agencies devoted to 5 improving community health is likely to affect the proclivity of hospitals to participate in community health promotion activities.25,26 Hospital markets also influence hospital behavior. Research has found that NPHs in general are more likely to be involved in community health initiatives compared with their for-profit counterparts;31,34,35 however, this difference disappears if an NPH is a sole provider in a market, suggesting that a monopolistic market can be a hindrance to hospitals’ involvement in community health.31 Similarly, Proenca and colleagues found that higher market concentration and higher levels of hospital competition lead to a greater degree of community orientation among hospitals.21 Lastly, community orientation in hospitals may be correlated with a community’s level of social capital. Conceptually social capital has been defined in different ways; however, it is generally described as the level of connectedness and quality of social relationships across individuals and groups.38 On a community-level, social capital can be considered a public good that helps facilitate collective work towards a set of common objectives.39 Research has generally established a strong, positive association between community social capital and population health, although much of the specifics for how it functions to improve health remain elusive.36 More specific to health systems, Mays and colleagues reported a strong, positive relationship between public health multisector collaboration and community-level social capital, and community health outcomes.13 As community social capital indicates how well a community functions to support its members, it offers a plausible causal pathway for explaining differences in population health outcomes. From this perspective, those communities with higher levels of social capital may be more influential in affecting the level of community orientation of their 6 local NPH through enforcement of local norms – for example, if community residents are more aware of their community’s health issues and are collectively active in redressing these issues. Likewise, hospitals operating in communities with high levels of social capital may be more collaborative than their counterparts with less social capital. Given the associations between social capital and population health, these factors may all contribute to healthier communities. Conceptual Model The conceptual model for this research is based on the theoretical framework put forth by Oliver, who combined the perspectives of resource dependence and institutional theories to define a list of determinants for explaining organizational responses to environmental pressures.24 Per Oliver’s framework, an organization’s response to regulatory and environmental pressures to conform their behavior is dependent on what the pressures are, which stakeholders are exerting pressure, how their peer organizations respond, and how and where pressure is exerted.21,24 Oliver identified five determinants of organizational response: cause, constituents, context, content, and control.24 Cause refers to the possible sources of pressure for why a hospital might opt to conform (or not) to playing a significant role in community health. Hospitals are subject to a range of expectations from a multitude of stakeholders. This includes the widespread expectation, especially among policymakers, that hospitals should provide cost-effective care while also caring for the community-at-large, stemming from the belief that more community-oriented hospitals are critical to lower health care costs, higher quality care, and better access to care.17 Larger firms are more likely to be visible to external 7 stakeholders – especially policymakers and regulators - and this likely extends to hospitals. Hospitals that serve larger constituencies are more likely to acquiesce to public pressures to invest in community health to avoid public inquiry;41 this is particularly important for NPHs given the extensive public scrutiny of the state and federal tax exemptions provided to NPHs. Because of their reliance on public research funding, this also likely extends to teaching hospitals. Finally, hospitals that are more financially secure may want to invest the time and resources to improve community health, thus earning positive public perception. Constituents refers to the external stakeholders who are powerful enough to impose expected norms and behaviors. Because of the fragmented nature of the US health care system, especially among payers, there are no shortage of stakeholders who have the power to enforce these expectations on hospitals. However, because these stakeholders may hold differing expectations, hospitals can shirk on some stakeholder requirements depending on the relative power of each stakeholder. Where hospitals rely on revenue from one payer, they may be more indebted to follow the requirements of that payer. Managed care organizations that attempt to mitigate risk through capitated contracts or discounted fee structures may push hospitals to adopt a community-oriented stance because of the potential for cost-savings if a hospital is successful at keeping patients from utilizing unnecessary services. Likewise, hospitals that are members of Accountable Care Organizations (ACOs) also will face pressure to mitigate risk by investing in upstream interventions (and/or collaborate with community organizations) that may prevent unnecessary health care service utilization. 8 Context refers to the environmental context in which organizations operate. Oliver suggests that context affects organizational behavior across two dimensions – through uncertainty and through organizational interconnectedness. Where the level of uncertainty is elevated, e.g., where risk of non-payment from their patients is high, hospitals may be more likely to resist investing in community health where the payoff is long-term and comparatively risky to further investments in acute care. Thus, policies that reduce uncertainty, such as expansion of patient coverage through Medicaid, may incentivize hospitals to invest in riskier community health initiatives and collaboration. Other community characteristics that may affect a hospital’s level of risk, such as the number of uninsured and the level of poverty, may also affect their strategic thinking. Diffusion of norms, including the extent to which a hospital adheres to organizational expectations around participating in community health initiatives, are more likely when these organizations are connected. This pressure to conform may take several forms. If a hospital is part of a provider network or system where community involvement is expected, it follows that that hospital adheres to the institutional expectations. Market competition may also spur hospitals to adopt community-oriented strategies, particularly if competing hospitals have gained legitimacy by adopting these strategies. Like any other type of organization, hospitals are also likely influenced by non-provider organizations, such as advocacy organizations, public health and social service agencies, civic groups, and educational institutions. To the extent these types of organizations operate alongside hospitals and can help hospitals manage constituent demands, hospitals may be more willing to invest in community health. This would suggest that communities in which the density of these organizations is sparse, e.g., rural 9 communities or communities where civic engagement is low, we would expect to see hospitals less likely to adopt a more community-oriented investment strategy. Control refers to how external demands operate – for example, through governmental authority or voluntary dissemination. While community orientation per se may be voluntary, the means by which community benefit regulation is enforced is through governmental regulations and all NPHs are subject to these federal regulations. Some NPHs also face additional regulatory pressure from state or local governments, but, at least as it pertains to policies that require NPHs to pursue community-oriented strategies via their community benefit plans, these policies only exist primarily in New York and Ohio.37 For NPHs in these two states, the external demands to pursue a community-oriented strategy is heightened. The remaining determinant identified by Oliver is also critical to understanding the range of pressures that hospitals face, but is difficult to measure accurately. Content refers to the extent hospitals will conform to demands that diverge or align with their own interests or values. In theory, the demand on NPHs to respond to community health needs should align with nonprofit ethos that prioritizes service to community, especially relative to for-profit hospitals that prioritize profitability. Thus, the regulatory demands of the expanded community benefit regulations should be compatible with most, if not all U.S. NPHs. In practice, however, NPHs will vary in how closely they adhere to this ethos – and likely depends on the beliefs and values of those in leadership positions at each NPHs. Some NPHs, for example, act more like their for-profit peers than a NPH; seven of the ten most ‘profitable’ hospitals as measured by net income from patient care services per adjusted discharge were NPHs.38 10 It is important to note that community health needs, such as abnormally high rates of disease incidence, are only relevant to hospitals’ behavior if high rates of disease are relevant to the different sources of demand on hospitals. For example, for a hospital serving a highly impoverished community with corresponding high rates of disease may reflect a need for more upstream interventions but fail to move hospital strategic thinking because of politically weak constituents or a context that requires less risky investments (e.g., expansions of acute care services). Figure 2.1 Conceptual Model of Hospital Compliance with IRS Expanded Community Benefit Regulations Figure 2.1 illustrates the conceptual model for this research based on Oliver’s theoretical framework. For a given hospital, CB strategy starts with decisions around 11 which partners to involve in completing their CHNA and implementation strategy, what these partners’ role should be, and for what purpose. These decisions around collaboration are built on and lead directly to hospital decisions around how best to address their identified priority health issues and the extent to which these strategies are community-oriented in their design. Hospital decision-making is also subject to a range of external and internal factors, including the state and local environment (i.e., rural vs urban; additional CB regulatory requirements from states and municipalities), local market characteristics, hospital size, and payer mix (e.g., hospital use of managed care and participation in alternative payment schemes; reliance on Medicaid). Hospital leadership also plays a role in determining strategy and the extent to which hospital executives and their board members value community-oriented strategies over other, more clinically-based strategies for addressing local health issues. Hospital leadership also learns over time, what works best given their competing demands and changes in community health, and then may adjust their strategies to be more (or less) community- oriented based on their experiences. As illustrated in Figure 2.1, the two research aims fits into the conceptual model as well:  Research Aim 1 examines how NPHs have organized themselves and their partners as they have complied with the expanded community benefit regulations to complete their CHNA and implementation strategies. How NPHs have organized their network of collaborators may change between CHNA cycles depending on changes in their local context (e.g., changes to state community 12 benefit regulations that require NPHs work with more expansive networks of collaborators).  Research Aim 2 examines the degree of community orientation in implementation strategies and how they are influenced by their network of community stakeholders, as well as the context, constituents, and causes that affected who NPHs worked with to complete their CHNAs. Finally, the extent to which hospitals expand their network of collaborators and invest in addressing their priority health issues leads to changes in their community social capital, i.e., through strengthening and/or expanding ties between organizations and through greater visibility within their community. These changes to community social capital may potentially lead to improvements in population health outcomes.36 In this chapter, I reviewed previous research related to hospital community orientation and its hospital- and community-level determinants. I also described the conceptual model guiding analysis of my three research aims. In Chapter 3, I discuss the primary and secondary data that will be used in analysis of the three research aims. 13 Chapter 3 Methodology This chapter lays out the methodology of this study, including my chosen approach to primary data collection, use of secondary data sources, as well as an overview of the analytic approach taken to addressing each research aim. A more detailed discussion of each research aim’s analysis plan can be found in each subsequent chapter. Data Sources & Study Population To obtain data for this research, I used content analysis - a method of extracting information from narrative documents using an iterative coding scheme39 on available CHNA and implementation strategies and which was then converted into a numeric database for quantitative analysis. I derived a list of all NPHs in the U.S. via the Centers for Medicare & Medicaid (CMS) Healthcare Cost Report Information System (HCRIS). All CHNA and implementation strategy reports were found by searching hospital websites via Google (i.e. “[NAME OF NPH] community health needs assessment”) because, according to IRS regulations, all NPHs must make their CHNAs available to the public, including through their websites. Searching started in April 2019 and finished in December 2020. Through these efforts, I have compiled CHNA reports for most NPHs in the US (approximately 9,000 CHNAs and implementation strategies). All data were collected using REDCap40 by a team of 3 coders, which included me and two graduate research assistants. 14 I collected data from a 4% sample (n=125) of the entire population of NPHs in the US (n = 2,946 hospitals). My list of NPHs was randomized to reduce the possibility of selection bias. To be included in the sample, NPHs had to have both a CHNA and implementation strategy for at least three CHNA “cycles”, defined as a period of 3-4 years since 2010 (i.e., 1st cycle = 2011-2014; 2nd cycle = 2015-2017; 3rd cycle = 2018- 2021). I also only included general short-term acute care hospitals. A majority of NPHs were missing either a CHNA and/or implementation strategy and thus were excluded from the final sample. Out of 2,946 U.S. NPHs, only 1,362 NPHs (46%) had both a CHNA and implementation strategy publicly available for all three cycles. On first inspection, many of these excluded NPHs were smaller, rural hospitals; however, several large health systems, such as Kaiser Permanente and Mayo Health System were also excluded. There are several possible reasons for missing CHNA and implementation strategy reporting, including:  NPHs have the option of making their implementation strategies available through their website or through their annual tax filing (Schedule H of Form 1090), with many opting for the latter. For those NPHs that opt to report their implementation strategies through their tax filing, all narrative of their strategies is retrospective and thus differs from most implementation strategies that are prospective;  Many NPHs remove CHNAs or implementation strategies if, for example, they revamp their website, or lose their CHNAs due to changes in ownership; and  Many NPHs have not finished their latest CHNA and implementation strategy; 15  Some NPHs may have not completed a CHNA and implementation strategy, or never made their reporting publicly available. Table 3.1 displays key hospital and county characteristics of my NPH sample compared to all US NPHs. Table 3.1 Nonprofit hospital sample characteristics (%) Characteristics Sample (n=125) U.S. Nonprofit Hospitals (N=3,049) Hospital Characteristics Size Small (<100 beds) 46% 44% Medium (101-299 beds) 30% 32% Large (300+ beds) 24% 24% Teaching Hospital 34% 31% Critical Access Hospital 33% 27% Participation in Medicare Alternative Payment Model 75% 72% Religious Affiliation 18% 11% Ownership Stand-Alone 18% 17% Health System / Network Member 82% 83% County Characteristics ≥10% of Households Earn Less Than or Equal To Federal Poverty Line 38% 43% Rural 38% 37% Urban 62% 63% Census Region Northeast 18% 18% South 29% 29% Midwest 33% 34% West 21% 18% Data Sources: (1) Community Benefit Insight; (2) 2016 Area Health Resource File; (3) Compendium of U.S. Health Systems 16 As Table 3.1 illustrates, my sample of NPHs compares well across most hospital and county characteristics, with most sample characteristics within 5% of national figures. My sample has a higher proportion of religiously affiliated hospitals as well as a higher proportion of critical access hospitals relative to the larger population of U.S. NPHs. Hospitals in my sample also serve in less impoverished counties, with only 38% of sample NPHs residing in counties where 10% of families (or more) make equal to or less than the Federal Poverty Line, relative to 43% of hospitals nationally. Stakeholder Classification & Role in CHNA / Implementation Strategy Process Using the following coding scheme, I catalogued information related to the collaborators and activities undertaken by NPHs, as documented in their CHNA and implementation strategies (see Table 3.2, below). The list of CHNA and implementation strategies activities was developed first, through a careful reading of the literature, particularly materials recommended by the Centers for Disease Control & Prevention (CDC)41, which includes guidance from the Catholic Health Association42 and the Association for Community Health Improvement.43 The list of CHNA activities and their definitions includes:  Activity 1: Plan and define CHNA: This initial activity can encompass a range of activities necessary to start a CHNA, including: (1) Defining the scope of the CHNA; (2) Defining the community; (3) Defining and engaging partners; (4) Assigning leadership / oversight roles; (5) Identifying models to follow for the assessment (e.g., MAPP); (6) Hiring consultants to help with the conduct of the CHNA. 17  Activity 2: Collect and analyze data: This activity can again cover several steps, including: (1) determining appropriate population health measures; (2) collecting available quantitative secondary data and/or statistics from public sources; (3) collecting quantitative and/or qualitative primary data from local constituents through surveys, key informant interviews, focus groups, etc.; and (4) analyzing resultant data.  Activity 3: Identify priority health issues: This activity involves identifying the most pressing community health needs using data collected from Activity 2. Note that hospitals will often ask individual stakeholders what they feel is/are the most pressing health issue(s) as part of their data collection effort, and then using this information to identify their priority health issues. From the hospital’s perspective, the former activity (data collection) should be considered part of Activity 2; the latter activity (assuming the respondents fit the criteria of ‘active’) part of Activity 3.  Activity 4: Communicate results: The next step in conducting a CHNA is to communicate the results internally and externally. This can involve: (1) writing up the final CHNA report; (2) discussing the results internally, including with a hospital’s board, who must approve the final CHNA document; (3) sharing the results with external stakeholders (e.g., local media, partnering community organizations); (4) sharing with relevant federal / state / local regulatory agencies, especially if required to share by state law, such as in Maryland. 18  Activity 5: Plan implementation strategies: Based on the CHNA results, hospitals must plan if and how they are going to address each health issue – i.e., what resources are already available to utilize, what resources can hospitals allocate to address each health issue, etc. This process may be done internally, or it may be done collaboratively with community partners. If hospitals opt to not address an issue, they must also describe why. For each activity [Q1], I captured which organizations (if any) were identified by the NPH as participants [Q1A] on each activity and their level of involvement [Q1B]. For the purposes of this research, I defined relevant stakeholders as those who are active participants in the CHNA / implementation strategy process. This means that passive participants (e.g., survey respondents, key informant interviews) are not sufficient to “count” as a stakeholder. Attending meeting(s), helping to plan or actively participate in an intervention – even if that organization is not primarily responsible for completing the activity or intervention – would suffice to count as an active stakeholder participant. 19 Table 3.2 Variables compiled from CHNA Reports and Implementation Strategies Variable Domain Categories Activities [Q1] CHNA Activities: (1) Plan CHNA (Define scope / community of CHNA); (2) Identify, collect, and analyze data; (3) Define health priorities; (4) Communicate results; IS Activities: (5) Plan implementation strategy; (6) Implement strategy Evaluation Activities: (7) Evaluate strategy Type of Organization [Q1A] Clinical partners: (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; Non-clinical partners: (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare -12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others Role [Q1B] (1) Primary responsibility; (2) Secondary responsibility Priority Health Issue [Q2] (1) Access & availability of health care services; (2) Quality of care; (3) Aging & elderly health conditions; (4) Behavioral & mental health; (5) Cancer; (6) Cardiovascular disease; (7) Diabetes; (8) Disability; (9) Hypertension & stroke; (10) Infectious diseases; (11) Kidney disease; (12) Respiratory diseases; (13) Sexually transmitted diseases; (14) Obesity; (15) Chronic disease; (16) Tobacco use; (17) Diet & exercise; (18) Alcohol & 20 drug use; (19) Sexual activity; (20) Education; (21) Employment; (22) Income; (23) Community safety; (24) Food security; (25) Environment; (26) Housing / transportation; (27) Maternal & child health; (28) Oral health; (29) Emergency preparedness; (30) Other Social determinants [Q3]: What is the social determinant that is being addressed? (1) Economic stability; (2) Education; (3) Social & community context; (4) Neighborhood & built environment; (5) Health & health care; (6) Other Level of prevention [Q3A]: If addressing health & health care, what is the lowest level of prevention being addressed with this activity? (1) Primary prevention; (2) Secondary prevention; (3) Tertiary prevention; (4) Not preventative intervention (acute care); (5) Not preventative intervention (other) Continuum of Care [Q4]: Does this activity describe a community-clinical linkage? (1) Activity describes clinical-community linkage; (2) Activity does not describe clinical-community linkage Interventions Target Population [Q5]: Does this activity build upon existing community infrastructure? (1) Community-at-large; (2) Vulnerable population; (3) Community-at-large but includes targeted outreach to DUHN communities Nature of intervention support [Q5A]: If building upon existing community infrastructure, what is the type of support being provided by the reporting hospital or health system for this activity? (1) Financial support; (2) Technical assistance; (3) Advocacy; (4) Equipment or other material donations; (5) In-kind support; (6) Does not describe provision of support 21 Implementation Strategy Characteristics In addition to collecting information on the stakeholders and relationships used to conduct each CHNA and implementation strategy, I also collected data on key attributes of each chosen implementation strategy activity using the framework put forth by the Public Health Institute (PHI) and their guidance document, Advancing the State of the Art in Community Benefit Toolkit (ASACB).20 PHI offers five principles for NPHs to use in determining how best to address their community’s priority health issues. Table 3.3 illustrates how each principle was operationalized. Note that more detailed definitions for each construct have been developed as part of the content analysis coding process. The specific principles that ASACB calls on NPHs to adhere to in designing their implementation strategies are:  Emphasis on communities with disproportionate unmet health-related needs: NPH CB programming will be targeted to benefit either the community-at-large or specific subgroups with disproportionate unmet health needs (DUHNS). Alignment with this principle requires NPHs to specify in their implementation strategies if a specific activity either targets individuals with DUHNs and/or identifies specific accommodations by which individuals with DUHNs could participate in community-at-large programming.20  Building a seamless continuum of care: ASACB calls for NPHs to develop links between clinical services and community-based services and activities. For clinical service programs, this would mean establishing links to prevention services and/or to community support services. For community-based prevention 22 programs, this would call for coordination with health care providers to identify and reduce the demand for relevant clinical services. Establishing such links would improve the continuum of care for patients and community-members.20  Community capacity building: Rather than developing interventions siloed away from what already exists in the community, ASACB calls for NPHs to build upon the current community infrastructure using their available charitable resources. This approach reinforces the collective action that should underlie upstream interventions; it also encourages shared accountability, improves programmatic efficacy and visibility, and prevents replication of effort.20  Collaborative governance: ASACB calls on NPHs to involve community partners not only in the implementation of IS programming, but also as partners in the IS decision-making process. Involving community stakeholders in decisions around program design, targeted populations, and oversight, again reinforces the collective action necessary to address upstream determinants of health while ensuring buy-in from affected community constituencies. This principle necessarily involves collaborative governance in the form of community stakeholder membership on a committee or board, with equitable say with their hospital partners.20  Emphasis on primary prevention: ASACB calls on NPHs to favor primary prevention programming in designing their implementation strategies. Per their definition, primary prevention includes three types of activities: o Health promotion: Health messages to encourage healthy lifestyles in the general population (e.g., health fairs; group education). 23 o Disease prevention: Targeted interventions for at-risk populations, such as a home fall prevention for seniors or clinically-based interventions, such as programs that extend clinical services (e.g., behavioral health care, pharmaceutical, physical therapy, dental care, etc.) to patients in non- clinical settings. o Health protection: Support for changes in local environments that promote good health behaviors.20 Table 3.3 Implementation strategy measures Measure Definition Applied To (1) Does IS activity target communities with disproportionate need? [Q5] Yes = 1 if activity targets / includes DUHN communities No = 0 if community-at-large All listed IS activities (2) Does IS activity emphasize primary prevention? [Q3 and Q3A] Yes, non-Health SDoH = 1 Yes, primary health prevention = 0.75 Yes, secondary health prevention = 0.5 Yes, tertiary health prevention =0.25 No = 0 All listed IS activities (3) Does IS activity work to build a continuum of care between clinical and community? [Q4] Yes = 1 if community-based prevention intervention describes a link to clinical service delivery Yes = 1 if clinical service delivery intervention describes a link to community prevention and/or community-based support system No=0 All listed IS activities (4) Does IS activity build on what is already in place in community? [Q1 – Activity 6 and Q5] Yes = 1 if community stakeholder involved in activity implementation AND NPH offers at least one form of support to stakeholder (financial / technical support, advocacy, promotion). No = 0 All listed IS activities (5) Are community stakeholders involved in any IS activity decision- making process? [Q1 – Activity 5] Yes = 1 if community stakeholder involved in IS planning No = 0 IS planning (Activity 5) IS = implementation strategy 24 Each identified implementation strategy activity was assigned a score across four of the five measures (Measures 1 through 4; see Table 3.3, above) based on adherence to the ASACB framework. Because ASACB only provides a definition for what it considers “primary prevention” (Measure 2), I adopted the Centers for Disease Control and Prevention (CDC)’s definitions for primary, secondary, and tertiary prevention, as well as to better differentiate between prevention levels than the ASACB definition:  Primary prevention: Intervening before health effects occur, through measures such as vaccinations, altering risky behaviors (poor eating habits, tobacco use), and banning substances known to be associated with a disease or health condition. I also include expansion of primary care services and/or outreach efforts to expand the number of insured, since these measures seek to improve the health of all community members, healthy or not.  Secondary prevention: Screening to identify diseases in the earliest stages (i.e., at- risk), before the onset of signs and symptoms, through measures such as mammography and regular blood pressure testing  Tertiary prevention: Managing disease post-diagnosis to slow or stop disease progression through measures such as chemotherapy, rehabilitation, and screening for complications. I also include measures such as chronic disease self- management; recruitment of specialty providers; or care coordination efforts for chronically ill patients, since all of these measures seek to slow or stop disease progression or complications. 25 Most NPHs opt to address their priority health issues by focusing on health and health care as a determinant; however, others may seek ways of addressing non-health determinants. To further differentiation between health and non-health social determinant interventions, I used the Healthy People 2020 categorization to classify what determinants hospitals are seeking to address.  Economic stability: Interventions that address employment; food insecurity; housing instability; and poverty  Education: Interventions that address / support early childhood education and development; enrollment in higher education; high school graduation; and language and literacy  Social & community context: Interventions that address / support civic participation; discrimination; incarceration; and social cohesion  Neighborhood & built environment: Interventions that support access to foods that support healthy eating patterns; crime and violence; environmental conditions; and quality of housing Using the CDC and Healthy People 2020 definitions as a framework, for Measure 2, I assigned each intervention a score between 0 and 1, with the highest score of 1 assigned to non-health social determinant interventions; a score of 0.75 to primary health prevention interventions; a score of 0.5 to secondary health prevention interventions; a score of 0.25 to tertiary health prevention interventions; and a score of 0 to interventions that either do not address health explicitly (e.g., coalition building) or is simply a continuation of acute care services (e.g., addressing cancer as a priority health issue through existing cancer services). 26 The final measure (Measure 5) was assessed based on described stakeholder involvement in developing the implementation strategy (Activity 5), where at least one community (i.e. non-clinical) organization played a role in the production of the implementation strategy. To calculate this measure, I use a measure of homophilization, which is a network measure that assesses the similarity (or differences) of stakeholders. I added up the number of stakeholder organizations mentioned that had either primary or secondary responsibility for completing the planning of the implementation strategy, and then counted up the number of community organizations; the homophilization measure, then, is the ratio of community organizations to all organizations that worked on planning. Hospital / Community Characteristics I collected key identifiers for each NPH, such as the Centers for Medicare & Medicaid Services’ (CMS) Certification Number (CCN), NPH employer-identification number (EIN), and county name and Federal Information Processing Standard (FIPS) codes for the primary county serviced by each NPH. This allowed me to merge my primary data with other secondary datasets. Once data were compiled and cleaned, I merged my primary data with several publicly-available datasets, including: (1) available data from Community Benefit Insight (website: http://www.communitybenefitinsight.org/). This dataset compiles information from other available datasets in one dataset, including IRS 990 filings from all 501(c)(3) hospitals and systems, American Hospital Association (AHA) hospital information, the CMS HCRIS, and several other sources; (2) Health Resources and Services Administration’s 27 Area Health Resource File (AHRF); (3) Agency for Healthcare Quality (AHRQ)’s Compendium of US Health Systems (CHS); See Table 3.4 (below) for specific variables and associated secondary data sources and definitions. Table 3.4 Key Secondary Data Sources and Measures Used Across Research Aims Variable Source Definition Hospital, Community, and State-Level Variables Hospital size* CMS HCRIS Small (<100 beds); Medium (101 – 299 beds); Large (300+ beds) Hospital religious affiliation* CMS HCRIS 0 = no religious affiliation; 1 = any reported religious affiliation Hospital teaching status* CMS HCRIS 0 = not a teaching hospital; 1 = teaching hospital Participation in alternative payment models CHS Accountable Care Organizations (ACOs) or Medicare’s bundled payment model Counties w/ high levels of poverty AHRF Counties where 10% or more of households make equal to or less than federal poverty line Rural / urban AHRF Rural Urban Continuum Codes; Urban (Metropolitan Counties); Rural (Nonmetropolitan Counties) Notes: (1) * = compiled from Community Benefit Insight; (2) CBI = Community Benefit Insight; AHRF = Area Health Resource File; CHS = Compendium of US Health Systems; CMS = Centers for Medicare & Medicaid; HCRIS = Healthcare Cost Report Information System Research Aim 1: Development of CHNA Network Typology Health care researchers rely on typologies to classify the heterogeneous organizations that comprise health systems, which allows for better comparison of performance and outcomes. Of relevance to this research is the empirical typology developed by Bazzoli et al. to categorize multiorganization health care delivery systems and networks.19 This typology has subsequently been adopted by other health care 28 researchers, such as Mays and colleagues,44 who applied the same categorical measures to multisectoral public health systems. Bazzoli and colleagues developed three general categories to group multiorganizational health care delivery system characteristics: differentiation, integration, and centralization. Each measure represents different structural and strategic dimensions of healthcare organizations.19 Measures of integration and centralization are especially relevant in describing the attributes of NPHs and their multisectoral community networks that have been formed in response to the enhanced community benefit requirements. Integration refers to the extent to which services or activities are offered through the relationships formed with other organizations. Centralization is defined as the extent to which responsibility for activities are distributed among NPHs and their network of community partners. A third measure used by Bazzoli and colleagues, differentiation, refers to the type and scope of services and programs that hospitals choose to offer. Because this research aim assesses how NPHs respond to the completion of a singular, required service – the completion of the CHNA – differentiation is less relevant to this research. All NPHs will generally follow the IRS’s prescribed process for completing a CHNA. However, because NPHs can choose the diversity of organizations to involve in their CHNA / implementation activities, a more relevant measure to determine how NPHs differentiate themselves from their peers is homophilization, or the extent to which NPHs establish relationships with similar organizations.45 For NPHs, homophilous networks of community stakeholders were those primarily comprised of clinical partners, as opposed to a more community-oriented array of partners from different sectors. This 29 is an important component to measure because research has shown that like-minded partners can influence strategic decision-making.46–48 For example, NPHs that involve primarily clinical partners may be more inclined to adopt clinical interventions to address their community’s priority health issues rather than more community-oriented interventions. Based on Bazzoli et al.’s original definitions for integration and centralization, I define “integration” as the average proportion of CHNA- and implementation strategy- related activities performed by each type of organization in the defined network affiliated with each NPH. Higher levels of integration would therefore be those networks where NPHs engage with multiple stakeholders for most or all CHNA/implementation strategy activities. I further define “centralization” as the average level of effort of the NPH across each activity, where level of effort is defined as the proportion of shared responsibility with other partners for each activity. Lastly, I define “homophilization” as the proportion of all identified clinical partners to all partners (see Table 3.2 for categories of clinical and non-clinical partners). Using these measures, I developed an empirical typology of NPH network structure using a stepwise agglomerative hierarchical clustering approach. This approach allowed for nesting of additional identifiable subgroups, rather than a partitioned approach resulting in single layer of data groupings. Further details can be found in Chapter 4. 30 Research Aim 2: Development of a Community Orientation Index Measure To better characterize the community orientation of NPHs, I developed a composite measure using the five defined ASACB sub-measures. A composite index score is a single measure derived from a set of indicators that individually reflect a particular characteristic or characteristics of a larger construct.49 When combined, the composite index measure is better able to illustrate how that construct varies over time and by internal and external characteristics. After compiling all available primary and secondary data, I first described the distribution of community-orientation index scores within my sample. I then used confirmatory factor analysis to ensure that combining the five measures into a single community orientation composite index score is appropriate. I then used measurement invariance to ensure the latent construct, community orientation, remained consistent across time periods. Further details can be found in Chapter 5. Interrater Reliability I worked with two research assistants who also collected a subsample of data, and recoded their subsample to test the interrater reliability. This comprised approximately 11% (n=14 hospitals) of the final sample. Interrater reliability was excellent for data collected for Research Aim #2 and for the five sub-measures used to develop the community orientation composite index 31 measure, ranging from a very high level of agreement of 96% between coders (Measure 1: Targeting of DUHN communities) to a low of 92% agreement between coders (Measure 2: Prevention Level). Interrater reliability was less satisfactory for data collected for Research Aim #1. Calculating the level of agreement between coders was admittedly difficult due to the number of factors used to calculate each variable. For example, there were 15 categorical measures for stakeholders for each CHNA activity for a total of 75 measures. Where there were large numbers of stakeholders – as high as 72 in one CHNA – each stakeholder was assigned a level of effort. Moreover, overlooking or missing text that identified stakeholders – which happened for several of the CHNA reports that were coded twice – led to differences between coders across multiple measures. Level of agreement generally was as low as 62% for several Activity 3 measures-related measures. However, despite a fair amount of disagreement, I calculated all network-related measures using both sets of data (data collected by me and data collected by the two other coders) and found that all measures for each NPH were well within a standard deviation of each other. I also used both sets of data to complete Research Aim #1 (as well as Research Aim #2) and found very little difference in results. Note that all data used for both of these research aims was produced by me. However, the poor interrater reliability scoring for this research measure does suggest a need for clearer criteria and definitions to help guide data collection. 32 Chapter 4 - Developing a Typology of Network Structures Associated with Conducting a Community Health Needs Assessment This chapter describes the analysis used to identify and group different patterns of collaboration between NPHs and other stakeholders related to completing their CHNAs. As discussed in Chapter 3, by clustering the various network models chosen by NPHs, we can better understand and compare (1) how NPHs have organized themselves and/or engaged other organizations in response to the expanded community benefit regulations; (2) if NPHs have changed their organizational structures to be more or less collaborative over time; and (3) if certain network models that, for example, are highly integrative of community organizations and/or delegate significant responsibility to community stakeholders subsequently result in more community-oriented in their implementation strategies (see the next chapter for further discussion). Using data collected via the methods described in Chapter 3, I first examine the distribution of the three network measures: integration, centralization, and homophilization. I then discuss using cluster analysis to group the available data into homogenous groups that are consistent across all three time periods. I then discuss the characteristics of each grouping and whether certain types of hospitals were more likely to pursue a certain strategy in how they conducted their CHNA. 33 Examination of Three Network Measures As discussed in Chapter 3, the three measures use data collected on the range of five CHNA-related activities: (1) Plan and define the CHNA; (2) Collect and analyze data; (3) Identify priority health issues; (4) Communicate results; and (5) Use CHNA results to plan implementation strategies. The three measures are defined as:  Integration: the average proportion of activities performed by each category of stakeholder. Please note that I have consolidated categories of stakeholders into fifteen separate categories.  Centralization: the average level of effort put forth by the reporting hospital over each of the five CHNA-related activities.  Homophilization: the proportion of clinical partners to all partners, averaged over the five activities These three measures were used to characterize different facets of the organizational structures put in place by NPHs in response to the expanded community benefit regulations. Prior to conducting cluster analysis, I analyzed the distribution of the three measures and how they have changed over the course of the three CHNA reporting periods. Figure 4.1 presents a box plot distribution for each of the three measures. As evidenced by the box plots, the mean changed very little over the three cycles for all three measures. The lower tail of the integration distribution noticeably shifted up in the 3rd cycle, though, which would suggest all hospitals in the sample incorporated at least some 34 other stakeholder group into the CHNA process. Correspondingly, the lower tail of the homophilization measure also shifted away from zero in the third CHNA cycle. Figure 4.1 Box Plot Distribution of Three CHNA Network Measures Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 Tables 4.1 through 4.3 also show how the three measures have varied over time in more detail. Two integration-related parameters are presented in Table 4.1. The first - the proportion of CHNAs that include contributions from each category of stakeholder - demonstrates which organizations are more likely to have played an active role in completing CHNAs over the three reporting cycles. In general, participation increased for all types of organizations from the baseline reporting period to the subsequent two 0 .2 .4 .6 .8 1 In te gr at io n 1 2 3 CHNA Reporting Cycle 0 .2 .4 .6 .8 1 C en tra liz at io n 1 2 3 CHNA Reporting Cycle 0 .2 .4 .6 .8 1 H om op hi liz at io n 1 2 3 CHNA Reporting Cycle 35 reporting periods. Two organizational categories saw a significant increase in participation between the first and third CHNA cycles: local health agencies (LHAs) and non-profit organizations and foundations. Other hospitals also participated in a majority of CHNAs for all three cycles. 36 Table 4.1 Measures of Integration Across Three CHNA Reporting Cycles (1) * = Difference is significant from 1st CHNA reporting at p < 0.05; (2) CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 CHNA Report Cycle 1 CHNA Report Cycle 2 CHNA Report Cycle 3 Variable Mean/Pct (SE) Mean/Pct (SE) Mean/Pct (SE) Integration: Proportion of CHNAs that include contributions from the following organizations A. Other hospitals 0.63 (0.04) 0.67 (0.04) 0.70 (0.04) B. Primary care orgs. 0.38 (0.04) 0.45 (0.04) 0.46 (0.04) C. Other provider orgs. 0.39 (0.04) 0.47 (0.04) 0.48 (0.04) D. Insurers/managed care orgs. 0.10 (0.03) 0.16 (0.03) 0.16 (0.03) E. State health agencies 0.18 (0.03) 0.22 (0.04) 0.18 (0.03) F. Local health agencies 0.58 (0.04) 0.67 (0.04) 0.71 (0.04)* G. Other state agencies 0.06 (0.02) 0.07 (0.02) 0.06 (0.02) H. Other local government 0.42 (0.04) 0.42 (0.04) 0.52 (0.04) I. Other govt. (federal/tribal) 0.24 (0.04) 0.28 (0.04) 0.26 (0.04) J. Faith orgs 0.20 (0.03) 0.24 (0.04) 0.19 (0.04) K. Schools 0.37 (0.04) 0.38 (0.04) 0.38 (0.04) L. Universities 0.41 (0.04) 0.44 (0.04) 0.46 (0.04) M. Non-profit organizations/foundations 0.62 (0.04) 0.68 (0.04) 0.75 (0.03)* N. Private consultants 0.44 (0.04) 0.49 (0.04) 0.46 (0.04) O. Others 0.50 (0.04) 0.47 (0.04) 0.54 (0.04) Integration: Average proportion of CHNA activities contributed by the following organizations when part of CHNA network A. Other hospitals 0.73 (0.03) 0.84 (0.03) 0.83 (0.03) B. Primary care orgs. 0.33 (0.03) 0.42 (0.03) 0.41 (0.03) C. Other provider orgs. 0.35 (0.03) 0.39 (0.03) 0.37 (0.03) D. Insurers/managed care orgs. 0.37 (0.06) 0.43 (0.06) 0.56 (0.08) E. State health agencies 0.43 (0.06) 0.49 (0.05) 0.56 (0.06) F. Local health agencies 0.54 (0.04) 0.59 (0.03) 0.58 (0.03) G. Other state agencies 0.29 (0.04) 0.31 (0.07) 0.34 (0.11) H. Other local government 0.37 (0.03) 0.37 (0.03) 0.39 (0.02) I. Other govt. (federal/tribal) 0.34 (0.04) 0.31 (0.03) 0.30 (0.03) J. Faith orgs 0.33 (0.05) 0.33 (0.04) 0.33 (0.04) K. Schools 0.34 (0.03) 0.36 (0.03) 0.36 (0.03) L. Universities 0.49 (0.04) 0.47 (0.04) 0.54 (0.03) M. Non-profit organizations/foundations 0.44 (0.03) 0.45 (0.03) 0.47 (0.03) N. Private consultants 0.53 (0.03) 0.51 (0.02) 0.52 (0.02) O. Others 0.30 (0.02) 0.27 (0.02) 0.27 (0.02) 37 The significant increase in the proportion of CHNAs that incorporate LHAs is expected. Like NPHs, local public health agencies are expected to regularly conduct CHNAs, as it has been a requirement for accreditation from the Public Health Accreditation Board, a private national organization dedicated to improving the delivery of public health services, since 2011.50 NPHs and LHAs are thus frequent collaborators on CHNAs when residing in a shared community - especially in the most recent CHNA cycle - as both groups of stakeholders realize the potential benefits associated with completing their CHNAs together. Some states, such as New York and Ohio, also now require LHAs and NPHs to conduct their CHNAs together.37 The second parameter – the mean proportion of CHNA activities with contributions from each stakeholder category when participating in CHA – remained consistent across all three reporting cycles. The highest level of integration into the CHNA process came from other hospitals, followed by local health agencies. On average, private consultants also consistently contributed on half of all CHNA activities. Several stakeholder groups saw their participation increase substantially from the baseline reporting cycle, including insurers / managed care organizations, state health agencies, and universities. Again, these observations are expected. Most organization types are minimally integrated into the CHNA process unless there is potential for a more significant role; more likely, these organizations will serve in an advisory role on 1-2 CHNA activities. Other organizations that are more integrated into the CHNA process either have needed knowledge or capacity that goes beyond an advisory role (e.g., private consultants; universities) and/or are also incentivized to be a part of a CHNA process (e.g., other 38 hospitals; LHAs). For the latter group, at a minimum, these organizations will likely provide some measure of oversight over all shared CHNA-related activities, which would explain their higher integration measurement. It is also not unexpected to observe non-profit organizations and foundations contributing significantly more over the three time periods while also not contributing on a majority of CHNA activities. As mentioned previously, the expanded community benefit regulations that that were part of the ACA and then finalized in December 2014 required NPHs to solicit input from both public health practitioners and representatives of underserved communities as they decide what health issues to prioritize.9 Those nonprofit organizations that are integrated into the CHNA process, then, are likely serving an advisory role on one or more activities. Table 4.2 Measures of Centralization Across Three CHNA Reporting Cycles CHNA Report Cycle 1 CHNA Report Cycle 2 CHNA Report Cycle 3 Variable Mean (SE) Mean (SE) Mean (SE) Centralization: Mean level of effort provided by reporting hospital over five CHNA activities Activity 1: Plan and define the CHNA 0.35 (0.03) 0.37 (0.03) 0.34 (0.03) Activity 2: Collect and analyze data 0.24 (0.02) 0.24 (0.02) 0.20 (0.02) Activity 3: Identify priority health issues 0.29 (0.03) 0.28 (0.03) 0.25 (0.02) Activity 4: Communicate results 0.40 (0.03) 0.38 (0.03) 0.34 (0.03) Activity 5: Use CHNA results to plan implementation strategies 0.64 (0.03) 0.56 (0.03) 0.58 (0.03) Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 39 Tables 4.2 and 4.3 show the detailed distribution of the centralization measure and homophilization measure across the five CHNA activities over the three reporting cycles. Both measures are fairly consistent across the three time periods. For the centralization measure, only Activity 4 (communicate results) and Activity 5 (use CHNA results to plan implementation strategy) show a decrease of 0.06 from the baseline to the third reporting period. Activity 5 is the activity with the highest level of effort from the reporting hospital and the highest level of homophilization, which illustrates how hospitals will assume a measure of autonomy over planning their own implementation strategy, regardless of how collaborative they are on other activities. Although slight, the decrease in centralization and homophilization for Activity 5 does seem to suggest that NPHs have been slightly more likely to incorporate external community organizations into the planning process. The two activities with both the lowest centralization and homophilization levels are Activities 2 (collect and analyze data) and 3 (identify priority health issues). Again, Activity 3 requires input from external organizations to complete, which would explain why it has lower centralization and homophilization measurement. Activity 2 is likely one that requires a significant amount of overall effort to complete, in addition to requiring specialized knowledge of available population-level data (e.g., for secondary data sources) and/or data collection practices if collecting primary data directly. It is predictable, then, that hospitals might utilize external resources to complete this activity, by hiring an organization that has the capacity and ability to organize and complete this step – from, for example, universities and/or private consulting firms. 40 Table 4.3 Measures of Homophilization Across Three CHNA Reporting Cycles CHNA Report Cycle 1 CHNA Report Cycle 2 CHNA Report Cycle 3 Variable Mean (SE) Mean (SE) Mean (SE) Homophilization: Mean proportion of clinical stakeholders to all stakeholders over five CHNA activities Activity 1: Plan and define the CHNA 0.69 (0.02) 0.69 (0.02) 0.67 (0.03) Activity 2: Collect and analyze data 0.54 (0.03) 0.53 (0.02) 0.50 (0.02) Activity 3: Identify priority health issues 0.44 (0.02) 0.43 (0.02) 0.40 (0.02) Activity 4: Communicate results 0.64 (0.03) 0.66 (0.02) 0.64 (0.02) Activity 5: Use CHNA results to plan implementation strategies 0.79 (0.03) 0.74 (0.03) 0.73 (0.03) Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 Cluster Analysis To place each CHNA-developed network into a mutually exclusive group and establish an empirical typology of NPH-centered networks, I conducted hierarchical cluster analysis using the three structural measures. I first standardized the three measures into z-scores and then grouped each network into a distinct cluster using different similarity algorithms. The Ward method of assigning networks with a squared Euclidean distance specification provided a good fit with the data. I also visually inspected the resulting dendogram to confirm that a five-cluster solution provided a parsimonious but conceptually well-defined solution. This approach was used first on the 41 baseline reporting cycle and then on the two subsequent reporting cycles to confirm that the five-cluster solution remained consistent longitudinally. To further confirm the five-cluster model, I also used the Calinski–Harabasz method for determining the appropriate number of clusters.51 A five-cluster solution consistently had the largest pseudo-F value relative to a four-group and six-group solution, suggesting the five-cluster solution is the most distinct. Table 4.4 provides the resulting mean values for all three measures delineated for all five different clusters over the three reporting cycles and the entire sample. At first glance, two pairs of network configurations (Clusters 1 and 2; Clusters 3 and 4) are differentiated primarily by their level of homophilization, with similar mean integration and centralization measurement. A third configuration (Cluster 5) is entirely distinct. As I will further discuss, the underlying organizational structures for each cluster are conceptually distinct and signal different approaches for completing CHNAs that both balances an NPH’s need to control some or all of the CHNA process given internal and external resources, while also delegating a proportion of the work to external organizations. Figure 4.2 displays the frequency of each grouping across the three time periods. As is evident, Cluster 2 was the grouping with the highest frequency across all three time periods, while Clusters 3, 4, and 5 each report the smallest membership for one reporting cycle. Cluster 5 also saw its membership clearly decline over the three time periods, while the other four clusters remained consistent or varied a little. Finally, Table 4.5 through 4.7 also breaks down the key factors that describe the membership of each grouping. 42 Table 4.4 Mean Integration, Centralization & Homophilization Values for Five CHNA Network Configurations CHNA Report Cycle 1 CHNA Report Cycle 2 CHNA Report Cycle 3 Overall Cluster 1: Medium integration; Medium to low centralization; High homophilization Mean (SE) Mean (SE) Mean (SE) Mean (SE) Integration 0.42 (0.02) 0.42 (0.02) 0.43 (0.02) 0.42 (0.01) Centralization 0.25 (0.03) 0.30 (0.03) 0.27 (0.03) 0.27 (0.02) Homophilization 0.80 (0.02) 0.78 (0.02) 0.76 (0.01) 0.78 (0.01) Cluster 2: Medium integration; Medium to low centralization; Medium homophilization Integration 0.39 (0.01) 0.41 (0.01) 0.41 (0.01) 0.40 (0.01) Centralization 0.38 (0.02) 0.30 (0.03) 0.34 (0.03) 0.34 (0.01) Homophilization 0.49 (0.02) 0.50 (0.02) 0.52 (0.01) 0.50 (0.01) Cluster 3: High integration; low centralization; medium- to medium-low homophilization Integration 0.72 (0.03) 0.71 (0.02) 0.69 (0.03) 0.71 (0.02) Centralization 0.17 (0.04) 0.13 (0.03) 0.11 (0.03) 0.14 (0.02) Homophilization 0.36 (0.03) 0.39 (0.03) 0.37 (0.03) 0.37 (0.02) Cluster 4: High integration; medium centralization; medium- to medium high homophilization Integration 0.73 (0.04) 0.68 (0.02) 0.70 (0.03) 0.70 (0.02) Centralization 0.39 (0.05) 0.40 (0.04) 0.43 (0.05) 0.41 (0.02) Homophilization 0.65 (0.05) 0.61 (0.03) 0.60 (0.04) 0.62 (0.02) Cluster 5: Low integration; high centralization; high homophilization Integration 0.24 (0.03) 0.25 (0.03) 0.27 (0.02) 0.25 (0.02) Centralization 0.79 (0.03) 0.81 (0.03) 0.78 (0.03) 0.79 (0.02) Homophilization 0.83 (0.02) 0.84 (0.02) 0.77 (0.04) 0.82 (0.01) Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 43 Figure 4.2 Five Cluster-Model Membership Frequency over 3 CHNA Reporting Periods Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 Clusters 1 and 2 had very similar levels of moderate to low levels of integration and centralization but differed significantly in terms of their respective levels of homophilization, with Cluster 1 showing significantly higher levels of clinical stakeholders relative to Cluster 2. These two clusters also differed substantially in the types of NPHs that comprised their groups. Urban hospitals, hospitals from a healthcare system, and medium/large hospitals were much more likely to be in Cluster 1, while smaller, stand-alone hospitals from rural areas were much more likely to be in Cluster 2. 0 5 10 15 20 25 30 35 40 45 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 N um be r o f C H N A N et w or ks CHNA Report Cycle 1 CHNA Report Cycle 2 CHNA Report Cycle 3 44 What unites NPHs from Clusters 1 and 2 is an organization external to the reporting hospital that is highly integrated into the CHNA process, i.e., that does a significant amount of the work necessary for completing a CHNA. Table 4.5 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 1 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Region Northeast 0.27 (6) 0.36 (8) 0.18 (4) 0.09 (2) 0.09 (2) South 0.22 (8) 0.33 (12) 0.19 (7) 0.08 (3) 0.17 (6) Midwest 0.20 (8) 0.34 (14) 0.17 (7) 0.10 (4) 0.20 (8) West 0.31 (8) 0.27 (7) 0.00 (0) 0.27 (7) 0.15 (4) Size Small (<100 beds) 0.18 (10) 0.43 (24) 0.13 (7) 0.07 (4) 0.20 (11) Medium (101-299 beds) 0.31 (12) 0.26 (10) 0.13 (5) 0.18 (7) 0.13 (5) Large (300+ beds) 0.27 (8) 0.23 (7) 0.20 (6) 0.17 (5) 0.13 (4) Ownership Stand-alone 0.00 (0) 0.57 (13) 0.13 (3) 0.04 (1) 0.26 (6) Health System 0.29 (30) 0.27 (28) 0.15 (15) 0.15 (15) 0.14 (14) Medicare APM Participant 0.31 (29) 0.28 (26) 0.15 (14) 0.14 (13) 0.12 (11) Non-Participant 0.03 (1) 0.47 (15) 0.13 (5) 0.09 (3) 0.28 (9) Environment Urban 0.28 (22) 0.28 (22) 0.15 (12) 0.13 (10) 0.15 (12) Rural 0.17 (8) 0.40 (19) 0.12 (6) 0.13 (6) 0.17 (8) Note: CHNA Cycle 1 includes the years 2011 through 2014 Table 4.6 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 2 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Region Northeast 0.23 (5) 0.23 (5) 0.36 (8) 0.04 (1) 0.14 (3) South 0.14 (5) 0.39 (14) 0.08 (3) 0.28 (10) 0.11 (4) Midwest 0.24 (10) 0.24 (10) 0.15 (6) 0.20 (8) 0.17 (7) West 0.23 (6) 0.46 (12) 0.00 (0) 0.19 (5) 0.12 (3) Size Small (<100 beds) 0.18 (10) 0.35 (20) 0.13 (7) 0.18 (10) 0.16 (9) Medium (101-299 beds) 0.23 (9) 0.31 (7) 0.10 (4) 0.21 (8) 0.15 (6) Large (300+ beds) 0.23 (7) 0.30 (9) 0.20 (6) 0.20 (6) 0.07 (2) 45 Ownership Stand-alone 0.04 (1) 0.43 (10) 0.13 (3) 0.22 (5) 0.17 (4) Health System 0.25 (25) 0.30 (31) 0.14 (14) 0.19 (19) 0.13 (13) Medicare APM Participant 0.27 (25) 0.27 (25) 0.14 (13) 0.19 (18) 0.13 (12) Non-Participant 0.03 (1) 0.50 (16) 0.13 (4) 0.19 (6) 0.16 (5) Environment Urban 0.22 (17) 0.35 (27) 0.10 (8) 0.22 (17) 0.12 (9) Rural 0.19 (9) 0.30 (14) 0.12 (9) 0.15 (7) 0.17 (8) Note: CHNA Cycle 2 includes 2015 through 2017 Table 4.7 Distribution of Clusters Across Different Hospital- and Community-Level Characteristics in CHNA Report Cycle 3 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Region Northeast 0.09 (2) 0.45 (10) 0.36 (8) 0.05 (1) 0.05 (1) South 0.28 (10) 0.36 (13) 0.08 (3) 0.22 (8) 0.06 (2) Midwest 0.22 (9) 0.22 (9) 0.15 (6) 0.24 (10) 0.17 (7) West 0.31 (8) 0.35 (9) 0.15 (4) 0.12 (3) 0.08 (2) Size Small (<100 beds) 0.20 (11) 0.41 (23) 0.14 (8) 0.14 (8) 0.11 (7) Medium (101-299 beds) 0.26 (10) 0.21 (8) 0.21 (8) 0.23 (9) 0.10 (4) Large (300+ beds) 0.27 (8) 0.33 (10) 0.17 (5) 0.17 (5) 0.07 (2) Ownership Stand-alone 0.04 (1) 0.57 (13) 0.13 (3) 0.17 (4) 0.09 (2) Health System 0.28 (28) 0.27 (28) 0.18 (18) 0.18 (18) 0.10 (10) Medicare APM Participant 0.30 (28) 0.27 (25) 0.16 (15) 0.17 (16) 0.10 (9) Non-Participant 0.03 (1) 0.50 (16) 0.19 (6) 0.19 (6) 0.09 (3) Environment Urban 0.28 (22) 0.31 (24) 0.14 (11) 0.18 (14) 0.09 (7) Rural 0.15 (7) 0.36 (17) 0.21 (10) 0.17 (8) 0.10 (5) Note: CHNA Cycle 3 includes the years 2018 through 2021 For those NPHs in Cluster 1, this external entity is often other hospitals – for example, if a reporting hospital is a part of a health system where hospital administrators from the flagship hospital are responsible for the majority of the work, or where an NPH is one of 46 several hospitals completing a CHNA for a shared geographic area. For NPHs in Cluster 2, the external organization is more likely to be, for example, a private consultant or a local health agency. In either case, a reporting NPH is likely still providing a moderate amount of input and will assume primary responsibility over at least one activity – most often, the activities where a reporting hospital might feel more comfortable controlling to ensure their CHNA process is compliant and/or where local control might feel more appropriate, i.e., Activities 1, 4, and 5, the planning or reporting-related activities. So, for example, a health system will conduct most of a CHNA for their member hospitals (or in the case of hospitals from Cluster 2, a consultant or LHA), but allow each reporting hospital to assume control over using the CHNA results to plan their implementation strategy. Both Clusters 1 and 2 also engage similar numbers of external stakeholder groups as part of their CHNA process, but – unless they have primary responsibility for completing some or all of the CHNA activities in Cluster 2 – tend to only have limited involvement in 1-2 activities, mostly Activity 3. The next model of CHNA collaboration, Cluster 3, is characterized by a highly decentralized CHNA process that is also highly integrative and with lower levels of homophilization. Geographically, NPHs from the Northeast were more likely to pursue this model of collaboration, especially in the last two reporting cycles - but otherwise, very little distinguished NPHs that chose this model. For many of the CHNAs being conducted using this model, a key participating stakeholder was the local health agency, who tended to be highly integrated into the process. Another common attribute for the CHNAs grouped into this cluster was the presence of an advisory committee that participated in all CHNA-related activities. Such committees were common in other 47 clusters as well, but did not have the same scope that those in Cluster 3 tended to have, such as participating in the implementation strategy planning process. Cluster 4 is characterized by a highly integrative CHNA process with moderate to moderately high levels of centralization and homophilization. Like Cluster 3, characteristics of NPHs following this model of CHNA collaboration are less distinctive than with Clusters 1 and 2; the first CHNA reporting cycle saw a large majority of Western hospitals use this model, but become less regionally distinct in subsequent reporting cycles. NPHs in this cluster also tended to engage smaller numbers of stakeholders – on average, CHNAs in this cluster only engaged four different stakeholder groups, while clusters 1-3 engaged about eight categories and cluster 5, five. However, those organizations that participated in the CHNA process tended to be highly integrated into the process, with all organizations sharing responsibility across the five activities. The organizations that reporting hospitals tended to integrate also were likely to be clinical in nature. The final Cluster 5 is highly centralized, highly homogenous, and with very low levels of integration of external organizations. For this model of CHNA, the reporting hospital was responsible for the majority of work on all activities. Like Cluster 4, collaboration with other organizations was limited and mostly confined to Activity 3. While the characteristics of NPHs in this cluster varied and were largely undistinguished like Clusters 3 and 4, it is clear that frequency with which NPHs pursued this model decreased over the three reporting cycles. This is understandable given the level of responsibility placed on NPHs to pursue this model. 48 Table 4.8 Transition Probabilities from CHNA Cycle 1 to CHNA Cycle 2 Table 4.9 Transition Probabilities from CHNA Cycle 2 to CHNA Cycle 3 CYCLE 3 C Y C L E 2 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 1 0.58 0.15 0.15 0.12 - Cluster 2 0.12 0.59 0.12 0.12 0.05 Cluster 3 0.18 0.18 0.47 0.18 - Cluster 4 0.25 0.17 0.13 0.33 0.13 Cluster 5 - 0.35 0.06 0.18 0.41 Transitions Between Clusters While the number of CHNAs in each cluster did not vary tremendously for each reporting cycle, there was still a moderate amount of transitioning between clusters. Tables 4.8 and 4.9 show the probability of transitioning from cluster to cluster between the 1st and 2nd cycles and the 2nd and 3rd cycles, respectively. While the most likely outcome was for a CHNA network to remain grouped within the same cluster after each subsequent cycle, there were still a reasonably good chance of being grouped in a different cluster, e.g., the probability of transitioning from Cluster 5 to Cluster 2 was 0.3 and 0.35 between the first two cycles and the last two cycles, respectively. CYCLE 2 C Y C L E 1 Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 1 0.43 0.20 0.13 0.20 0.03 Cluster 2 0.17 0.49 0.07 0.22 0.05 Cluster 3 - 0.22 0.56 0.17 0.05 Cluster 4 0.25 0.31 - 0.31 0.13 Cluster 5 0.10 0.30 - 0.05 0.55 49 Discussion of Results This chapter has detailed the approach I used to classify the different CHNA- related networks that NPHs have utilized to complete their CHNA requirement. As detailed above, I found that a five-cluster model fit the data well across all three CHNA reporting cycles and also provided thematically unique groupings. To better characterize what each group looks like in practice, I found example CHNA reports from each cluster that were closest to the mean integration, centralization, and homophilization for that cluster. A brief description of both the reporting hospital as well as the adopted assessment process follows: Cluster 1 (Medium integration; Medium to low centralization; High homophilization): The reporting hospital for this CHNA from the first reporting cycle is a small critical access hospital (CAH) located in western Wisconsin, and is part of a large regional health system. The health system hired a private consulting firm to assist them with data collection and analysis, solicit input from external organizations to help identify priority health issues, and write up a report for a regionwide CHNA for each member hospital. As part of that process, a leadership committee was formed comprised of representation from each hospital, including the reporting hospital, as well as from the administrative office of the health system. This leadership committee had primary responsibility over planning the CHNA as well as deciding which health issues were the most pressing for each member hospitals’ subregion after weighing input from representatives of different social service agencies and analysis conducted on available primary and secondary sources. The 50 reporting hospital then planned their own implementation strategy using the results of the assessment. The process adopted by the health system and its reporting hospitals is one in which the clinical stakeholders involved in the CHNA’s production controlled the key planning-related activities as well as the analysis related to choosing which health issues were priorities. Some control was ceded to the consulting firm, as least as it pertains to the data collection and analysis and the authorship of the report, but otherwise, all processes were controlled internally. Cluster #2 (Medium integration; Medium to low centralization; Medium homophilization): This small CAH in rural Pennsylvania became affiliated with a regional health network in 2015; however, its first reporting cycle CHNA was conducted when it was a stand-alone hospital in 2012. This first cycle CHNA was co-led by the reporting hospital and the local county health agency; a number of other community and clinical organizations also helped plan the CHNA. While much of the CHNA was conducted with the local health agency, the actual implementation strategy was still largely planned by the reporting hospital. While much of the responsibility for completing the CHNA was shared with the local health agency, the reporting hospital still assumed primary responsibility over some key activities. As a result, the network formed by the reporting hospital had a similar level of integration and centralization relative to those from Cluster 1; however, because the local health agency assumed a significant amount of responsibility, the level of homophilization was much less. 51 Cluster #3 (High integration; low centralization; medium- to medium-low homophilization): This hospital is located outside of Boston, and is part of a large regional health system. The third cycle CHNA for this reporting hospital was largely led by the health system administrators, with input from member hospitals, and with some of the data collection and analysis handled by a private consultant. It also incorporated the input of a “Community Benefits Advisory Committee” comprised almost entirely of community organizations, who advised and/or led on all five CHNA activities, including the planning of the implementation strategy, but who also regularly oversaw the hospital’s community benefit efforts. The identification of the priority health issues was also decided by this committee in collaboration with the hospitals and health system. A large (20+) number of community organizations also provided a significant amount of input into this identification process as well. The reporting hospital’s responsibilities were thus fairly limited for this CHNA process, with the exception of the planning of the implementation strategy. Every CHNA activity involved 15 stakeholders at a minimum, many of them from community-based organizations, who also seemed to wield significant influence over the process. As a result, this CHNA process was highly integrated and decentralized, as well as heterogenous in the types of organizations involved. Cluster #4 (High integration; medium centralization; medium- to medium high homophilization): This hospital is located outside of Austin, Texas and is part of a large national health system. For this hospital’s third cycle CHNA, responsibilities for 52 conducting the CHNA were shared with a nearby hospital that was also part of the same health system. Some external oversight was provided by an advisory committee comprised of a mix of clinical and community-based organizations. This committee was responsible for helping the two hospitals identify priority health issues, as well as advise on data collection processes and the planning of the implementation strategy; the two hospitals completed everything else. Because both the advisory committee and the other hospital were highly involved in at least some capacity for most of the CHNA activities, the level of integration involved in completing the CHNA was high. Because this advisory committee was comprised of a mix of clinical and community representation, homophilization was likewise moderate. The process was also moderately centralized because primary responsibility for completing the CHNA was shared with another hospital. Cluster #5 (Low integration; high centralization; high homophilization): This CAH is located in a remote part of South Dakota and is part of a large regional health care system. For this reporting hospital’s first cycle CHNA, much of the work completing the CHNA was done by hospital staff, with the only input from community-based organizations limited to helping identify community health issues. Primary responsibility for completing each CHNA activity remained with the reporting hospital, with no other input from other organizations on the majority of activities. This type of model is one in which hospitals maintain total control of the CHNA process. Hospitals that use this framework for completing their CHNA are likely 53 satisficing with the federal community benefit regulations that require input from certain types of stakeholders, by limiting their interactions with external organizations to determining which health issues to prioritize. In the case of this South Dakota hospital, it may be that there is a lack of resources, both internally (e.g., limited funding) and externally (e.g., a lack of available organizations with which they can collaborate such as a local health agency) for the hospital to pursue a more collaborative framework. Conclusions As the discussion of these example reporting hospitals and their CHNAs makes clear, each cluster essentially falls on a continuum between controlling the CHNA process internally and dispersing responsibility to community organizations. In between these two points, hospitals will choose how ‘community-based’ their network of collaborators will be, with those hospitals in Clusters 1 and 4 relying primarily on their peer hospitals, health system administrators and other clinical stakeholders to comprise a majority of their collaborators (and those in Cluster 5 only relying only on themselves) and those in Clusters 2 and 3 relying more on community stakeholders, especially local health agencies. As might also be apparent, there are a number of external factors that clearly influence and/or limit how much autonomy a reporting hospital has in choosing their CHNA process. Those hospitals that are part of a regional health system, for example, are beholden to how their system leadership decides to conduct their CHNA – does the system allow the hospital to pursue their own assessment process autonomously, or is much of the work done by administrators at the system level? Likewise, having a highly 54 integrated, community-based CHNA process also necessitates there being sufficient local organizations with the capacity and willingness to participate. In a rural community, these conditions might not always present, especially when public health and/or social service infrastructure is not well-resourced. These external factors might also change between reporting cycles, which is likely what has led to hospitals changing how they conduct their CHNAs (and changing their cluster membership). These external factors also extend to relevant policy - in particular state or local policies pertaining to how NPHs should conduct their CHNAs. For example, New York and Ohio both have implemented regulations that require or highly encourage NPHs to work with their LHAs to conduct their CHNAs in 2015 and 2017, respectively.37 As such, these states’ NPHs tend to have less centralized, more integrated, and more heterogenous CHNA networks and thus, are more likely to be grouped into Clusters 2 or 3. Another state with a unique approach to conducting CHNAs is Maine, whose state health department, together with the four largest healthcare systems in the state, completes most of the assessment work themselves, but leaves communicating results and planning the implementation strategy to each reporting hospital.37 NPHs are expected to still have primary responsibility over several of the key CHNA activities, but an external agency is responsible for completing the rest. As a result, Maine hospitals (at least in my sample) are all entirely grouped into Cluster 2. Having established a method for classifying NPHs into groups based on how they have organized themselves in response to the expanded federal community benefit regulations, it will be much more straightforward determining if these different organizational structures are associated with potential community social capital-related 55 changes. For example, have the federal regulations requiring hospitals to assess their communities’ health status and connect with local organizations led to any changes in how hospitals and communities interact over the past ten years? For hospitals and community organizations that have developed strong relationships via their CHNAs: are they more likely to work together and develop community-oriented strategies to address their identified health issues? Are there potential positive (or negative) externalities that rely on strong hospital-community relationships, such as improvements in care transitions for patients in need of social services to manage their health? Or changes to norms governing local health-seeking behaviors due to the presence of a more visible local hospital that is better integrated with the operations of community organizations? These questions will be further explored in the next chapter, using this framework to compare and contrast these different underlying organizational CHNA networks and their associated outcomes. 56 Chapter 5 - Development of Community Orientation Composite Measure This chapter details the chosen set of analytic approaches used to develop and test a set of measures meant to approximate the latent variable, community orientation, for each implementation strategy within my sample. After using the data collection methods detailed in Chapter 3, I first examine the distribution of the five measures associated with community orientation. I then discuss the use of confirmatory factor analysis to ensure that combining the five measures into a single community orientation composite index score is appropriate. I then use measurement invariance to ensure the latent construct, community orientation, remains consistent across time periods. Finally, I discuss using cluster analysis to divide the implementation strategies into similar groups – again using the five community orientation measures - and characterize each group, offering example implementation strategies as case studies. Examination of Five Community Orientation Measures As discussed in Chapter 3, the five measures I have developed to approximate the community orientation of a given NPH’s implementation strategy are: (1) emphasis on prevention; (2) targeting of communities with disproportionate unmet health needs (DUHNs); (3) maintenance of community and clinical continuum; (4) building upon existing community infrastructure; and (5) use of community stakeholders in planning of 57 the implementation strategy. A more detailed discussion for how these five measures were constructed was discussed in Chapter 3. Figure 5.1 displays a set of box plots for these five measures and how they have changed over the course of the three CHNA cycles. As evidenced by each set of box plots, all five measures have seen their means increase in value over the course of the three CHNA cycles. Figure 5.1 Box Plot Distribution of 5 Community Orientation Measures Over 3 CHNA Cycles Note: (1) CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021; (2) DUHN = Disproportionate Unmet Health Needs Four measures – targeting of DUHN communities, community / clinical continuum, use of community infrastructure, and community involvement in planning of the 0 .2 .4 .6 .8 1 Sc or e 1 2 3 CHNA Report Cycle Prevention Level 0 .2 .4 .6 .8 1 Sc or e 1 2 3 CHNA Report Cycle Targeting of DUHN Communities 0 .2 .4 .6 .8 1 Sc or e 1 2 3 CHNA Report Cycle Community + Clinical Continuum 0 .2 .4 .6 .8 1 Sc or e 1 2 3 CHNA Report Cycle Use of Community Infrastructure 0 .2 .4 .6 .8 1 Sc or e 1 2 3 CHNA Report Cycle Community Involvement in Planning 58 implementation strategy - were skewed towards the lower end of their distributions. The fifth measure – prevention level – was less skewed and had a higher mean over the course of the three CHNA cycles. In addition, four of the measures – prevention level, targeting of DUHN communities, community/clinical continuum, and use of community infrastructure – have outliers that affect the kurtosis of each measure’s distribution. Table 5.1 further illustrates how the mean and standard deviation of each measure change over the three time periods. Table 5.1 Summary Statistics of 5 Community Orientation Measures Over 3 CHNA Cycles Mean Standard Deviation Skew Kurtosis Prevention Level Cycle 1 (n=125) 0.595 0.132 -0.887 4.729 Cycle 2 (n=125) 0.600 0.151 -1.271 6.031 Cycle 3 (n=125) 0.641 0.140 -0.768 4.648 Targeting of DUHN Communities Cycle 1 0.215 0.222 1.333 5.001 Cycle 2 0.237 0.238 1.260 4.394 Cycle 3 0.287 0.282 1.083 3.430 Community/Clinical Continuum Cycle 1 0.183 0.221 1.922 7.252 Cycle 2 0.195 0.202 1.157 3.786 Cycle 3 0.248 0.208 0.801 3.095 Use of Community Infrastructure Cycle 1 0.117 0.149 1.336 4.182 Cycle 2 0.175 0.185 0.826 2.714 Cycle 3 0.225 0.243 1.185 3.733 Community Involvement in Planning Cycle 1 0.201 0.300 0.956 2.160 Cycle 2 0.251 0.306 0.654 1.828 Cycle 3 0.272 0.324 0.539 1.550 Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 59 The distribution and variation of these measures suggests that, on average, NPHs tended to focus their implementation strategies on the higher end of the prevention scale, i.e., on average, they were more likely to use interventions that addressed non-health social determinants and/or primary prevention interventions. Despite the preference for preventative interventions, most NPHs used community stakeholders sparingly, both during the implementation strategy planning process and in the implementation of chosen activities. In practice, this means NPHs tended to prefer interventions where clinical staff educated the public-at-large on health prevention topics. Chosen interventions also tended to focus on targeting the community-at large, as opposed to DUHN communities. While examining the distribution of these five measures is illustrative for establishing some initial patterns contained within each measure, it is difficult to conceptualize how these measures relate to each other (i.e., how each measure varies with other measures) or how these measures vary given certain hospital and community characteristics. Thus, developing a composite index measure that aggregates these five measures into a single community-orientation measure, while also maximizing the variation that exists between variables, would be helpful for visualizing how community orientation varies between NPHs and over time. My hypothesis is that all five measures are affected by a single latent variable, community orientation, that describes an internal characteristic of each hospital, and that this characteristic is then reflected in how hospitals approach addressing their community’s most pressing health issues; thus, the most appropriate methodological approach is factor analysis. 60 Exploratory and Confirmatory Factor Analysis To determine whether it is appropriate to aggregate these measures into a single composite index score reflecting the latent construct of community orientation, I need to use either exploratory or confirmatory factor analysis. Exploratory factor analysis (EFA) is a statistical approach for determining the underlying data structure associated with a latent construct and identifying factors by maximizing the amount of explained variance. Typically, it is more appropriate to conduct EFA when little is known or hypothesized about the underlying data structure, which allows flexibility in the number of factors identified.52 Confirmatory factor analysis (CFA) is a companion methodological approach to EFA, in that it helps understand the shared variance of measured variables attributable to a latent construct, but it also requires you to hypothesize in advance, the number of factors, whether these factors are correlated, and the items that load on to which factors. CFA usually requires prior knowledge based on theory and/or empirical evidence.53–55 I opted to use CFA because I am hypothesizing a priori that the five measures all load well onto a single-factor model and these five measures together adequately reflect the latent construct community orientation. I am also using a previously defined framework in defining community orientation, which again, suggests the need to use CFA. Once I established factor loadings through EFA on the first CHNA cycle data, I then conducted CFA on the subsequent two CHNA cycles. Based on the results of the EFA, I hypothesized that the latent measure community orientation was driving much of the common variation between the five observed variables, and that the five measures would have consistent factor loadings on this single factor across the three time periods. 61 Prior to conducting CFA, I first examined whether the data supports conducting CFA – that is, determining whether there is a sufficient sample size; whether the multivariate distribution warrants a more robust approach; and whether there is sufficient internal consistency between variables.56 First, I calculated Chronbach’s alpha for the entire sample and each individual CHNA cycle to determine the extent to which all items measure the same construct. The use of alpha as a coefficient of internal consistency between items is not always reliable, as it is susceptible to, for example, the number of items being evaluated; however, it is an appropriate measure when assessing a single common factor as I am estimating.57,58 A value of alpha over 0.6 is usually considered sufficient for demonstrating composite reliability, especially when assessing a construct that has not been thoroughly evaluated. I found a value for alpha over 0.6 for the overall sample and for each cycle, except for the 1st cycle, which had an alpha of 0.46. While this is worrisome, it is important to consider this in light of other measures and features of the data – including how well a single factor model fits with the data.58 Table 5.2 Diagnostic Tests for Conducting Confirmatory Factor Analysis Test Statistic Cronbach’s Alpha Overall 0.60 1st Cycle 0.46 2nd Cycle 0.61 3rd Cycle 0.63 Bartlett’s Test of Sphericity Overall (Chi-Square) 293.1* 62 1st Cycle (Chi-Square) 48.7* 2nd Cycle (Chi-Square) 112.4* 3rd Cycle (Chi-Square) 133.3* Kaiser-Meyer-Olkin Test of Sampling Adequacy Overall 0.72 1st Cycle 0.63 2nd Cycle 0.71 3rd Cycle 0.68 Note: * = significant at α ≤ 0.05; CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 I also calculated another suggested measure before conducting CFA: Bartlett’s test of sphericity. This test assesses the extent to which the correlation matrix of the data to be factored differs significantly from the identity matrix.59 The null hypothesis of the test is that the variables are orthogonal, or not correlated and the alternative hypothesis is that they are correlated enough to be significantly different from the identify matrix.59 Because all Chi-Square values are significant, we can confirm that the variables are correlated and have a structure that can be identified through factor analysis. The sample size (n=125) is also likely sufficient for CFA. I calculated a Kaiser- Meyer-Olkin (KMO) test for each subset of CHNA cycle data, which found at least a ‘mediocre’ sample size for the all three CHNA cycles (Cycle 1 = 0.63; Cycle 2 = 0.71; Cycle 3 = 0.68).60,61 The KMO test for the entire sample (KMO = 0.72) was rated slightly better (‘middling’ per Kaiser’s categorization).60,61 As the box plots from Figure 5.1 make clear, the five measures are also skewed and possibly non-normal with most of the measures having several outliers – all of which may be a concern.62 However, skewness and kurtosis statistics from Table 5.1 are within the acceptable range of 2 and 762, 63 respectively, apart from the kurtosis value for the community / clinical continuum measure in CHNA cycle 1. Table 5.3 Pairwise Correlation Matrix for First CHNA Reporting Cycle Pr ev en tio n L ev el T ar ge tin g of D U H N C om m un iti es C om m un ity /C l in ic al C on tin uu m U se o f C om m un ity In fr as tr uc tu re C om m un ity In vo lv em en t i n Pl an ni ng Prevention Level 1.000 Targeting of DUHN Communities 0.0002 1.000 Community/Clinical Continuum 0.2623* 0.2061* 1.000 Use of Community Infrastructure 0.1691 0.2606* 0.4224* 1.000 Community Involvement in Planning 0.1756 0.0548 0.1286 0.0922 1.000 * Significant at α = 0.05 Table 5.4 Pairwise Correlation Matrix for Second CHNA Reporting Cycle Pr ev en tio n L ev el T ar ge tin g of D U H N C om m un iti es C om m un ity /C l in ic al C on tin uu m U se o f C om m un ity In fr as tr uc tu re C om m un ity In vo lv em en t i n Pl an ni ng Prevention Level 1.000 Targeting of DUHN Communities 0.2579* 1.000 Community/Clinical Continuum 0.3454* 0.3716* 1.000 Use of Community Infrastructure 0.3320* 0.3522* 0.6271* 1.000 Community Involvement in Planning 0.2057* 0.0204 0.1710 0.2231* 1.000 * Significant at α = 0.05 64 Table 5.5 Pairwise Correlation Matrix for Third CHNA Reporting Cycle Pr ev en tio n L ev el T ar ge tin g of D U H N C om m un iti es C om m un ity /C l in ic al C on tin uu m U se o f C om m un ity In fr as tr uc tu re C om m un ity In vo lv em en t i n Pl an ni ng Prevention Level 1.000 Targeting of DUHN Communities 0.3300* 1.000 Community/Clinical Continuum 0.4466* 0.3168* 1.000 Use of Community Infrastructure 0.4132* 0.4389* 0.6392* 1.000 Community Involvement in Planning 0.0808 0.1513 0.1922* 0.0475 1.000 * Significant at α = 0.05 Table 5.3 through 5.5 shows the correlation matrices for each of the three time periods. As evidenced by the three correlation matrices, correlations between variables are much stronger 65 within the second and third CHNA cycles – in particular, the relationships between the prevention level and targeting of DUHN community variables are much more correlated relative to the first CHNA cycle. The planning variable also has the strongest relationships to the other four measures in the second and third CHNA cycles. The strongest bivariate relationship across all three time periods is the relationship between the community / clinical continuum and use of community infrastructure measures. Although it would appear to have a noisy data structure, there is nothing to suggest the five measures do not provide a reasonable data structure for conducting EFA. CFA and all diagnostic tests were conducted using Stata v14.2. For CFA, I used Stata’s sem command, specifying maximum likelihood for my estimation method. When assessing the single-factor model using the entire sample, I used a more robust estimation model by clustering errors around each hospital to account for the within unit variance. Figure 5.2 Five Measure Single-Factor Model for Community Orientation 66 Table 5.6 Standardized Factor Loadings for Community Orientation Measures Using Confirmatory Factor Analysis Note: (1) * = this model used clustered standard errors to adjust for using panel data; (2) CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 CHNA Cycle 1 CHNA Cycle 2 CHNA Cycle 3 Overall* Items Factor Loading (SE) R- squared Factor Loading (SE) R- squared Factor Loading (SE) R- squared Factor Loading (SE) R- squared Targeting of DUHN Communities 0.32 (0.11) 0.10 0.46 (0.08) 0.21 0.50 (0.08) 0.25 0.46 (0.06) 0.22 Prevention Level 0.33 (0.11) 0.11 0.45 (0.08) 0.20 0.55 (0.08) 0.30 0.47 (0.06) 0.22 Community/Clinical Continuum 0.70 (0.12) 0.49 0.79 (0.06) 0.63 0.78 (0.06) 0.60 0.74 (0.05) 0.54 Use of Community Infrastructure 0.61 (0.12) 0.37 0.79 (0.06) 0.62 0.81 (0.06) 0.66 0.77 (0.05) 0.59 Community Involvement in Planning 0.20 (0.11) 0.04 0.25 (0.10) 0.06 0.15 (0.10) 0.02 0.21 (0.07) 0.04 67 Confirmatory Factor Analysis Results Table 5.6 displays the factor loadings by CHNA report cycle and overall, as well as the associated proportion of variance in each measure explained by the latent variable community orientation (i.e., the R-squared values). In general, factor loadings had consistent values from cycle to cycle, and overall. All factor loadings were statistically significant from zero. The variables with the highest factor loadings were the use of community infrastructure and community / clinical continuum measures, with factor loadings above 0.6 and 0.7, respectively. Intuitively, this makes sense, as these measures both reflect direct use of community resources and knowledge, and thus is an ideal representation of community orientation. Two other variables, targeting of DUHN communities and the prevention level measures, had factor loadings above 0.4, except for the first cycle CHNA data. The final variable, planning of the implementation strategy, consistently had the lowest factor loadings across all three time periods, with a factor loading ranging from 0.15 to 0.25. There is no fixed cutoff for keeping or dropping variables due to their factor loadings, as decisions about what items have low factor loadings are dependent on, for example, the theoretical construct being tested and how well the model being tested fits the data.63–65 Factor loadings as low as 0.2 can be included in the construction of latent variables if a clear justification for its inclusion can be made.64,65 Given that this measure was developed as part of an external framework and that it reflects direct use of community knowledge, there is sufficient reasoning for its inclusion as part of the latent construct community orientation. However, the low factor loadings of the planning 68 variable are somewhat concerning, and further examination of the underlying data is warranted before deciding to keep or drop this variable. In addition to assessing the factor loadings, I also assessed the model fit with a variety of goodness-of-fit measures. As evidenced in Table 5.7, the CFA single-factor model fit the data well, especially in the first two CHNA cycles; the third cycle goodness- of-fit measures were more mixed (please note that, because the full sample used clustered errors, I could only report on two model fit indexes):  The model chi-square values for the three CHNA cycles are not statistically significant, which indicates that the model reproduces the observed covariances among the five items well.66 CHNA Cycle 3 data is close to being statistically significant, however, as it is slightly above a p-value of 0.05.  The root mean square error of approximation (RMSEA) is an absolute fit index (i.e., one that compares the hypothesized model to a perfect fit model that assesses the extent to which the hypothesized model is different from a perfect model. The first two CHNA cycles had a RMSEA less than 0.05 signifying a close fit; the third cycle had a RMSEA less than 0.10 signifying an acceptable fit.67  The Comparative Fit Index (CFI) is an incremental fit index (i.e., one that compares the hypothesized model to the baseline model). All three CHNA cycles had acceptable CFI values over the suggested cutoff of 0.95.66,67  The Tucker-Lewis Index (TLI) is another incremental fit index, with values over 0.95 signifying acceptable fit. The third cycle CHNA model’s TLI value of 0.91 is below the suggested cutoff point, while the other two cycles had TLI over one, suggesting excellent fit.66,67 69  The standardized root mean square residual (SRMR) is an index of the average standardized residuals between the hypothesized and observed covariance matrices.66,67 The SRMR for all three CHNA cycle models and for the full sample model were all below 0.05, signifying a good fit.  The coefficient of determination – otherwise known as R-squared – all had high values for the three CHNA cycles and the overall sample, suggesting the model was able to explain most of the variability observed in the data.67 Table 5.7 Goodness of Fit Measures for Single-Factor Community Orientation Model CHNA Cycle 1 CHNA Cycle 2 CHNA Cycle 3 Overall Model Chi-Square (p-value) 4.92 (0.42) 4.32 (0.51) 10.7 (0.06) - RMSEA 0.00 0.00 0.09 - Prob. of RMSEA <= 0.05 0.60 0.66 0.14 - Comparative Fit Index 1.00 1.00 0.96 - Tucker-Lewis Index 1.01 1.01 0.91 - Standardized Root Mean Squared Residual 0.04 0.04 0.04 0.02 Coefficient of Determination 0.65 0.80 0.81 0.77 Note: RMSEA = Root mean squared error of approximation Taken together, the six goodness-of-fit measures suggest the single-factor model fits the data well, although it does not perform as well with the third CHNA cycle data. 70 There is nothing to suggest that the five measures are not a good approximation for community orientation, although further examination of the planning measure is still warranted. A four-measure single-factor model that excludes the planning measure would need to provide an equal or better fit than the five-measure single-factor model. Measurement Invariance As discussed, the CFA provided evidence to support the five-measure single- factor model of community orientation. Based on initial inspection, the factor loadings all remained relatively consistent across the three time periods and with the overall sample. However, I need to systematically confirm that my conceptualization of community orientation is consistent across time – that is, that the structures of the latent factor and associated parameters do not significantly change depending on when they were measured – before I can, for example, compare mean community orientation scores by report cycle.68 To confirm this, I used measurement invariance, a technique involving a series of tests of equivalence that correspondingly constrict parameters across time periods or groups.68–70 I followed the measurement invariance hierarchical approach suggested by the UCLA Institute for Digital Research & Education71, which draws upon the work of both Gregorich and Acock suggestions for establishing measurement invariance.72,73 Measurement invariance again uses a CFA framework wherein: (1) a model where all parameters are freely estimated (Model 1) is compared to a model where factor scores are held invariant (Model 2), to confirm metric (or pattern) invariance; (2) Model 2 is compared to a model with invariant loadings and intercepts, to confirm strong invariance (Model 3); (3) Model 3 is compared with a model holding factor loadings, intercepts and 71 residuals invariant to confirm strict invariance (Model 4); (4) Model 4 is compared with a model holding factor loadings, intercepts, residuals, and factor means invariant (Model 5); and (5) Model 5 is compared with a model that holds loadings, intercepts, residuals, factor mean, and variances invariant (Model 6).71 With each model comparison, I noted the difference in chi-square values; a non-significant difference in chi-square values between models indicated confirmation of that level of measure invariance. As illustrated in Table 5.8, I demonstrated both metric invariance and strong invariance, based on a non-significant difference in chi-square values.70 I did not demonstrate any level of strict invariance, however.70 By confirming metric invariance between the three time periods, I have verified that scaling of the latent construct community orientation is consistent across the three time periods, and no factor loading is different regardless of the time period.69,70 I have also found that scalar invariance is supported, which confirms that the intercept is equivalent across the three time periods.69,70 Despite not validating strict invariance, I can still compare mean differences between time periods because unequal residuals are inconsequential to the interpretation of latent mean differences.70 72 Table 5.8 Invariance Testing Across Three Time Periods Model 𝑿𝟐(df) CFI RMSEA Model Comparison ∆𝑿𝟐(df) p Model 1: All parameters free 20.11 (15) 0.981 0.052 -- -- -- Model 2: Metric invariance (loadings invariant) 25.90 (23) 0.989 0.032 1 vs 2 5.79 (8) 0.67 Model 3: Strong invariance (loadings/intercepts invariant) 32.07 (31) 0.996 0.017 2 vs 3 6.18 (8) 0.63 Model 4: Strict invariance (loadings/intercepts /residuals invariant) 61.15 (41) 0.926 0.063 3 vs 4 29.08 (10) 0.00* Model 5: Strict invariance plus equal factor means 79.74 (43) 0.865 0.083 4 vs 5 18.59 (2) 0.00* Model 6: Strict invariance plus equal factor means & variances 97.09 (45) 0.809 0.096 5 vs 6 17.35 (2) 0.00* Cluster Analysis While the CFA provided evidence to support the one-factor five measure, one of the measures - the community involvement in planning measure - had a very low factor loading, suggesting that it might be better to drop it from the model. Before dropping it, however, it may be helpful to further clarify the complexity underlying the data structure to see if I can ascertain why it performs poorly as a measure of community orientation. One option for clarifying this data structure is to use a different multivariate statistical tool that, like factor analysis, helps to identify relationships among items and/or subjects: cluster analysis.74 As described in Chapter 4, cluster analysis is an analytic 73 approach for identifying homogenous groups in data. Researchers have used both methods in tandem as a way of either confirming a simple data structure if both methods have similar conclusions, or for identifying a more complex data structure that might not have been discovered through either method alone.74,75 For the purposes of identifying unique clusters, I opted to use a hierarchical agglomerative approach to identify similar clusters based on the five community orientation measures. I experimented with several different approaches for discerning appropriate clustering methods and found that Ward’s clustering method produced the clearest demarcations between groups.76 I determined the number of clustered groups based initially on a visual inspection of the dendrogram, and then through a more in- depth inspection of the clusters to confirm that they were conceptually distinct.76 I also used the Calinski–Harabasz method for determining the appropriate number of clusters.51 A three-cluster solution had the largest pseudo-F value of 60.0, indicating that the three- group solution is the most distinct compared with the two-group and four-group solutions. As I had done using CFA, above, I conducted a similar three clustering solution for each of the three time periods and then for the entire sample and confirmed that three clusters provided a parsimonious solution, regardless of the sample used. Table 5.9 shows the three groups identified through the cluster analysis and how the three clusters varied across the five measures:  Cluster 1 is characterized by minimal involvement of community stakeholders in either planning or implementation, and minimal targeting of DUHN communities. These hospitals also tended to use fewer preventative interventions relative to the other two clusters. 74  Cluster 2 is characterized by high level of community involvement and use of community resources as part of the implementation strategy but low involvement of community stakeholders in planning their implementation strategy. This cluster of hospitals also tended to target DUHN communities more than either of the other two clusters.  Cluster 3 is characterized by high level of community involvement, but only as it pertains to the implementation strategy planning process; use of community stakeholders and resources was somewhat lower as part of implementing chosen strategies. These hospitals also targeted DUHN communities slightly less than hospitals from Cluster 2. Factor Analysis and Cluster Analysis Results How might we parse the CFA findings of a five-measure one-factor model with the cluster analysis findings of a three-cluster model? A likely solution is that the underlying data is better thought of as a single “chained” cluster, as defined by Gorman and Primavera, where the structures placed on the data by the clustering algorithm are identifying opposite ends of this single cluster.74 Thus, cluster 1 may be best thought of as a lower end of this chained cluster, while both clusters 2 and 3 are on the higher end. As the two higher-end clusters are most differentiated by their planning measure values, this helps to elucidate why the planning measure has the lowest factor loadings of the five measures and is a comparatively poor measure of community orientation: when a hospital has a low planning measure, the remaining four measures may also be low (and grouped 75 in cluster 1) or high (and grouped in cluster 2). This finding would seem to support dropping the measure from the model. Still, the results of the cluster analysis provide a good framework for (1) further exploring the characteristics of hospitals that are more and less community oriented by defining cluster membership; and (2) determining where, when, and why hospitals might have changed their strategy over the course of the three time periods. 76 Table 5.9 Description of Three Cluster Model Overall and Across Three CHNA Cycles Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 1st Cycle 2nd Cycle 3rd Cycle Overall Cluster 1: Low to Medium Prevention Level; Low Community Involvement in Planning and/or Implementation; Minimal Targeting of DUHN Communities Number of Hospitals 66 52 45 163 Prevention Level (mean, sd) 0.574 (0.139) 0.538 (0.180) 0.578 (0.151) 0.564 (0.157) Targeting of DUHN Communities (mean, sd) 0.139 (0.141) 0.147 (0.153) 0.127 (0.144) 0.138 (0.145) Community/Clinical Continuum (mean, sd) 0.086 (0.105) 0.089 (0.105) 0.131 (0.115) 0.099 (0.109) Use of Community Infrastructure (mean, sd) 0.067 (0.098) 0.097 (0.123) 0.088 (0.099) 0.082 (0.107) Community Involvement in Planning (mean, sd) 0.003 (0.021) 0.000 (0.000) 0.000 (0.000) 0.001 (0.013) Cluster 2: High Prevention Level; High Community Involvement in Implementation; Minimal Community Involvement in Planning; High Targeting of DUHN Communities Number of Hospitals 20 18 25 63 Prevention Level 0.620 (0.110) 0.679 (0.102) 0.733 (0.091) 0.682 (0.110) Targeting of DUHN Communities 0.446 (0.298) 0.541 (0.305) 0.518 (0.299) 0.502 (0.298) Community/Clinical Continuum 0.448 (0.275) 0.417 (0.231) 0.389 (0.186) 0.416 (0.228) Use of Community Infrastructure 0.246 (0.200) 0.316 (0.200) 0.483 (0.271) 0.360 (0.250) Community Involvement in Planning 0.023 (0.078) 0.020 (0.060) 0.025 (0.086) 0.023 (0.076) Cluster 3: High Prevention Level; High Community Involvement in Planning; Low to Medium Community Involvement in Implementation; Low to Medium Targeting of DUHN Communities Number of Hospitals 39 55 55 149 Prevention Level 0.618 (0.128) 0.633 (0.109) 0.651 (0.125) 0.636 (0.120) Targeting of DUHN Communities 0.224 (0.209) 0.223 (0.200) 0.314 (0.282) 0.256 (0.238) Community/Clinical Continuum 0.211 (0.221) 0.223 (0.195) 0.280 (0.227) 0.241 (0.215) Use of Community Infrastructure 0.136 (0.152) 0.202 (0.197) 0.220 (0.219) 0.191 (0.197) Community Involvement in Planning 0.636 (0.141) 0.569 (0.195) 0.612 (0.189) 0.602 (0.181) 77 Cluster Characteristics To further differentiate between the three clusters, I also looked at some of the key characteristics of hospitals comprising each cluster. As shown in Table 5.10, there are some noticeable patterns in cluster membership, including:  Cluster 1 is comprised of a high proportion of medium-sized hospitals, which is consistent across the three time periods. Larger hospitals trended away from Cluster 1 in CHNA cycle 1 towards clusters 2 and 3 in the later two time period CHNA cycles, while smaller hospitals likewise moved away from cluster 1 towards cluster 3.  Stand-alone hospitals were mostly grouped in cluster 1 across all three CHNA cycles, while hospitals that were part of a health system moved from cluster 1 to cluster 3 over the three time periods.  Almost all northeast hospitals moved away from cluster 1 in CHNA cycle 1 to cluster 3 in CHNA cycle 3. Western hospitals likewise moved to clusters 2 and 3 by the second and third CHNA cycles. Southern hospitals were more consistent across time, with most hospitals remaining in cluster 1. Midwestern hospitals moved slightly away from cluster 1 towards cluster 3 in the last two CHNA cycles.  The proportion of rural and urban hospitals were about equal for all three clusters in all three CHNA cycles and, except for some slight movement towards cluster 2 in CHNA cycles 2 and 3 from urban hospitals, did not change significantly. 78 Table 5.10 Cluster Membership Over 3 CHNA Cycles CHNA Cycle 1 CHNA Cycle 2 CHNA Cycle 3 Clus 1 % (n) Clus 2 % (n) Clus 3 % (n) Clus 1 % (n) Clus 2 % (n) Clus 3 % (n) Clus 1 % (n) Clus 2 % (n) Clus 3 % (n) Size Small (<100 beds) 50 (28) 13 (7) 38 (21) 36 (20) 11 (6) 54 (30) 36 (20) 18 (10) 46 (26) Medium (101-299 beds) 59 (23) 15 (6) 26 (10) 59 (23) 15 (6) 26 (10) 44 (17) 18 (7) 38 (15) Large (300+ beds) 50 (15) 23 (7) 27 (8) 30 (9) 20 (6) 50 (15) 27 (8) 27 (8) 47 (14) Ownership Stand-alone 57 (13) 0 (0) 43 (10) 52 (12) 0 (0) 48 (11) 57 (13) 4 (1) 39 (9) Health System 52 (29) 20 (20) 28 (53) 39 (40) 18 (18) 43 (44) 31 (32) 24 (24) 45 (46) Medicare APM Participant 54 (50) 22 (20) 25 (23) 39 (36) 19 (18) 42 (39) 32 (30) 25 (23) 43 (40) Non-Participant 50 (16) 0 (0) 50 (16) 50 (16) 0 (0) 50 (16) 47 (15) 6 (2) 47 (15) Region Northeast 45 (10) 18 (4) 36 (8) 32 (7) 18 (4) 50 (11) 5 (1) 5 (1) 91 (20) South 56 (20) 6 (2) 38 (14) 56 (20) 6 (2) 39 (14) 47 (17) 19 (7) 33 (12) Midwest 49 (20) 17 (7) 34 (14) 39 (16) 12 (5) 49 (20) 39 (16) 20 (8) 42 (17) West 61 (16) 27 (7) 12 (3) 34 (9) 27 (7) 38 (10) 42 (11) 35 (9) 23 (6) Environment Urban 54 (42) 16 (13) 29 (23) 41 (32) 18 (14) 41 (32) 35 (27) 24 (19) 41 (32) Rural 51 (24) 15 (7) 34 (16) 43 (20) 9 (4) 49 (23) 38 (18) 13 (6) 49 (23) Note: CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 79 Change in Cluster Grouping Figure 5.3 Transitions Between Community Orientation Groupings To further identify patterns in how each NPH sorted itself into clusters, I examined the underlying data to note when and where NPHs changed cluster membership, which is represented by Figure 5.3. As is apparent, most NPHs remained in their cluster grouping from cycle to cycle, although there was also a fair amount of movement between clusters, representing a significant change in NPH strategy. Of the 125 NPHs in the sample, 48 remained in the same cluster over the three time periods; the remaining NPHs changed cluster membership at least once. This represents a significant amount of change in strategy from cycle to cycle, and closer examination is warranted to understand the nature of these patterns. Cluster 1 n = 66 Cluster 2 n = 20 Cluster 3 n =39 Cluster 3 n = 55 Cluster 3 n = 55 Cluster 2 n = 18 Cluster 2 n = 25 Cluster 1 n = 52 Cluster 1 n = 45 43 6 17 2 10 8 7 2 30 30 9 13 6 8 4 9 8 38 CHNA Cycle 1 (2011-2014) CHNA Cycle 2 (2015-2017) CHNA Cycle 3 (2018-2021) 80 Examination of CFA Results Over the course of this chapter, I have: (1) discussed the operationalization of five measures, using the ASACB framework, and established an initial composite measure of community orientation based on a single-factor model with reasonably good fit to the data; (2) verified that this model of community orientation is consistent across three time periods by testing measure invariance; and (3) conducted cluster analysis and found a three-cluster model that both clarified the underlying data structure and provided a structure for discussing changes in implementation plan strategy. In addition, I also weighed the possibility of dropping the implementation planning measure due to its low factor loading. As further evidenced by the cluster analysis, the planning measure does not appear to provide a good measure of community orientation. What follows is a discussion of these results, contextualized by returning to the original text documents to inform the discussion. Change in Strategy Among Hospitals Figure 5.3, which mapped out how cluster membership changed over the course of the three CHNA cycles, showed a substantial amount of change in strategy among most NPHs. To highlight some of these trends and provide context for why NPHs might opt to change their strategy, I identified three NPHs and their associated implementation 81 strategies to explore and discuss in more detail as case examples. I chose NPHs that experienced a change in grouping at least once over the three CHNA cycles. Example #1: This 164-bed hospital located in upstate New York is a member of small health system. For their 2013 CHNA, this hospital identified two priority health issues: prevention of chronic diseases and promotion of healthy women, infants, and children. To address chronic disease in their community, this hospital chose in situ interventions that were clinical in nature and/or conducted by the hospital, such as: employee wellness initiatives, including smoking cessation classes; making meals sold and served at the hospital healthier; free cardiac screenings; and continued participation in various local health initiatives, such as the local rural health network and the local county health services advisory board. Initiatives that were proposed to address maternal and child health were similarly clinical and implemented by the hospital, although they tended to target DUHN communities more, such as developing a crisis intervention and support services for victims of sexual assault and abuse and offering workshops to local groups on healthy relationships. For the first and second CHNA cycles, the hospital’s implementation strategies were grouped into Cluster 1: generally focused on primary and secondary prevention-type interventions with minimal contributions from other community stakeholders at the intervention stage and no contribution to the planning of strategies. There was also minimal targeting of DUHN communities and patients. The 2019 3rd cycle implementation strategy saw a significant shift in strategy that, as reflected in the narrative of the published plan, is due at least in part to changes in New York’s community benefit requirements. Since 2013, New York has required NPHs to 82 submit ‘community service plans’ (CSPs), wherein hospitals would specify how they would address the state’s Department of Health (DOH) pre-identified priority areas based on the results of each NPHs’ CHNA (CSPs could also comply with the federal community benefit requirements related to developing implementation strategies).17 In November 2015, the New York State DOH Commissioner compelled New York NPHs to submit a single community health improvement plan (CHIP) with their local public health agency (LPHA), essentially mandating that NPHs work with their LPHA in some capacity on a countywide CHIP.77 The CHIP (or implementation strategy) released by this hospital in 2019 is very different from previous implementation strategies, in that it clearly is responsive to the NY DOH regulations. Rather than an implementation strategy that describes the planned activities of the NPH only, the CHIP describes a range of planned activities, some of which involve the NPH and some that involve other stakeholders. Most chosen interventions involve working with at least one community stakeholder, usually the LPHA, but also local schools, social service agencies, and advocacy groups, and include some activities that address non-health social determinants. Interestingly, there are several interventions mentioned in the CHIP that are like some of the previous NPH implementation strategies, such as offering cancer screenings, but that also discuss utilizing local partners to help implement these activities, making them more community- oriented. Finally, the planning of the hospital’s implementation strategy involved a facilitated discussion between the hospital and a range of local stakeholders, including the LPHA, ensuring community stakeholders provided input into the planning process. As a result, the third CHNA cycle implementation strategy for this NPH is much more 83 community oriented, as reflected in both their composite index score and in their movement to Cluster 3. Example #2: This 25-bed critical access hospital (CAH) is in a very rural community in central North Dakota. It belongs to a large religiously affiliated regional health care system that formally launched in 2016 (although this CAH was owned by a large national corporation prior to 2016); this system includes a tertiary hospital as well as six other CAHs. This CAH’s first implementation strategy in 2013 was characterized by hospital- led interventions that sought to address a range of priority health issues, from chronic conditions (diabetes, cancer, and obesity) to access to care and excessive alcohol consumption. Interventions tended to be clinically-oriented – for example, training to improve diabetes care and expanding the range of cancer care services – that were also targeted to the community-at-large. Implementation strategy planning was done entirely by hospital staff. As a result, this CAH’s first implementation strategy started in Cluster 1. The 2016 implementation strategy differed from the 2013 strategy in several ways that moved the implementation strategy to Cluster 3. First, the identified priority health issues were different from the 2013 CHNA. Planning of the implementation strategy was much more community-oriented, in that it solicited significant input from the county public health agency. The interventions themselves were also very community-oriented, involving the public health agency, law enforcement, county government officials, area schools, and a local social service provider. While the activities themselves were not targeted to underserved communities, the prevention level of the interventions were much 84 higher than the 2013 strategies and included several efforts to address non-health social determinants; this included a restorative justice intervention to address bullying in schools and exploring ways of expanding childcare options in the community. The 2019 implementation strategy further advanced the CAH’s 2016 chosen strategy. Many of the interventions still relied heavily on community involvement and incorporating hospital stakeholders into the workflow of different community stakeholders. Community stakeholders also continued to contribute significantly to the development of the implementation strategies. Many of the strategies also expressly addressed several non-health determinants, such as housing and school safety, making this iteration even more preventative than the 2016 set of activities, although the CAH’s strategy remained situated in Cluster 3. As with the previous strategy, interventions were again targeted towards the community-at-large. There are several possible reasons for the apparent change towards a more community-oriented strategy. First, as discussed in the 3rd cycle implementation strategy, one of the most pressing health issues for this community (and the CAH) was a need to recruit and retain practicing physicians, especially primary care physicians. Many of the interventions, particularly those that addressed non-health social determinants, were part of a broader community-wide strategy to improve living conditions in the community and encourage young families to relocate to their area. For example, the hospital was one of several stakeholders that contributed to the creation of a housing resource guide, while also advocating for more equitable city/county ordinances to make housing more affordable. Thus, the hospital’s needs dovetailed with the community’s needs, and as a 85 result, they could re-orient their implementation strategy away from offering clinical interventions. While not explicitly stated (but potentially important), the reorganization of the CAH’s ownership that occurred in 2016 may have also made the NPH more open to community collaboration. The transition to a more community-oriented implementation strategy coincided with a transition to a regional health system with more localized oversight. This transition seemed to impact how the CHNA was conducted, at least going by the change in appearance to a much more professional, organized CHNA report; it would not be too much of a stretch to say that the transition likewise resulted in a change in leadership philosophy. Example #3: This small independent rural hospital in central Texas started its first CHNA cycle with a fairly community-oriented implementation plan, with much of its activities targeted to underserved communities and at least somewhat reliant on community partners to help implement the plan. Activities included, for example, a public health-led effort to educate community members on healthy eating at restaurants that targeted local Spanish-speaking residents. Also noticeable is how involved different community stakeholder organizations were in both the conduct of the CHNA, as well as the planning process. This included a community summit that hosted over seventy community leaders to help identify priority health issues and develop the implementation plan. As a result, this NPH’s implementation plan was grouped into cluster 3. The next CHNA cycle saw this NPH’s implementation strategy change to be less community oriented, although it still went through a process of soliciting significant input 86 from community stakeholders in planning the implementation strategy. Chosen strategies were less preventative and largely implemented only by the NPH. Because of the extensive community involvement in the planning process, and because several activities still targeted underserved communities, the NPH’s implementation strategy was still grouped into cluster 3. The third CHNA cycle implementation strategy was the least community oriented, with very few community-based interventions. Most activities were related to increasing access to different health care services (e.g., improving access to telemedicine; expanding access to geriatric psychiatry). There was also no community input solicited in the implementation strategy planning process, and as a result, this NPH’s strategy was grouped in cluster 1. There is no accompanying narrative describing why this NPH opted to be less community oriented in the later CHNA cycles, so any hypothesis as to what happened is speculative. It is safe to say that this NPH started with a highly collaborative process in conducting both its CHNA and implementation strategy that became much less community oriented by the latest CHNA cycle. Both theory78 and empirical research79 have suggested that sustaining collaborative efforts between different organizations can be inherently difficult and complex, especially when formed in response to social issues.80 Engaging and coordinating different community and clinical stakeholders across a range of different CHNA and implementation strategy activities is likely to require a significant amount of effort and resources from NPHs. Where stakeholders are intrinsically different – e.g., where there are significant differences in resources, in the types of personnel that 87 comprise each organization, in organizational motivation, norms, and values – the more difficult it might be to sustain partnerships to address social issues and causes.80 To the extent that it is possible for an NPH to shirk on some responsibilities (maintaining community partnerships) in favor of responsibilities that are more immediate and/or familiar (maintaining or expanding acute care services and increasing revenue), at least some, if not many NPHs will likely opt to shirk. Federal oversight of the enhanced community benefit regulations likely contributed to the possibility of NPHs’ noncompliance and prevented NPHs from maintaining their level of community orientation. Because of the ambiguity written into the tax regulations and because it does not have the authority or capacity to provide sufficient oversight, the Internal Revenue Service (IRS) has not provided rigorous oversight of the community benefit regulations.81 By the federal government’s own admission, the IRS community benefit standards do not define the activities or services that an NPH must provide to justify their tax exemption, so there is tremendous uncertainty over what NPHs should do. Per the Government Accountability Office (GAO)’s 2020 report to Congress on the oversight of hospitals’ tax-exempt status, “… a hospital could, in theory, maintain a tax exemption by operating an emergency room open to all and accepting patients on Medicare or Medicaid… while spending little to no money on charity care or other community benefit activities.” (p. 15).81 The GAO’s report also analyzed NPH’s Schedule H tax data and found hundreds of hospitals that either reported no financial assistance, no community benefit or spent less than 1 percent community benefit spending, yet no action was taken against these NPHs.81 Because of this lack of oversight, which may have been learned by NPHs as they initially complied 88 with the regulations in the first and second CHNA cycles, NPHs (such as this Texas NPH) might feel that they can satisfice and not expend resources to engage with the community like they had in previous CHNA cycles. These examples highlight some of the reasons why NPHs might have changed their strategy at some point over the past ten years. Whether its explicitly clear in the narrative - i.e., in the case of the New York hospital – or is possibly implied, such as through a change in leadership or a need to conserve resources, the reasons why a majority of NPHs have changed how they conduct their implementation strategies are likely varied and complex. Further clarity on the reasons behind a change in strategy among NPHs would be helpful for determining how influential policy can be at incentivizing NPHs to be more community oriented. Use of the Community Involvement in Planning Measure To further examine why it may be prudent to drop the community involvement in planning measure, I again found an example NPH – in this case, a typical hospital from cluster 2. Cluster 2 NPHs were very similar to cluster 3 NPHs – high use of community organizations and resources as part of the implementation strategy activities, high targeting of underserved communities, highly preventative activities - but were much less likely to engage with community stakeholders as part of the implementation strategy planning process. How might this be possible? 89 Example #4: This large (300+ beds) hospital in an urban part of California is a typical cluster 2 hospital. All three of its implementation strategies were highly community oriented, with activities that addressed non-health social determinants, that worked with different community-based organizations that they connected to their clinical delivery, and that were highly targeted to underserved communities. Some example interventions from their first CHNA include: providing temporary housing through a housing agency to homeless patients that also provided wrap-around health care services; working with an existing local organization that served an underserved neighborhood to address high- blood pressure and heart disease; and working with certain community-based organizations that served immigrants to help recently discharged high-risk patients with needed services to ensure compliance. In particular, this NPH’s first cycle CHNA seemed to describe programming that had already been in place, and that was well- connected to the community; thus, when describing how they put together their implementation strategy, it appeared to be entirely done internally. This example highlights a key issue with the planning measure, in particular with those hospitals grouped into cluster 2: a low value is not indicative of low community orientation. In the case of this example NPH, it is likely that either the administrators at this NPH did not need to consult with community stakeholders because they engaged with them on a regular basis already, or that any consultation with community stakeholders as part of the planning process was not mentioned in the narrative of the report. A highly collaborative planning process involving community stakeholders is also not always indicative of a highly community-oriented implementation plan. The 90 aforementioned Texas NPH, for example, discussed a high level of engagement with community stakeholders as part of their second CHNA cycle implementation strategy, but the plan itself was actually not community oriented. It may be that an NPH is willing to solicit feedback from community members during the planning process but pursue a strategy that is less reliant on community organizations; likewise, community organizations may want a more clinically oriented implementation strategy from their NPHs and suggest as much during the planning process. Based on these contradictory results, I concluded that the planning measure was too inconsistent in measuring community orientation – particularly among hospitals that scored highly on the other four measures – and that dropping the planning measure from the latent variable community orientation was an appropriate step. To confirm that a single-factor model with the remaining four measures provided a good fit for the remaining data while remaining measure invariant across the three time periods, I reran all analyses with the four-measure model. These tables can be found in below (Tables 5.11 through 5.13) and show that (1) the factor loadings for the four measures remained almost exactly the same; (2) the goodness of fit measures improved for CHNA cycle 3 and cycle 2, but slightly worsened for CHNA cycle 1; and (3) community orientation, as reflected by the four-measure single-factor model, maintained both metric and strong invariance across the three time periods. As such, the four- measure community orientation model was used to calculate the factor scores, which I used as my composite index score. 91 Table 5.11 Standardized Factor Loadings for Four Community Orientation Measures Using Confirmatory Factor Analysis * Note: this model used clustered standard errors to adjust for using panel data; SE = Standard Error Table 5.12 Goodness of Fit Measures for Single-Factor Four-Measure Community Orientation Model CHNA Cycle 1 CHNA Cycle 2 CHNA Cycle 3 Overall Model Chi-Square (p-value) 3.1 (0.21) 0.66 (0.72) 4.40 (0.11) - RMSEA 0.06 0.00 0.09 - Prob. of RMSEA <= 0.05 0.31 0.79 0.18 - Comparative Fit Index 0.98 1.00 0.98 - Tucker-Lewis Index 0.92 1.04 0.95 - CHNA Cycle 1 CHNA Cycle 2 CHNA Cycle 3 Overall* Items Factor Loading (SE) R- squared Factor Loading (SE) R- squared Factor Loading (SE) R- squared Factor Loading (SE) R- squared Targeting of DUHN Communities 0.32 (0.11) 0.11 0.47 (0.08) 0.21 0.50 (0.08) 0.25 0.46 (0.06) 0.22 Prevention Level 0.31 (0.11) 0.10 0.44 (0.08) 0.19 0.54 (0.08) 0.29 0.46 (0.06) 0.21 Community/Clinical Continuum 0.70 (0.12) 0.48 0.81 (0.07) 0.65 0.76 (0.06) 0.58 0.73 (0.06) 0.53 Use of Community Infrastructure 0.62 (0.13) 0.39 0.77 (0.07) 0.60 0.83 (0.06) 0.69 0.78 (0.06) 0.61 92 Standardized Root Mean Squared Residual 0.04 0.02 0.03 0.02 Coefficient of Determination 0.64 0.79 0.81 0.77 Note: RMSEA = Root mean squared error of approximation Table 5.13 Invariance Testing Across Three Time Periods for Four-Measure Single Factor Model Model 𝑿𝟐(df) CFI RMSEA Model Comparison ∆𝑿𝟐(df) p Model 1: All parameters free 8.14 (6) 0.99 0.05 -- -- -- Model 2: Metric invariance (loadings invariant) 13.87 (12) 0.99 0.04 1 vs 2 5.73 (6) 0.45 Model 3: Strong invariance (loadings/intercepts invariant) 18.92 (18) 1.00 0.02 2 vs 3 5.05 (6) 0.54 Model 4: Strict invariance (loadings/intercepts /residuals invariant) 44.24 (26) 0.93 0.08 3 vs 4 25.32 (8) 0.00 Model 5: Strict invariance plus equal factor means 62.69 (28) 0.87 0.10 4 vs 5 18.59 (2) 0.00 Model 6: Strict invariance plus equal factor means & variances 82.16 (30) 0.80 0.12 5 vs 6 19.47 (2) 0.00 93 Establishing A Composite Index Measure The results of the CFA found that four of the five community orientation measures loaded onto the first factor reasonably well, regardless of sample. As such, the next step required determining how best to combine the four measures into a composite index measure and analyze the characteristics that predict community orientation. A composite index score is a single measure derived from a set of indicators that individually reflect a particular characteristic or characteristics of a larger construct. When combined, the composite index measure is better able to illustrate how that construct varies over time and by internal and external characteristics.82 There are a multitude of possible approaches available for deriving a composite score. For example, a weighting scheme may be appropriate when certain attributes of a given construct are more important than other attributes. There are also ranking methodologies that list how a particular unit fares relative to other units across measures and then combined into a composite ranking. I chose to use the factor scores predicted through linear regression, which is a common approach to producing composite measures.82,83 Factor scores produced through linear regression result in standardized scores with variance equal to the squared multiple correlation between the estimated factor score and the true factor value.52,84 While this treats all variables with equal weight, regardless of factor loadings, it also ensures any resultant index score best reflects the data structure of the associated measures.52 Other approaches, such as summing the raw values of each measure,85 were also tested as a composite measure, with very little difference found when determining mean NPH community orientation values. 94 Analysis of Community Orientation Composite Measure Table 5.14 shows the mean values of community orientation over the three time periods. The overall trend for NPHs in this sample was an increase in community orientation, as the community orientation score in the first reporting cycle is significantly less than the community orientation in the third reporting cycle. This provides some empirical evidence that, in aggregate, the mean value of community orientation in hospitals has increased over time Table 5.14 also illustrates how the community orientation measure varies by different hospital- and community-level characteristics overall and by reporting period. In general, regardless of hospital type, there was an overall trend of increasing community orientation. Certain hospital and community characteristics were associated with higher community orientation scores relative to their peers by the second and/or third reporting cycle, including being a member of a health system, participating in a Medicare alternative payment model (APM), residing in a state that expanded Medicaid, and residing in a non-Southern state. 95 Table 5.14 Mean Community Orientation Overall and by Hospital- and Community Characteristics Reporting Cycle 1 Reporting Cycle 2 Reporting Cycle 3 Overall Overall Reporting Cycle 1 (Standard Error) - - - 0.21 (0.01) Reporting Cycle 2 - - - 0.24 (0.01) Reporting Cycle 3 - - - 0.29 (0.01)* Hospital Network Groupings Group 5 (Low Int; High Cent; High Homoph) 0.19 (0.01) 0.21 (0.03) 0.27 (0.04) 0.22 (0.01) Group 4 (High Int; Med Cent; Med Homoph) 0.22 (0.02) 0.23 (0.02) 0.30 (0.04) 0.25 (0.02) Group 3 (High Int; Low Cent; Med Homoph) 0.25 (0.02)* 0.29 (0.03)* 0.35 (0.03) 0.30 (0.01)* Group 2 (Med Int; Med Cent; Med Homoph) 0.21 (0.01) 0.24 (0.01) 0.27 (0.02) 0.24 (0.01) Group 1 (Med Int; Low Cent; High Homoph) 0.22 (0.01)* 0.26 (0.02) 0.30 (0.02) 0.26 (0.01)* Teaching Status Not Teaching Hospital 0.21 (0.01) 0.24 (0.01) 0.29 (0.01) 0.25 (0.01) Teaching Hospital 0.23 (0.01) 0.26 (0.02) 0.31 (0.02) 0.27 (0.01) Size Small (<100 beds) 0.22 (0.01) 0.21 (0.01) 0.30 (0.02) 0.25 (0.01) Medium (101-299 beds) 0.21 (0.01) 0.23 (0.01) 0.27 (0.02) 0.24 (0.01) Large (300+ beds) 0.23 (0.01) 0.27 (0.02)* 0.32 (0.02) 0.27 (0.01) Ownership Stand-alone 0.21 (0.01) 0.20 (0.02) 0.24 (0.02) 0.21 (0.01) Health System 0.22 (0.01) 0.25 (0.04)* 0.31 (0.01)* 0.26 (0.01)* Medicare APM Non-Participant 0.21 (0.01) 0.22 (0.02) 0.25 (0.02) 0.23 (0.01) Participant 0.22 (0.01) 0.25 (0.01) 0.31 (0.01)* 0.26 (0.01)* Medicaid Expansion Non-Expansion State 0.20 (0.01) 0.18 (0.01) 0.23 (0.02) 0.20 (0.01) Expansion State 0.22 (0.01) 0.26 (0.01)* 0.31 (0.01)* 0.26 (0.01)* Region South 0.20 (0.01) 0.20 (0.01) 0.23 (0.01) 0.21 (0.01) Northeast 0.22 (0.01) 0.26 (0.02)* 0.30 (0.02)* 0.26 (0.01)* Midwest 0.22 (0.01) 0.26 (0.02)* 0.34 (0.02)* 0.27 (0.01)* West 0.23 (0.01) 0.26 (0.02)* 0.31 (0.02)* 0.27 (0.01)* Environment Urban 0.22 (0.01) 0.24 (0.01) 0.29 (0.01) 0.25 (0.01) Rural 0.22 (0.01) 0.25 (0.01) 0.30 (0.02) 0.25 (0.01) 96 Notes: (1) * = Significant difference (α ≤ 0.05). Means are compared to top reference group within each column and category; (2) APM = Advanced Payment Models; (3) CHNA Report Cycle 1 includes the years 2011 through 2014; Cycle 2 includes 2015 through 2017; Cycle 3 includes the years 2018 through 2021 In addition, community orientation by the homogenous network categorization developed in Chapter 4 again shows a general increase in community orientation across CHNA cycles, regardless of categorization. However, Group 3, which was characterized by high integration, low centralization, and medium levels of homophilization, consistently had higher levels of community orientation relative to other groups. This difference was significant relative to Group 5, which was characterized by low integration, high centralization, and high homophilization. This suggests more ‘community-oriented’ NPH networks that involved more community stakeholder groups may have in turn led to more community-oriented implementation strategies. Conclusions This chapter has explored the development and use of a community orientation composite measure to describe a latent characteristic of NPHs and their relationship to their communities. I discussed the process and framework used to define the composite measure and analyze the underlying data structure of the sub-measures, as well as the rigorous analytic process used to eliminate one of the sub-measures from the final composite measure. I also explored case examples of NPHs that had changed their strategies to be more or less community oriented to inform our understanding of the 97 factors that influenced hospital decision-making. Finally, I also analyzed the resulting composite measure to determine how it varied by certain hospital and community characteristics, including the homogenous CHNA network groupings developed in Chapter 4. As the analysis of the mean community orientation across different hospital and community characteristics showed, NPHs in Medicaid expansion states (i.e., largely non- Southern states) and part of health systems were more likely to be community oriented, especially in later CHNA cycles. It is also important to note what characteristics were not associated with community orientation, i.e., rural hospitals were just as likely to be community oriented as urban hospitals; teaching hospitals were just as community- oriented as non-teaching hospitals; and hospitals participating in advanced payment models (APMs) were only more community-oriented than their non-APM peers by the final CHNA cycle. The finding that the overall NPH community orientation increased over time is also significant. As discussed in Chapter 2, a key policy goal of the expanded community benefit regulations, as well as a general policy goal of health reform efforts, was to engender greater cooperation between NPHs and community organizations and agencies to address community health. This policy goal, at least as reflected in NPHs’ implementation strategies, appears to have been met. What are some possible reasons for why NPHs became more community-oriented over the three CHNA cycles? For one, it may be a reflection of developing organizational relationships between NPHs and other community organizations over time. Given that those NPHs that were part of a decentralized, integrated CHNA process 98 (i.e., in Group 3) also had more community-oriented implementation strategies, it may be, for example, that NPHs strengthened their community relationships first via the CHNA process, which led to NPHs trusting their community counterparts to implement effective programming. The specific component of the expanded community benefit regulations that required NPHs to consult with certain community stakeholders, then, may have contributed to this trend. Another factor that may have led to NPHs recognizing the value of a community- oriented strategy is the recognition that more upstream, targeted interventions were integral to keeping patients out of the hospital, which would also help NPHs reduce costs. This would explain why those NPHs that participated in APMs were more likely to be community oriented by the third CHNA cycle. While this would be a promising development, it is less optimal if NPHs were implementing these upstream interventions themselves and possibly replicating work done by other community organizations, i.e., setting up temporary housing for patients without homes at discharge, rather than supporting existing homeless shelters.86,87 The initial analysis of the five community- orientation sub-measures at the start of this chapter showed that, while there was increasing reliance on community infrastructure and continued integration of community and clinical workflows over the three CHNA cycles, it was not common for most chosen interventions. Again: hospitals prefer to implement their strategies on their own. Next steps will be to determine if increased integration of hospital and community organization work via implementation strategies has led to any significant changes in key population health outcomes. There has been some recent research that has examined this question, albeit on a smaller scale. A recent 2021 study from Santos and 99 Lindrooth found that New York counties where NPHs and local health agencies jointly prioritized substance use as a health issue and addressed it via a county-level community health improvement plan (CHIP) experienced a deceleration of substance use-related mortality of 8 deaths per 100,000.88 This is a promising result that suggests the effects of increased NPH community orientation are potentially beneficial. To determine if such effects on health are observed nationwide, it will be important to use the community orientation composite measure to differentiate between NPH strategies reflected in the implementation strategies. The composite measure offers a method for characterizing the community-based work NPHs have done over the past decade and will be helpful for any further policy analysis that attempts to assess the impacts of the expanded community benefit regulations. The composite community orientation measure also provides a more specific framework for NPHs to follow than what is offered through the original ASACB guidance document. I have worked to define each sub-measure beyond the general principles elucidated by the ASACB. Some NPHs in my sample have used the ASACB framework to guide their implementation strategies for many years, but have relied on their own set of definitions to assess whether their strategies align with the framework. These definitions are not consistent between NPHs, nor do they reflect a truly ‘community-oriented’ strategy, e.g., suggesting an intervention that involves communication with other medical providers adheres to the community infrastructure principle. 100 Chapter 6 - Conclusion The previous chapters have provided a set of analyses by which I have attempted to clarify how nonprofit hospitals have conducted themselves in relationship to their communities due to the changes in federal community benefit regulations. This chapter provides some final remarks on what conclusions I can draw from these results. I first review the conceptual model used to frame this research in light of each chapter’s findings. I also discuss the possible direct implications for researchers and policymakers, as well as for hospital leadership weighing decisions about the extent to which they might invest and support community organizations. I also review some potential broader implications for how community benefit regulations might be strengthened in the future. Finally, I also note some limitations to the approaches I have used to conduct this research as well as strengths, and what some next steps might be to further our understanding of hospital-community relationships. Discussion of Findings Chapters 4 and 5 examined different reflections of hospital community orientation. Chapter 5 sought to characterize community orientation via NPH implementation strategies and adherence to a community orientated approach to improving community health. Chapter 4 established a typology of networks used to conduct each CHNA that ranged in levels of community orientation, from those that were 101 least community oriented (i.e., highly centralized and homophilous, with low levels of community integration) to those that were very community oriented (i.e., highly decentralized and heterophilous, with high levels of community integration). Much of these analyses are predicated on noting the effects of changing strategies – that is, what happens when an NPH decides to become more integrated with community organizations. What are some of the hospital- and community-level factors that might have caused NPHs to change their strategies to be more or less community oriented over time? While there appears to be a holistic increase in community orientation across all types of hospitals over the three reporting cycles, it is also clear that certain types of hospitals are more likely to be community oriented relative to their peers. To think through what might be driving the changes in NPH strategy cycle to cycle, it is helpful to return to Oliver’s theoretical framework and conceptual model and consider the results within the different determinants of community orientation. Note that I am using a more inclusive definition of community orientation to include how the CHNA was conducted and the subsequent implementation strategy for this discussion. For one, the context of an NPH’s environment seems to affect the level of community-orientation of NPHs. NPHs in Southern states, for example, are much less likely to have become more community-oriented over time, either in how they conduct their CHNA or in their implementation strategies, while NPHs in states that expanded Medicaid are much more likely to be community-oriented. Some of the effects of environmental context may be related to local political norms, given the positive association between Medicaid expansion and higher community orientation scores. At least some of the effects of environmental context is also likely driven by state and/or 102 local constituent norms and expectations. Some state regulators have advanced their own community benefit requirements for NPHs to dovetail with the federal CHNA requirements in exchange for state tax exemptions. New York’s requirement that NPHs work with their local health departments on their CHNA and then jointly to select at least two health priorities from the New York Prevention Agenda to address collaboratively17 has resulted in New York-based NPHs spending, on average, $393,000 to $786,000 more per year on population health initiatives than other states’ NPHs.37 Within this study’s subsample of New York-based NPHs, community orientation scores almost all increased significantly over time - many because they were more likely to utilize community resources and liaise with community stakeholders as part of their implementation strategies. Cause-related determinants may be less predictive of hospital community orientation, with no noticeable differences between rural and urban hospitals despite the latter NPHs’ larger constituencies and more prominent visibility. It may be that such determinants are less relevant when the enforcement mechanism is weak, at least as it pertains to NPHs working with community stakeholders on their implementation strategy to comply with the federal regulations. Both theory78 and empirical research79 have suggested that sustaining collaborative efforts between different organizations can be inherently difficult and complex, especially when formed in response to social issues.80 Engaging and coordinating different community and clinical stakeholders across a range of different CHNA and implementation strategy activities is likely to require a significant amount of effort and resources from NPHs. Where stakeholders are intrinsically different – e.g., where there are significant differences in resources, in the types of personnel that 103 comprise each organization, in organizational motivation, norms, and values – the more difficult it might be to sustain partnerships to address social issues and causes.80 To the extent that it is possible for an NPH to shirk on some responsibilities (maintaining community partnerships) in favor of responsibilities that are more immediate and/or familiar (maintaining or expanding acute care services and increasing revenue), at least some, if not many NPHs will likely opt to shirk – particularly if enforcement of these responsibilities is weak. Federal oversight of the enhanced community benefit regulations likely contributed to at least some NPHs’ less community-oriented strategies. Because of the ambiguity written into the tax regulations and because it does not have the authority or capacity to provide sufficient oversight, the Internal Revenue Service (IRS) has not successfully ensured compliance with the community benefit regulations.81 By the federal government’s own admission, the IRS community benefit standards do not define the activities or services that an NPH must provide to justify their tax exemption, so there is tremendous uncertainty over what NPHs should do. This includes, for example, ensuring NPHs comply with the CHNA requirements to solicit input from certain community stakeholders. The GAO’s report also analyzed NPH’s Schedule H tax data and found hundreds of hospitals that either reported no financial assistance, no community benefit or spent less than 1 percent of their functional expenses on community benefit-related spending, yet no action was taken against these NPHs.81 Because of this, the lack of federal oversight, which may have been learned by NPHs as they initially complied with the regulations in the first reporting cycle, coupled with a lack of significant oversight from either state or local stakeholders over their community benefit activities, might lead 104 to NPHs feeling that they can satisfice and not expend resources to fully engage with the community. This leads me to note the effects of key control-related determinants. As discussed, all NPHs are required to comply with the poorly-regulated federal community benefit requirements, which they do with varying levels of compliance - but this does not mean governmental control has no effect on hospital community orientation. Where there are state laws enforcing additional requirements, particularly as it pertains to ensuring NPHs work with their LHAs on their CHNAs and implementation strategies, NPHs are more responsive and compliant to the state-level community benefit regulations. I have frequently noted throughout this dissertation that many New York-based NPHs were motivated to be more community oriented via their implementation strategies once their state mandated that they work with their LHAs. Although more recent, Ohio has enacted similar community benefit regulations89 and the latest CHNA cycle for Ohio-based NPHs in this cycle likewise seemed to showed increased community orientation via their implementation strategies. This suggests control-related determinants are an effective source of pressure to ensure NPHs are more community oriented, but it relies on NPHs respecting the potential for punishment if they are non-compliant. The effects of the final set of content-related determinants on hospital community orientation are undoubtedly significant, but likely vary from NPH to NPH - and unfortunately, are also difficult to measure. The mission and values of an NPH in practice are a reflection of what those in leadership positions value, i.e., are they truly ‘charitable organizations’ that prioritize the welfare of their community above profitability, or are they private firms that prioritize the financial well-being of their business? NPHs have 105 been given wide latitude to identify and address the most pressing local health issues as they see fit, and many have responded by doing what is most familiar: working with their peer clinical partners and investing in interventions that are based within an acute care model. While this type of strategy may lead to improvements in community health, NPHs may also anticipate other direct, ancillary benefits, such as increased revenue and/or improved profitability. Many commonly-adopted implementation strategy activities, such as health fairs that test residents for key risk factors for disease, can, for example, also serve to promote the NPH locally, increase the number of patients served, and/or increase the use of reimbursable healthcare services – all despite a lack of evidence supporting these types of interventions.90 An additional common underlying assumption of these clinically-based interventions is that those most at high risk of contracting the diseases that have been prioritized during the CHNA process will have no problem accessing these needed services, which is a highly dubious assumption. As a result, unless efforts are made to reach those with access issues, the benefits of these interventions will likely accrue to those with the means to access them. Interventions that address upstream determinants and/or that expand on community resources are likely to be supported with empirical evidence and thus more likely to help people.91 Adopting a strategy that relies on such interventions, however, does not have the direct financial benefits of expanding clinically-based interventions (although there are possible longer-term and/or indirect benefits associated with such interventions92). For NPH leadership, a clinically-based approach is safer and more likely to be directly beneficial to the NPH, but the community-based approach is more likely to be in-line with the stated mission and values of NPHs and the general ethos of the 106 community benefit program. NPH leadership is therefore a key content-related determinant and very consequential when it comes to decisions around pursuing community-oriented or clinically-oriented strategies. While much of this discussion of community orientation is attempting to make sense of these research findings based on a theoretical model, it is also at least somewhat confirmed by a 2016 study conducted by Chen and colleagues that asked hospital leaders why they chose to get more invested in community partnerships. As noted by the authors, there were a myriad of reasons cited by interviewees, many of them ‘idiosyncratic’, but some of the overriding reasons include: (1) pressures from state regulators (several of the hospitals were located in New York); (2) belief in seeing a financial benefit despite a lack of reimbursement where NPHs were involved with a value-based care model; and (3) a commitment to being more community oriented because it was “the right thing to do”.93 Implications of Findings Because this research was broad in scope and overlapped across different areas of research and practice, there are potential direct implications for many different disciplines – in particular, for those that advance and study health policy. Direct Research & Practice Implications 107 Health policy researchers and policymakers may see these results as confirmation that policy per se can motivate further integration of hospitals and community organizations, and that requiring hospitals to interact with community stakeholders may have positive impacts on community health. States also have policy levers available to them that can work in conjunction with the federal community benefit regulations and further incentivize hospitals to become more community oriented, as has been observed in New York and Ohio. While there is still an acute need for identifying the specific mechanisms that must be in place to affect certain population health outcomes (e.g., a robust public health and social service infrastructure), this research provides empirical support for the types of policies that require hospitals to engage with non-clinical partners as part of their regular practice and centered on improving community health. More broadly, this study is one of many that grapples with whether regulations over if and how NPHs “earn” their tax exemption are actually worth the public investment. While this research is less direct than other studies that, for example, explicitly estimate the value of foregone taxes against a value of community benefits94,95 or measure changes in community benefit-related expenditures96, it does suggest that hospitals became more community oriented over time in response to these regulations, which is a worthy policy goal. Policymakers, researchers, and advocacy organizations from across the political spectrum have argued that community benefit regulations need to be reformed, if not thrown out entirely, in light of egregious NPH behavior that goes against the community benefit ethos, such as excessive measures to collect on patient debt.97–99 This research is agnostic when it comes to whether a significant overhaul of the community benefit 108 regulations is necessary for greater hospital accountability from their communities. It does suggest, however, that requiring NPHs to regularly interact with community-based organizations and agencies in some capacity may have led to a holistic change in how hospitals interact with - and possibly even invest in – community organizations and agencies. As analysis of the community orientation measure showed at the end of Chapter 5, NPHs that had the most integrated, decentralized, and heterogenous CHNA networks were those that had the highest community orientation scores, often lead by local health agencies (LHAs). There are also potential policy-related implications of this research for those operating on a local level. Municipal leaders that recognize the alignment between their priorities for their communities and the needs of their local hospital(s) are likely to find willing clinical partners for their initiatives. The North Dakota-based rural hospital I described in Chapter 5 that worked with their local city leaders to improve housing and educational services in order to attract physicians to their community is one example. Establishing a shared local vision and priorities is also a way of holding NPHs to account for their obligations to their local communities. While I have noted many different implications for policymakers, hospitals should also note the ways in which their peers have benefited from community-oriented strategies. Research has found that hospitals that have been able to lower their unnecessary hospitalizations, for example, are generally hospitals that have improved local access to high-quality primary care; increased access to self-management supports and population health resources; and improved care coordination across a variety of support services.100 Hospital community orientation could potentially affect all of these 109 characteristics by increasing buy-in from community members via strengthened relationships with local organizations that have earned broad community trust. This could help hospitals increase the reach of their preventative health care services, encourage community members to access primary care, and/or ensure that vulnerable patients are better able to access needed social services to maintain their health. Chapter 5 noted the example of a New York hospital whose strategy became more community oriented in later CHNA cycles not by changing the types of chosen interventions from previously utilized prevention strategies, but by integrating community organizations to improve the reach of those prevention strategies. More research is needed to examine whether the hospitals most integrated with their communities also are exemplary at hospital quality management and patient engagement. Finally, for those researchers interested in examining how very different types of organizations with varying levels of power and resources can successfully collaborate, this study offers a starting point for further inquiry. As mentioned throughout this narrative, hospitals and community organizations and agencies are very different, with different types of employees, different sources of revenue, different sources of public pressure and motivations, etc. – with perhaps the only unifying factor a commitment to improving the health and well-being of their community. I have highlighted some examples of where hospitals have successfully worked with community-based organizations, particularly in Chapter 5, with health agencies a common lead partner for organizing and maintaining community-based coalitions. Having a strong, well-resourced leadership organization, such as an LHA, that is trusted by both NPHs and other stakeholders and committed in the short- and long-term to advancing community health, 110 may be what is needed to ensure NPHs’ implementation strategies are targeted effectively. This may necessitate that hospitals place some measure of faith in the abilities of their community partners, which will likely be difficult as evidenced by the results in Chapter 5; hospitals have strongly preferred to work independently of others and rely on their own resources in executing their implementation strategies. Implications for Future Community Benefit Policy Given these research findings suggesting NPHs have become more community oriented in how they conduct their CHNAs and implementation strategies over the past ten years, what are some potential changes to current community benefit regulation to ensure these gains in hospital community orientation continues or accelerates over the next CHNA cycles? As has been mentioned previously, there is a clear need for more oversight, and for NPHs to be held to account for their financial obligations to their communities in exchange for their significant tax exemptions. The findings from this research have posited that greater integration of community stakeholders into the CHNA process, particularly LHAs, has led to an increased likelihood of a highly community-oriented implementation strategy. Thus, utilizing community organizations to help hold NPHs accountable and willing to participate in community-based multisectoral partnerships may be a potential policy approach. LHAs in particular are well-positioned to help hold NPHs accountable, especially given that LHAs have their own requirement to conduct a community health 111 assessment.50 One possible policy approach to encourage greater NPH and LHA cooperation would be to require or encourage NPHs and LHAs to conduct their required assessments together. As discussed in Chapter 4, most CHNAs do include contributions from LHAs but for many, the contribution is minimal and often limited to identifying and prioritizing health issues and/or assisting with data analysis. It is very inefficient for two nationwide networks of different organization types that (1) are both dedicated to improving health and (2) also required to repeatedly assess their communities’ health status are not more regularly conducting assessments together. Other studies have confirmed that LHAs and NPHs do not collaborate on their respective CHNAs with any frequency - and even when they collaborated, the contribution from LHAs was minimal.101 In other words, NPHs were satisficing with the IRS requirements to consult with public health experts and felt no obligation to strengthen these relationships any further. Even if future research finds that there are no associated benefits for community health when NPHs and LHAs develop closer relationships, it would still be worthwhile for policymakers to ensure these two organizational types conducted their health assessments together. This may require, for example, extending the timeline for NPHs to conduct their CHNAs to every five years, rather than three, to match the assessment cycles for LHAs. Anecdotally, the asynchronous assessment cycles for NPHs and LHAs was a commonly cited reason in CHNA narratives for why NPHs did not conduct their CHNAs with their LHAs. As mentioned frequently throughout this study, a handful of states have led the way in requiring hospitals and health agencies to work together on their assessments, as it prevents a duplication of effort. 112 Such a policy change in how LHAs and NPHs conduct their assessments may also lead to greater NPH involvement in community health improvement plans (CHIPs), which are long-term community-wide improvement plans that are typically overseen by health agencies.41 Again, this research supports such a policy approach given the association between integrated CHNA networks and higher community orientation scores. Other studies, such as a 2018 study conducted by Carlton and Singh, likewise have found that jointly conducted CHNAs increases the probability of increased hospital involvement in CHIPs, as well as increased hospital investment in community infrastructure.15 Mandating (or at least strongly encouraging) NPHs to collaborate more often with community stakeholders like LHAs in order to engender more community oriented NPH implementation strategies via state and/or federal policy seems like a worthy policy pursuit, but to ensure it leads to more community-oriented NPHs, it must also be accompanied by a strong regulatory response. I have previously noted the lax oversight of the federal community benefit program provided by the IRS81, which has allowed NPHs to essentially decide their level of compliance with the community benefit regulations without fear of punishment. As an agency, the IRS is notoriously underfunded and overstretched,102 ensuring poor oversight over the NPH community benefit program. However, even with proper resources, the IRS is not a health or health care services agency, and so it likely does not have the capacity or expertise to confirm whether NPHs are, for example, implementing evidence-based interventions that actually benefit their communities or are pursuing strategies that are beneficial for increasing revenue but have a minimal effect on community health. Thus, other regulatory agencies with health- 113 related experience, such as the CDC or CMS, may have a role to play in helping the IRS ensure NPH compliance with community benefit regulations, by, for example, periodically auditing NPH’s submitted implementation strategies. New York is the clearest example of this need for stronger oversight and public investment. New York-based NPHs (at least those in my sample) are largely compliant with the state’s policy that requires NPHs to work with their LHAs on their assessment and a CHIP for each county,103 and this has led to greater investments in communities from NPHs.37 But stimulating such a response has required the New York Department of Health to take a top-down regulatory approach wherein the agency regularly develops a Prevention Agenda that specifies which health domains New York NPHs and LHAs can choose to prioritize, as well as identify possible evidence-based interventions for each health domain.77 The Prevention Agenda is also incorporated into other of New York’s state-based health reform, such as their Delivery System Reform Incentive Payment (DSRIP). The DSRIP is New York’s initiative to transform their Medicaid fee-for- service model to a value-based care payment model using community-based collaboration to advance a 25% reduction in avoidable hospitalizations over five years.77 The level of state involvement in aligning all of these initiatives is significant but necessary for achieving their policy goals. Future Research Directions Because this study examined a topic that has not been the subject of extensive research and used a unique source of data, there are no shortages of possible directions for future research. For one, future research that utilizes either the NPH network 114 typology and/or community orientation composite measure should further examine the potential broader implications of more community-oriented hospitals on health outcomes. In particular, breaking down the community orientation measure to determine which of its subcomponents might be most responsible for improvements in community health will be critical for helping determine where hospitals should invest their resources. Further examination of community-oriented NPHs’ thoughts on working with community organizations, both benefits and issues, might also clarify some of the mechanisms mentioned above regarding how and why community orientation might be affecting community health. This would also be helpful in assessing where and how successful coalitions have been able to work together over time and address some of the inter- organizational issues that might arise with continued collaboration. From a policy perspective there likewise is no shortage of possibilities for future research. Of particular importance is determining whether states like New York and Ohio that have advanced policies meant to induce greater NPH and LHA collaboration and increase NPH community orientation have seen any corresponding improvements in community health relative to other states. I also think it is important to consider the ways in which other federal policy initiatives, particularly APMs, have dovetailed with implementation strategies. For example, to what extent are NPHs leveraging relationships formed via the CHNA and implementation strategy process to improve their profitability of APMs? There is also a need to examine whether community orientation of a hospital has other potential impacts and/or extends to other hospital practices. This includes determining if NPHs that are more community orientated have more equitable internal 115 hiring practices, for example, or are less litigious with their poorer patients because of non-payment. As mentioned above, LHAs have been a particularly strong partner for community-oriented hospitals, and further examining the potential impacts of community orientation on LHAs’ capacity and quality of service should be a point of future research as well. Are agencies better able to pursue their mission with hospitals as partners and provide higher quality services? How have agencies been able to integrate hospital resources, if at all, into their own programming? Answering these questions will determine if other organizations like health agencies have been particularly integral to achieving positive change in community health. Study Limitations This research approach has a number of limitations that are important to note. First, in terms of the adopted data collection methodology, the use of content analysis relies on our interpretation of the narrative contained in each implementation strategy. These interpretations might not be correct, either because they are misrepresented in the text or misinterpreted by the team of coders. In addition, a larger sample might have improved the rigor of the results, although the current sample size was likely sufficient for preliminary results. The large number of hospitals that were missing one or more of their implementation strategies – and thus, excluded from my sample – may also be affecting these results. While my sample appears to be similar to the larger population of U.S. NPHs, there may also be something systematically different about the NPHs with one or more missing CHNA and/or implementation strategies that could affect the final 116 results. Those NPHs with missing CHNAs, for example, are not in compliance with the requirements of the community benefit regulations; those with missing implementation strategies are technically compliant but are being less transparent than other NPHs about their plans to address priority health issues. NPHs opting to be exiguous in following their community benefit regulatory obligations are also likely to be less community- oriented relative to other NPHs. The finding that NPHs became more community oriented over time may actually be only relevant among compliant NPHs. There are also limitations to note within each research chapter. In Chapter 4, I used the compiled data to develop three measures of NPH networks, which were then grouped into different homogenous groups using cluster analysis. As a methodological approach, cluster analysis has been criticized for finding structure where there is none74, and while both the Calínski and Harabasz pseudo-F index and the Duda and Hart index suggested a five-group model was the most parsimonious result, it may be that these three measures were not enough to differentiate the various collaborative approaches used to complete a CHNA. The comparatively poor interrater reliability measures for the data used for this analysis also suggested a need to clarify definitions and criteria for extracting data from the CHNA reports. As noted in Chapter 3, much of the error stems from missing key narrative within each report detailing relevant contributions of organizations, which in turn exacerbated differences between my collected data and the data collected by my research assistants. However, for this research chapter, I have used data entirely collected by myself, which at the very least maintains some consistency. 117 In Chapter 5, I weighed a number of different methodologies for characterizing community orientation of NPHs using five measures and developing a composite index measure to reflect this characteristic, finally deciding on a CFA approach. I felt that this was the appropriate method to adopt given the hypothesis that I was characterizing a latent characteristic of hospitals – not just that I was trying to consolidate measures. Other methods may have been better suited, particularly if my conceptualization of ‘community orientation’ as an NPH characteristic that had meaning beyond the implementation strategies is wrong. I also opted to drop the fifth measure related to community involvement in planning the implementation strategy due to low factor loadings. Again, this may have been the wrong approach, particularly if this fifth measure reflected low factor loadings due to, for example, measurement error (i.e., NPHs using community stakeholders to develop their implementation strategy but not mentioning it in their narrative). While I feel that I answered this particular limitation by highlighting why it might not have been a good measure of community orientation based on further analysis and reviews of the underlying narrative, it is still a limitation to note. Despite these limitations, this study provides significant evidence that the expanded community benefit regulations led to NPHs participating in community-based collective action to improve population health, which was a key policy goal. It also established that community orientation as reflected in the implementation strategies may have broader implications for community health, and that it is worth monitoring the contributions of NPHs via community benefit activities. 118 References 1. National Research Council, Committee on Population. US Health in International Perspective: Shorter Lives, Poorer Health. National Academies Press; 2013. 2. Braveman P, Gottlieb L. The social determinants of health: it’s time to consider the causes of the causes. Public health reports. 2014;129(1_suppl2):19-31. 3. Bradley EH, Canavan M, Rogan E, et al. Variation in health outcomes: the role of spending on social services, public health, and health care, 2000–09. Health Affairs. 2016;35(5):760- 768. 4. Bradley EH, Elkins BR, Herrin J, Elbel B. Health and social services expenditures: associations with health outcomes. BMJ quality & safety. Published online 2011:bmjqs- 2010. 5. Cramer GR, Singh SR, Flaherty S, Young GJ. The Progress of US Hospitals in Addressing Community Health Needs. Am J Public Health. 2016;107(2):255-261. doi:10.2105/AJPH.2016.303570 6. Mays GP, Scutchfield FD. Improving public health system performance through multiorganizational partnerships. Preventing chronic disease. 2010;7(6). 7. Prybil L, Scutchfield FD, Killian R, et al. Improving community health through hospital- public health collaboration: Insights and lessons learned from successful partnerships. Published online 2014. 8. Corrigan J, Fisher E, Heiser S. Hospital community benefit programs: increasing benefits to communities. Jama. 2015;313(12):1211-1212. 9. Internal Revenue Service. New Requirements for 501c3 Hospitals Under the Affordable Care Act | Internal Revenue Service. Accessed February 5, 2018. https://www.irs.gov/charities-non-profits/charitable-organizations/new-requirements-for- 501c3-hospitals-under-the-affordable-care-act 10. Hogg RA, Varda D. Insights into collaborative networks of nonprofit, private, and public organizations that address complex health issues. Health Affairs. 2016;35(11):2014-2019. 11. Singh SR, Bakken E, Kindig DA, Young GJ. Hospital community benefit in the context of the larger public health system: a state-level analysis of hospital and governmental public health spending across the United States. Journal of Public Health Management and Practice. 2016;22(2):164-174. 12. Stoto MA, Klaiman T, Davis M. A Community Health Needs Assessment Environment Scan. Available at SSRN 3230259. Published online 2018. 13. Mays GP, Mamaril CB, Timsina LR. Preventable Death Rates Fell Where Communities Expanded Population Health Activities Through Multisector Networks. Health Affairs. 2016;35(11):2005-2013. doi:10.1377/hlthaff.2016.0848 119 14. Isehunwa OO, Dobalian A, Ahn S, Relyea G, Carlton EL. Local Health Department and Hospital Collaboration Around Community Health Needs Assessment to Improve Health Outcomes. Family & Community Health. 2021;44(3):136-145. doi:10.1097/FCH.0000000000000280 15. Carlton EL, Singh SR. Joint Community Health Needs Assessments as a Path for Coordinating Community-Wide Health Improvement Efforts Between Hospitals and Local Health Departments. American journal of public health. 2018;108(5):676-682. 16. Conrad DA, Cave SH, Lucas M, et al. Community care networks: Linking vision to outcomes for community health improvement. Medical Care Research and Review. 2003;60(4_suppl):95S-129S. 17. New York State Department of Health. Prevention Agenda 2013-2018: New York State’s Health Improvement Plan. Accessed June 22, 2021. https://www.health.ny.gov/prevention/prevention_agenda/2013-2017/ 18. Santos T, Lee SYD, East C, Lindrooth RC. Can Collaboration Between Nonprofit Hospitals and Local Health Departments Influence Population Health Investments by Nonprofit Hospitals? Medical Care. 2021;Publish Ahead of Print. doi:10.1097/MLR.0000000000001561 19. Bazzoli GJ, Shortell SM, Dubbs N, Chan C, Kralovec P. A taxonomy of health networks and systems: bringing order out of chaos. Health services research. 1999;33(6):1683. 20. Barnett, K. Advancing the State of the Art in Community Benefit Toolkit: Instructional Materials and Templates for the Comprehensive Review and Enhancement of Community Benefit Programs. Public Health Institute; 2008. Accessed June 20, 2019. https://www.phi.org/resources/?resource=advancing-the-state-of-the-art-in-community- benefit-toolkit 21. Proenca EJ, Rosko MD, Zinn JS. Community orientation in hospitals: an institutional and resource dependence perspective. Health Services Research. 2000;35(5 Pt 1):1011. 22. Meyer JW, Rowan B. Institutionalized organizations: Formal structure as myth and ceremony. American journal of sociology. 1977;83(2):340-363. 23. Pfeffer J, Salancik GR. The External Control of Organizations: A Resource Dependence Perspective. Stanford University Press; 2003. 24. Oliver, C. Sustainable competitive advantage: combining institutional and resource-based views. Strategic Management Journal. 1998;18(9):697-713. doi:10.1002/(SICI)1097- 0266(199710)18:9<697::AID-SMJ909>3.0.CO;2-C 25. Proenca EJ, Rosko MD, Zinn JS. Correlates of hospital provision of prevention and health promotion services. Medical Care Research and Review. 2003;60(1):58-78. 26. Olden PC, Hoffman KE. Hospitals’ health promotion services in their communities: Findings from a literature review. Health care management review. 2011;36(2):104-113. 120 27. Jennings J, Landry AY, Hearld LR, Snyder SW, Maldonado RW, Patrician PA. Examining the Relationship between Community Orientation and Hospital Financial Performance. Journal of Organizational Psychology. 2017;17(2):50-60. 28. Orlitzky M, Schmidt FL, Rynes SL. Corporate social and financial performance: A meta- analysis. Organization studies. 2003;24(3):403-441. 29. Wang Q, Dou J, Jia S. A meta-analytic review of corporate social responsibility and corporate financial performance: The moderating effect of contextual factors. Business & Society. 2016;55(8):1083-1121. 30. Barnett ML. Stakeholder influence capacity and the variability of financial returns to corporate social responsibility. Academy of management review. 2007;32(3):794-816. 31. Zhao M, Carretta HJ, Hurley RE. Sole hospital commitment to health promotion and disease prevention (HPDP) services: does ownership matter? Journal of health and human services administration. Published online 2003:93-118. 32. Hendryx MS. Rural hospital health promotion: Programs, methods, resource limitations. Journal of community health. 1993;18(4):241-250. 33. Olden PC, Szydlowski SJ, Armstrong NG. Health promotion and disease prevention by small rural hospitals: Reasons, obstacles, and enablers. Journal of Healthcare Management. 2004;49(2):89. 34. Boscarino JA, Chang J. Nontraditional services provided by nonprofit and for-profit hospitals: Implications for community health. Journal of Healthcare Management. 2000;45(2):119-135. 35. Ginn GO, Moseley CB, Beck LM. Community health orientation, community-based quality improvement, and health promotion services in hospitals. Journal of Healthcare Management. 2004;49(5):293. 36. Shiell A, Hawe P, Kavanagh S. Evidence suggests a need to rethink social capital and social capital interventions. Social Science & Medicine. Published online September 8, 2018:111930. doi:10.1016/j.socscimed.2018.09.006 37. Santos T, Lee SYD, East C, Lindrooth RC. Can Collaboration Between Nonprofit Hospitals and Local Health Departments Influence Population Health Investments by Nonprofit Hospitals? Medical Care. 2021;59(8):687-693. 38. Bai G, Anderson GF. A more detailed understanding of factors associated with hospital profitability. Health Affairs. 2016;35(5):889-897. 39. Stemler S. An overview of content analysis. Practical assessment, research, and evaluation. 2000;7(1):17. 40. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. Journal of biomedical informatics. 2009;42(2):377-381. 121 41. Centers for Disease Control & Prevention. Assessment and Plans - Community Health Assessment - STLT Gateway. Published April 6, 2019. Accessed June 1, 2019. https://www.cdc.gov/publichealthgateway/cha/plan.html 42. Catholic Health Association of the United States. Assessing & Addressing Community Health Needs. Published online June 2013. Accessed June 1, 2019. https://www.chausa.org/docs/default-source/general-files/cb_assessingaddressing- pdf.pdf?sfvrsn=4 43. Association for Community Health Improvement. Community Health Assessment Toolkit. Accessed January 2, 2020. https://www.healthycommunities.org/resources/community- health-assessment-toolkit 44. Mays GP, Scutchfield FD, Bhandari MW, Smith SA. Understanding the organization of public health delivery systems: an empirical typology. The Milbank Quarterly. 2010;88(1):81-111. 45. Perry BL, Pescosolido BA, Borgatti SP. Egocentric Network Analysis: Foundations, Methods, and Models. Vol 44. Cambridge University Press; 2018. 46. Burns LR, Wholey DR. Adoption and abandonment of matrix management programs: Effects of organizational characteristics and interorganizational networks. Academy of management journal. 1993;36(1):106-138. 47. Westphal JD, Gulati R, Shortell SM. Customization or conformity? An institutional and network perspective on the content and consequences of TQM adoption. Administrative science quarterly. Published online 1997:366-394. 48. Greenhalgh T, Robert G, Macfarlane F, Bate P, Kyriakidou O. Diffusion of innovations in service organizations: systematic review and recommendations. The Milbank Quarterly. 2004;82(4):581-629. 49. Joint Research Centre-European Commission. Handbook on Constructing Composite Indicators: Methodology and User Guide. OECD publishing; 2008. 50. Public Health Accreditation Board. Seven Steps of Accreditation. Public Health Accreditation Board. Accessed May 13, 2022. https://phaboard.org/seven-steps-of- accreditation/ 51. Caliński T, Harabasz J. A dendrite method for cluster analysis. Communications in Statistics-theory and Methods. 1974;3(1):1-27. 52. Pett MA, Lackey NR, Sullivan JJ. Making Sense of Factor Analysis: The Use of Factor Analysis for Instrument Development in Health Care Research. sage; 2003. 53. Thompson B. Exploratory and confirmatory factor analysis: Understanding concepts and applications. Washington, DC. 2004;10694. 54. Schmitt TA. Current methodological considerations in exploratory and confirmatory factor analysis. Journal of Psychoeducational Assessment. 2011;29(4):304-321. 122 55. Hurley AE, Scandura TA, Schriesheim CA, et al. Exploratory and confirmatory factor analysis: Guidelines, issues, and alternatives. Journal of organizational behavior. Published online 1997:667-683. 56. Bandalos DL, Finney SJ. Factor analysis: Exploratory and confirmatory. In: The Reviewer’s Guide to Quantitative Methods in the Social Sciences. Routledge; 2018:98-122. 57. Shelby LB. Beyond Cronbach’s alpha: Considering confirmatory factor analysis and segmentation. Human dimensions of wildlife. 2011;16(2):142-148. 58. Vaske JJ, Beaman J, Sponarski CC. Rethinking internal consistency in Cronbach’s alpha. Leisure Sciences. 2017;39(2):163-173. 59. Tobias S, Carlson JE. Brief report: Bartlett’s test of sphericity and chance findings in factor analysis. Multivariate behavioral research. 1969;4(3):375-377. 60. Kaiser HF. An index of factorial simplicity. Psychometrika. 1974;39(1):31-36. 61. Schreiber JB. Issues and recommendations for exploratory factor analysis and principal component analysis. Research in Social and Administrative Pharmacy. 2021;17(5):1004- 1011. 62. Zygmont C, Smith MR. Robust factor analysis in the presence of normality violations, missing data, and outliers: Empirical questions and possible solutions. The Quantitative Methods for Psychology. 2014;10(1):40-55. 63. Wieland A, Durach CF, Kembro J, Treiblmaier H. Statistical and judgmental criteria for scale purification. Supply Chain Management: An International Journal. Published online 2017. 64. Ximénez C. A Monte Carlo study of recovery of weak factor loadings in confirmatory factor analysis. Structural Equation Modeling. 2006;13(4):587-614. 65. Brown TA, Moore MT. Confirmatory factor analysis. Handbook of structural equation modeling. Published online 2012:361-379. 66. Perry JL, Nicholls AR, Clough PJ, Crust L. Assessing model fit: Caveats and recommendations for confirmatory factor analysis and exploratory structural equation modeling. Measurement in physical education and exercise science. 2015;19(1):12-21. 67. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural equation modeling: a multidisciplinary journal. 1999;6(1):1-55. 68. Meredith W, Horn J. The role of factorial invariance in modeling growth and change. Published online 2001. 69. Bontempo DE, Hofer SM, Lawrence F. Assessing factorial invariance in cross-sectional and longitudinal studies. Handbook of methods in positive psychology. 2007;153175. 123 70. Vandenberg RJ, Lance CE. A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational research methods. 2000;3(1):4-70. 71. Stata. How can I check measurement invariance using the sem command? | Stata FAQ. Accessed December 18, 2021. https://stats.idre.ucla.edu/stata/faq/how-can-i-check- measurement-invariance-using-the-sem-command-stata-12/ 72. Gregorich SE. Do self-report instruments allow meaningful comparisons across diverse population groups? Testing measurement invariance using the confirmatory factor analysis framework. Medical care. 2006;44(11 Suppl 3):S78. 73. Acock AC. Discovering structural equation modeling using Stata. Stata Press Books. Published online 2013. 74. Gorman BS, Primavera LH. The complementary use of cluster and factor analysis methods. The Journal of experimental education. 1983;51(4):165-168. 75. Sheppard AG. The sequence of factor analysis and cluster analysis: Differences in segmentation and dimensionality through the use of raw and factor scores. Tourism analysis. 1996;1(1):49-57. 76. Ward Jr JH. Hierarchical grouping to optimize an objective function. Journal of the American statistical association. 1963;58(301):236-244. 77. Libman K, Ukeje C, Griffin K. Alignment across Hospital Investments toward Building a Culture of Health in New York State. https://www.nyam.org/publications/publication/alignment-across-hospital-investments- toward-building-culture-health-new-york-state/ 78. Wood DJ, Gray B. Toward a comprehensive theory of collaboration. The Journal of applied behavioral science. 1991;27(2):139-162. 79. Fichtenberg C, Delva J, Minyard K, Gottlieb LM. Health And Human Services Integration: Generating Sustained Health And Equity Improvements: An overview of collaborations, partnerships, and other integration efforts between health care and social services organizations. Health Affairs. 2020;39(4):567-573. 80. Selsky JW, Parker B. Cross-sector partnerships to address social issues: Challenges to theory and practice. Journal of management. 2005;31(6):849-873. 81. Office USGA. Tax Administration: Opportunities Exist to Improve Oversight of Hospitals’ Tax-Exempt Status. Accessed December 18, 2021. https://www.gao.gov/products/gao-20- 679 82. Joint Research Centre-European Commission. Handbook on Constructing Composite Indicators: Methodology and User Guide. OECD publishing; 2008. 83. Metge CJ, Chateau D, Prior H, Soodeen R, De Coster C, Barre L. Composite Measures/Indices of Health and Health System Performance. Manitoba Centre for Health Policy, Department of Community Health Sciences …; 2009. 124 84. Horn JL. An empirical comparison of methods for estimating factor scores. Educational and Psychological Measurement. 1965;25(2):313-322. 85. DiStefano C, Zhu M, Mindrila D. Understanding and using factor scores: Considerations for the applied researcher. Practical Assessment, Research, and Evaluation. 2009;14(1):20. 86. Castrucci B, Auerbach J. Meeting individual social needs falls short of addressing social determinants of health. Health Affairs Blog. 2019;10(10.1377). 87. American Hospital Association. Social Determinants of Health Series: Housing and the Role of Hospitals. Accessed June 17, 2022. https://www.aha.org/ahahret-guides/2017-08- 22-social-determinants-health-series-housing-and-role-hospitals 88. Santos T, Lindrooth RC. Nonprofit Hospital Community Benefits: Collaboration With Local Health Departments to Address the Drug Epidemic. Medical Care. 2021;59(9):829- 835. doi:10.1097/MLR.0000000000001595 89. Santos T. Non-profit hospital targeted health priorities and collaboration with local health departments in the first round post-ACA: a national descriptive study. Frontiers in Public Health. 2020;8:124. 90. Berwick DM. Screening in health fairs: A critical review of benefits, risks, and costs. Jama. 1985;254(11):1492-1498. 91. Naik Y, Baker P, Ismail SA, et al. Going upstream – an umbrella review of the macroeconomic determinants of health and health inequalities. BMC Public Health. 2019;19(1):1678. doi:10.1186/s12889-019-7895-6 92. Rozier M, Enard K. Leveraging Community Benefit To Increase Health Care Value. Health Affairs Forefront. Published May 26, 2022. Accessed August 17, 2022. https://www.healthaffairs.org/do/10.1377/forefront.20220519.808327/full/ 93. Chen M, Unruh M, Pesko M, et al. Hospitals’ engagement in population health: moving past the medicine and into the community. Health Affairs Blog. Published online 2016. 94. Propheter G. Property Tax Exemptions for Nonprofit Hospitals: What Are They Worth? Do They Earn Them? Evidence From New York City. Public Budgeting & Finance. 2019;39(2):3-22. 95. Rosenbaum S, Kindig DA, Bao J, Byrnes MK, O’Laughlin C. The value of the nonprofit hospital tax exemption was $24.6 billion in 2011. Health Affairs. 2015;34(7):1225-1233. 96. Young GJ, Flaherty S, Zepeda ED, Singh SR, Rosen Cramer G. Community benefit spending by tax-exempt hospitals changed little after ACA. Health Affairs. 2018;37(1):121- 124. 97. Crossley M. Health and taxes: hospitals, community health and the IRS. Yale J Health Pol’y L & Ethics. 2016;16:51. 98. Caffrey A, Pointer C, Steward D, Vohra S. The role of community health needs assessments in medicalizing poverty. Journal of Law, Medicine & Ethics. 2018;46(3):615-621. 125 99. Buxton ZJ. Community Benefit 501 (R) edux: An Analysis of the Patient Protection and Affordable Care Act’s Limitations Under Community Benefit Reform. . Louis UJ Health L & Pol’y. 2013;7:449. 100. Muenchberger H, Kendall E. Predictors of preventable hospitalization in chronic disease: Priorities for change. J Public Health Pol. 2010;31(2):150-163. doi:10.1057/jphp.2010.3 101. Beatty KE, Wilson KD, Ciecior A, Stringer L. Collaboration among Missouri nonprofit hospitals and local health departments: content analysis of community health needs assessments. American journal of public health. 2015;105(S2):S337-S344. 102. Fort D. Perspective | The IRS is underfunded, but it needs more than cash to stop tax cheats. Washington Post. https://www.washingtonpost.com/outlook/2021/05/06/irs-tax-evasion- funding/. Accessed June 17, 2022. 103. New York State Department of Health. Prevention Agenda 2013-2018: New York State’s Health Improvement Plan. Accessed April 14, 2022. https://www.health.ny.gov/prevention/prevention_agenda/2013-2017/ 126 Appendix A: Detailed Codebook I. Introduction This code book provides general directions and guidance to coding the set of community health needs assessments (CHNAs) and implementation strategies plans gathered. It also provides detailed set of definitions for key concepts, which will guide coders as they categorize information gleaned from reports. II. Directions Coders should also have access to the data repository of CHNA and IS reports in Google Docs as well as the list of hospitals in Excel. This list will note what reports should be available, as well as relevant notes on the reports. A few things coders should note on the set of files available in each folder:  I’ve done my best to ensure everything is accurate, including the name of the hospital, the availability of reports, dates of reports, etc. However, I’ve likely made some mistakes – especially as it pertains to the availability of IS plans, which are often included with the CHNA report, and which I might’ve missed. One common issue is the frequent name changes hospitals undergo (e.g., because they’ve changed ownership, or because they want to brand themselves differently). I’ve tried to note where hospitals have changed names, but the easiest way to check is to simply Google the name of the hospital on the report. Google can generally come up with the newest name of the hospital, which hopefully matches later reports.  What reports are available in each file will vary quite a bit, even from year to year with the same hospital. For example, one year a hospital might opt to publish the IS plan separately and the next year it is part of the CHNA. I’ve also downloaded most everything linked on a hospital’s webpage, which occasionally can include superfluous documents. For example, many hospitals work with local public health agencies to produce concurrent or separate CHNAs, which may or may not be relevant to the hospital’s CHNA or IS plan. There is absolutely no need to read each document thoroughly – if you’re able to get all relevant information from a given CHNA and IS plan, you can take a quick look at what is in other documents to see if it is relevant and move on. All coders should have access to the instrument in REDCap, the software that we’re using to gather information. 1. For each assigned hospital, find the relevant folder with all available CHNA / IS reports. All files are organized by state, so first note the state associated with the hospital, then the hospital name. Within each hospital’s folder, I have also organized the reports by the year they were released. Because the implementation plans are connected to the CHNA by the identified health priorities, I’ve kept the plans in the same ‘Year’ folder as the associated CHNA, even if the plan was released in a different year (for example, if a CHNA was conducted in 2013 but the implementation plan came out in 2014, I would still keep the documents together in the ‘2013’ folder). To the extent that you can, check to make sure you have the correct hospital’s reports for the correct year(s). 2. I also recommend briefly reading through the report to gain a reasonably good understanding for what the report includes and doesn’t include, where you might be able to find the relevant 127 information for our study, if the report covers multiple hospitals, etc. I’ve found using a highlighter and/or comments to select relevant passages on an initial reading to be helpful. 3. Answer each question asked in the instrument to the best of your ability. 4. I recommend starting with the earliest CHNA / implementation strategy for a given hospital, and then once you’ve finished entering it, moving on to that same hospital’s next CHNA / implementation strategy in chronological order. 5. You may need to do a little research on each organization that qualifies as a stakeholder. I recommend Google; most organizations should be easily found. III. Definitions The following definitions should be used to help guide coders as they enter information into each instrument.  CHNA: From the CDC, “A community health assessment (sometimes called a CHA), also known as community health needs assessment (sometimes called a CHNA), refers to a state, tribal, local, or territorial health assessment that identifies key health needs and issues through systematic, comprehensive data collection and analysis.” For the purposes of this research, a CHNA needs to be associated with a hospital and be used to fulfill the requirements of non-profit hospitals under the Affordable Care Act (ACA). It is acceptable for hospitals to use CHNAs conducted by other entities, such as public health agencies, to fulfill their own requirements under the ACA, as long as the hospital participates.  Implementation Strategy (IS): From the IRS regulations: “A hospital facility’s implementation strategy must be a written plan that, for each significant health need identified, either: o Describes how the hospital facility plans to address the health need, or o Identifies the health need as one the hospital facility does not intend to address and explains why it does not intend to address the health need. o Although an implementation strategy must consider all of the significant health needs identified through a hospital facility’s CHNA, the implementation strategy is not limited to considering only those health needs and may describe activities to address health needs that the hospital facility identifies in other ways.” For the purposes of this research, implementation strategies must identify the priority health issue or issues each activity is going to address or has addressed.  Stakeholder: A person or organization that has an interest in another organization and its activities, strategies, etc. For the purposes of this research, we are defining relevant stakeholders as those who are active participants in the CHNA / IS process. This means that passive participants (e.g., survey respondents, key informant interviews) are not enough to “count” as a stakeholder. Attending meeting(s), helping to plan or actively participate in an intervention – even if that organization is not primarily responsible for completing the activity or intervention – would suffice to count as a participant. One further note: occasionally, it’s challenging to differentiate between an individual hospital’s participation and their parent system’s participation. Many times, a health system will conduct a CHNA, for example, with minimal or even no input from the reporting hospital. This is fine. Where this has occurred, I have checked the “Hospitals in the same network / system” box and noted that it was the system completing the activity (e.g., “Kaiser Permanente (system)”. 128 We’ve tried to identify who relevant participants are for each activity below. CHNA and IS Activities: The set of activities described in the instrument are as follows: (1) Activity 1: Plan and Define CHNA: This initial activity can encompass a range of activities necessary to start a CHNA, including: (1) Defining the scope of the CHNA; (2) Defining the community; (3) Defining and engaging partners; (4) Assigning leadership / oversight roles; (5) Identifying models to follow for the assessment (e.g., MAPP); (6) Hiring consultants to help with the conduct of the CHNA. Relevant participants should be reasonably clear to identify, as hospitals will typically identify the committee overseeing this initial process (and other processes) and there is very little possibility of passive participation. Note that this activity will most often be conducted solely by hospitals, but there can be instances where other organizations play a significant role, such as when following the lead of a public health agency. Both ACHI and the Catholic Health Association break this initial step into 2-3 activities; however, after reviewing many CHNAs, based on the amount of information typically found that describes these initial processes, it makes sense to combine these activities into one. (2) Activity 2: Collect and analyze data: This activity can again cover several steps, including: (1) determining appropriate population health measures; (2) collecting available quantitative secondary data and/or statistics from public sources; (3) collecting quantitative and/or qualitative primary data from local constituents through surveys, key informant interviews, focus groups, etc; and (4) analyzing resultant data. Again, I am grouping data collection and analysis together because they are usually treated as a singular activity in CHNA reports. (3) Activity 3: Identifying priority health issues: This activity involves identifying the most pressing community health needs using data collected from Activity 2. Note that hospitals will often ask individual stakeholders what they feel is/are the most pressing health issue(s) as part of their data collection effort, and then using this information to identify their priority health issues. From the hospital’s perspective, the former activity (data collection) should be considered part of Activity 2; the latter activity (assuming the respondents fit the criteria of ‘active’) part of Activity 3. Occasionally the distinction is a little difficult to discern within the narrative. Note also that often stakeholders will contribute data that is then used to identify priority health issues – how stakeholders contribute this data is relevant to determining whether they are part of a hospital’s collaboration network. Relevant stakeholders are again those who are active participants, i.e., those who are not just responding to an interview request or survey, but, for example, attend a meeting where priority health issues are identified, or attend a focus group. For a stakeholder to “count”, they must do more than just share their opinion. In the case of a community forum or focus group, stakeholders must travel to a specific location; this is sufficiently active to count for the purposes of this research. (4) Activity 4: Communicate results: The final step in conducting a CHNA is to communicate the results internally and externally. This can involve: (1) writing up the final CHNA report; (2) discussing the results internally, including with a hospital’s board, who must approve the final CHNA document; (3) sharing the results with external stakeholders (e.g., local media, partnering community organizations); (4) sharing with relevant federal / state / local regulatory agencies, especially if required to share by state law, such as in Maryland. 129 (5) Activity 5: Plan implementation strategies: Based on the CHNA results, hospitals must plan if and how they are going to address each health issue – i.e., what resources are already available to utilize, what resources can hospitals allocate to address each health issue, etc. This process may be done internally, or it may be done collaboratively with community partners. If hospitals opt to not address an issue, they must also describe why. I think it is important to clarify here: this activity encompasses the overall planning process of the implementation strategy. It does NOT include one or more activities of the implementation strategy that may include planning activities. For example, many implementation strategies include planning activities (e.g., “Discuss with Stakeholder X about how best to implement program” or “Explore the possibility of doing Y program”); these activities wouldn’t be considered a part of Activity 5. However, narrative that describes the initial planning process for deciding what the implementation plan should look like and who to partner with would be part of this activity. (6) Activity 6: Implement strategies: This step puts into action the identified plan from the previous step to address prioritized community health needs. These activities can range from planning (e.g. “Discuss with Stakeholder X about how best to implement program”; “Explore the possibility of doing Y program”) and collaboration (e.g., “Form a network of stakeholders to address chronic disease prevention”), to the actual implementation of programming. It is important to note that many hospitals will describe their planned implementation strategy prospectively – i.e., what they are planning on doing in the subsequent year(s) to address their priority health issues. Other hospitals will relay what they did retrospectively to address their priority health issues as part of the required evaluation of their strategy (which, per the IRS regulations, must be included in the subsequent CHNA report – see Activity 7, below). Some hospitals will do both. Where hospitals report both a retrospective and prospective description of their implementation strategy activities, report on the prospective description. This is because there are comparatively few retrospective descriptions of their implementation strategies and I think it makes sense to remain as consistent as possible. This will likely be the hardest activity to identify which stakeholders are active participants. For example, often hospitals will send medical professionals to talk at schools, clubs, businesses, etc. Often it will be hard to discern whether a stakeholder is an active participant because of vague language, particularly because IS plans are often projections of what hospitals would like to accomplish. For now, the criteria we’ll be using is:  If a particular setting is targeted (e.g., schools, workplaces) where it is clear the hospital is responsible for implementation, the targeted stakeholder groups should count as active stakeholders, but should only be counted as having secondary responsibility.  If hospitals use phrases such as “partner with” or “plan with”, I think that is sufficient to inferring shared primary responsibility.  Occasionally, a hospital’s implementation strategy will list individual departments within that hospital involved in implementing a given strategy. I have NOT been listing these departments, since they are part of the hospital; however, if representatives from the health SYSTEM are involved in a given strategy, then I have included the system as a separate stakeholder from the hospital. 130  Where there is vagueness, I recommend erring on the side of NOT including stakeholders. Because the implementation strategy is comprised of a series of defined activities, we will be collecting information on stakeholder involvement per defined activity, where available in the narrative. This will allow better Level of Effort calculations when it comes time for analysis. If the report narrative does not describe specific stakeholder roles (but DOES include a list of stakeholders involved), we will still collect data for this activity (see Q11 – Nonrepeating, below in Form 1). (7) Activity 7 - Evaluate strategies: Hospitals must also assess their chosen implementation strategies and report on their results in the subsequent CHNA report. The evaluation may be done internally by hospital staff, or externally by hired consultants, or some other arrangement entirely. As with implementation plans, hospitals may report on their evaluation prospectively (how they’re planning on evaluating their strategies in the coming years) and/or retrospectively (how they have conducted their evaluation). Unlike with Activity 6, however, we are only collecting information from retrospective narrative that describes what was done. This is because prospective evaluation narratives seem to be rare among CHNA reports, and for those that describe their future evaluation plans, there is very little specificity. Stakeholder Groups: (1) Reporting Hospital: The hospital responsible for producing the CHNA / IS reports. (2) Hospitals in same network / system: Those hospitals taking part in a collaborative CHNA / IS process and in the same network or system as the reporting hospital. Please note: where an activity is implemented at a systems-level, rather than at a hospital-level, check only this box, and not ‘Reporting Hospital’. This will allow for a more accurate calculation of my centralization measure. (3) Unaffiliated hospital: Those hospitals participating in a collaborative CHNA / IS process and unaffiliated with the reporting hospital. This will often happen with regional CHNA / IS’s, where multiple area hospitals join together to produce a single CHNA report. Note: this is inclusive of specialty hospitals that provide inpatient services (e.g., children’s hospitals, long-term care facilities). (4) Physician practices: Single-specialty or multispecialty group practice with at least two practicing physicians. (5) Mental health / substance abuse providers: [Definition from NAMI]: These provider / provider organizations can either offer therapy or assessment services (i.e., psychologists, councilors, clinical social workers, therapists) and/or prescribe medications (i.e., psychiatrists, mental health nurse practitioners, primary care providers, psychiatric pharmacists). Substance use organizations employ many of the same types of professionals, but also focus on addiction prevention, intervention, treatment, recovery support, and education. (6) Community health centers / FQHCs [Definition from HRSA]: Community health centers (CHCs) are private, nonprofit organizations that directly or indirectly (through contracts and cooperative agreements) provide primary health services and related services to residents of a 131 defined geographic area that is medically underserved. FQHCs are community-based health care providers that receive funds from the HRSA Health Center Program to provide primary care services in underserved areas. They must meet a stringent set of requirements, including providing care on a sliding fee scale based on ability to pay and operating under a governing board that includes patients. Federally Qualified Health Centers may be Community Health Centers, Migrant Health Centers, Health Care for the Homeless, and Health Centers for Residents of Public Housing. (7) Other provider organizations: This can be a range of organizations whose employees provide direct patient care, including pharmacies, post-acute care providers, SNFs / nursing homes, home health agencies, dentists. (8) Private health insurers / managed care organizations: Any private organization that provides coverage for health care services or provides for the delivery of services through contracted arrangements (e.g., from Medicaid) that accept a set per member per month (capitation) payment for these services. (9) State health agencies: This can include any state agency devoted to health – public health, rural health, the state Medicaid agency, health statistics, etc. (10) Local health agencies: Again, this can include any non-state / federal agency devoted to health. It is important to note here, that this can include satellite offices of health serving a particular county or region, even if it serves under the auspices of the state health agency. (11) Other state gov't agencies: (12) Other local gov't agencies: Can include locally elected officials, as well as other agencies, such as housing / development agencies. (13) Federal gov't agencies (14) Faith communities: Note that this does not include community organizations with a religious mission. (15) Tribal organizations (16) Criminal justice / police system (17) Schools (Daycare-12): Can be public or private (should also include daycares in this group). (18) Colleges / Universities: This can include research centers, cooperative extensions, etc. (19) Other community-based non-profits and foundations: These can be private social service agencies, religiously-affiliated organizations that fill certain roles, as well as foundations that work on a local, state, or regional level. (20) Private businesses (21) Private consultants: These types of businesses are often hired to conduct some or all of the CHNA and IS analysis, organization, planning activities, and we’ll probably run into them quite often. Note that researchers from universities are also often hired to fill a similar role and it’ll be important to make this distinction. (22) Advocacy & professional associations: These can include local hospital agencies, chambers of commerce, or larger national organizations, such as the American Hospital Association. (23) Public: Any individual who does not represent a particular organization, agency, or business. (24) Others: Stakeholder Roles: The following categories describe the relative level of effort put forth by hospitals and their partners on each of the seven CHNA / IS categories, ranging from minimal involvement (advisory) in a given activity, to maximum involvement (implementation). (1) Primary responsibility: The highest level of involvement in a given activity. Here, an organization or agency must be the entity responsible for completing a given activity. This is 132 clearer for some activities (e.g., data collection and analysis) than others (e.g., planning the CHNA or the implementation strategy). For the latter activities, I think it is important to consider the nature of what is accomplished by the end of each activity. Where the end goal of an activity is a plan, those responsible for completing the necessary steps to coming up with a plan have primary responsibility; planning may also be one of the steps taken as part of other activities (such as planning the data collection or analysis, or planning as an implementation strategy) as well. (2) Secondary responsibility: The minimal level of involvement for a given activity. Each of the CHNA and IS activities listed above can often involve hospitals soliciting advise from different stakeholders. For example, many hospitals may put together an advisory committee comprised of different stakeholders to periodically provide advise on all or some of the CHNA activities. Many hospitals will also gather different stakeholders for a meeting in determining the most pressing health needs; those participating organizations have secondary responsibility. Note that this could also include organizations overseeing other entities as they complete a task. For example, any time a hospital pays money to a private consultant or organization to conduct a particular activity, that hospital would be providing oversight responsibilities, which is secondary to the consultants, who are actually responsible for completing the activity. Level of responsibility might be most difficult to discern for implementing IS activities. Where an implementation strategy describes “partnering” with an organization, I think it makes sense to assign both organizations primary responsibility. However, if an organization “participates” in an activity where another organization is leading, I think here the “participating” organization should be assigned secondary responsibility Social Determinants: Hospitals will address one or more determinants of health through their choice of intervention. Most will opt to address their priority health issues by focusing on health and health care as a determinant; others may seek ways of addressing non-health determinants. We will use the Healthy People 2020 categorization of determinants to classify what determinants hospitals are seeking to address. (1) Economic stability: Interventions that address employment; food insecurity; housing instability; and poverty (2) Education: Interventions that address / support early childhood education and development; enrollment in higher education; high school graduation; and language and literacy (3) Social & community context: Interventions that address / support civic participation; discrimination; incarceration; and social cohesion (4) Neighborhood & built environment: Interventions that support access to foods that support healthy eating patterns; crime and violence; environmental conditions; and quality of housing (5) Health & health care: Interventions that address/support access to health care; access to primary care; and health literacy (6) Other: This seems like a good spot for coalition building or activities where it’s not clear what is being addressed. Prevention Level: Most interventions that address health as a determinant (which will be most activities) will be focused on prevention. These definitions are from the CDC: 133 (1) Primary prevention: Intervening before health effects occur, through measures such as vaccinations, altering risky behaviors (poor eating habits, tobacco use), and banning substances known to be associated with a disease or health condition. I also include expansion of primary care services and/or outreach efforts to expand the number of insured, since these measures seek to improve the health of all community members, healthy or not. (2) Secondary prevention: Screening to identify diseases in the earliest stages (i.e., at-risk), before the onset of signs and symptoms, through measures such as mammography and regular blood pressure testing (3) Tertiary prevention: Managing disease post diagnosis to slow or stop disease progression through measures such as chemotherapy, rehabilitation, and screening for complications. I also include measures such as chronic disease self-management; recruitment of specialty providers; or care coordination efforts for chronically ill patients, since all of these measures seek to slow or stop disease progression or complications. (4) Not preventive intervention (acute care): Occasionally, hospitals will say they’re addressing a particular priority health issue through the regular provision of acute care services (e.g., addressing chronic disease by maintaining an emergency room). (5) Not preventive intervention (other): This can be planning, collaboration efforts, environmental scans, etc. Health Priority: These are the categories of priority health issues identified by hospitals and then addressed through their implementation strategies. Categorization comes first, from the County Health Rankings Model and second, from the American Hospital Association’s CHNA Finder (note that the AHA’s tool uses the County Health Ranking Model to build their categories). I wanted a more granular categorization of medical conditions, which comprises the lion share of identified health issues, and which the CHNA Finder provides. (1) Access & availability of health care services: This can include, for example: (1) Availability - adequate and appropriate medical staffing, ability to get appointments/diagnostic testing/treatment when necessary, clinical staff recruitment, physical space for medical buildings; (2) Coverage & affordability of care - uninsurance and underinsurance rates, access to insurance programs (e.g., Medicaid, CHIP), affordability of insurance, affordability of care; (3) Emergency care - availability and accessibility of emergency department services; inappropriate use/over utilization of emergency department services; readmissions; (4) Mental health services - availability and accessibility of mental health services and providers; coverage by insurance; (5) Oral health care - availability and accessibility of dental care and providers, including screenings, fluoride treatment, dental exams and follow up, and preventive oral health; (6) Prenatal care - availability and accessibility of obstetricians and/or nurse midwives; (7) Primary and preventive care - availability and accessibility of primary care providers, screening services; (8) Specialty care - availability and accessibility of specialists; (9) Transportation - ability to reach the appropriate health care services; includes transportation availability and distance/time needed to reach appropriate facility. Hospitals may opt to categorize some of these categories (e.g., access to mental health or oral health services) under those particular priority health issues (2) Quality of care: High quality health care is timely, safe, effective, and affordable–the right care for the right person at the right time. Includes care coordination - coordination among providers and between systems; integration and coordination of care between providers across the care continuum; clinical integration; need for care navigators, community health workers, etc. 134 (3) Aging & elderly health conditions: end of life care, elder health issues (osteoporosis, arthritis, joint/chronic pain, dementia), depression, social isolation, fall prevention, Alzheimer’s and dementia (4) Behavioral & mental health: depression, bipolar disorder, personality disorders, ADHD, autism, affective disorders, anxiety-related mental disorders, emotional stability, schizophrenia, psychosis, eating disorders, and suicide (5) Cancer: All types of cancer. Include prevention, screening and treatment (6) Cardiovascular disease: heart and blood vessel diseases. Includes heart attack, heart failure, ischemic stroke, hypercholesterolemia, etc. Refers to acute events and prevention (7) Diabetes: Type 1 or 2 diabetes or pre-diabetes. Includes prevalence and screening (8) Disability: physical and mental disabilities (e.g., CP, Parkinson's, post-workplace injury) (9) Hypertension & stroke: uncontrolled hypertension, including hemorrhagic stroke (10) Infectious diseases: vaccination rates, influenza, pneumonia, hospital acquired infections (e.g., C difficile, CAUTIs, etc.) (11) Kidney disease: renal failure, dialysis rates. Includes prevalence, screening, treatment and prevention (12) Respiratory diseases: allergies, chronic asthma, COPD, chronic bronchitis (13) Sexually transmitted diseases: sexually transmitted infections including HIV/AIDS, HPV, herpes, etc. Includes prevention and treatment (14) Obesity: elevated BMI, includes overweight and obesity in children and adults (15) Chronic disease / community wellness: A general category for hospitals that just list ‘chronic disease’ or ‘community wellness’ as a priority health issue; chronic condition management, chronic disease health care services, education activities and oversight to help patients control their chronic disease and maintain a reasonable quality of life. Again, not disease specific (16) Tobacco use: habitual use of tobacco, including cigarettes, pipes, cigars, vape pens, etc. (17) Diet & exercise: consumption of healthy foods, including fruits and vegetables; understanding of healthy eating guidelines; access to and use of recreation and fitness facilities; promotion of active lifestyle (18) Alcohol & drug use: binge or heavy drinking, underage drinking, alcohol use by pregnant women; prescription and elicit drug use (e.g., heroin, meth, cocaine), opioid abuse, clean needles (19) Sexual activity: prevention of unwanted pregnancy; access to and use of contraceptives; teen pregnancy rates and prevention (20) Education: high school graduation rates, literacy rates, early childhood education, vocational training (21) Employment: job security, employment rates, availability of jobs (22) Income: poverty, living wage (23) Community safety: violence, trauma, unintentional injury and violent crimes (e.g., assaults, homicide, gun violence), burns, domestic violence, child abuse and bullying (24) Food security: Access and affordability of healthy food choices. Access to food stamps/SNAP (25) Environment: air and water quality, neighborhood recovery and restoration, safe space (26) Housing / transportation: access, affordability and safety of housing options (e.g., lead poisoning, mold, etc.); housing stability; (27) Maternal & child health (28) Other: 135 Advancing the State of the Art in Community Benefit Toolkit (ASACB) Questions: The final set of questions come from the Public Health Institute’s ASACB Toolkit and its five principals for hospitals to follow as they design their implementation strategy (three principals are covered with these questions and the remaining two are covered with other questions – I’ll eventually add a section describing how measures will be calculated with the information we collect). Much of the language for these definitions comes from this document. Community-Clinical Linkage: Ideally, hospitals will seek to establish a seamless continuum of care with how they design their set of implementation activities. This should mean, to the extent possible, clinician and community stakeholders working together to ensure community-based interventions are not siloed from clinicians and clinical service delivery interventions are not siloed from relevant community stakeholders – i.e., integrated into the respective workflows. To answer ‘Yes’ to the question ‘Does this activity describe a community-clinical linkage?’, the following criteria must be met:  The narrative describes links to clinical settings / stakeholders that occur regularly (e.g., regular meetings between clinical and community stakeholders; patient / client needs are regularly addressed by both community and clinical stakeholders), and  Involve BOTH clinical and community stakeholders in implementation and/or decision- making. For example, a health fair or a screening event put on by the hospital in a community setting would not be sufficient, even if it occurs regularly, because community stakeholders are not involved in how it is implemented; however, a community organization that partners with clinicians to regularly screen, educate, etc. the clients they serve would suffice because such services are integrated into both organizations’ workflow and both have a say in how it is implemented. Likewise, a clinically-based intervention that regularly refers patients to a community-based organization to fulfill other needs (with the full knowledge and partnership of that community organization) would also suffice. Disproportionate Unmet Health Needs (DUHN) Populations: Community benefit programs generally fall into one of two categories: programs that focus exclusively on vulnerable populations or programs that serve the community at large. For ASACB, vulnerable populations are clearly defined as populations that face financial or nonfinancial barriers to care (legal, transportation, language, culture, etc.) and/or have physical or psychological disabilities. If the targeted population does not meet this criterion (e.g., if it focuses on a general population subcategory such as women, children, or seniors), it should be placed in the community at large category for this question. To answer ‘Yes’ to the question ‘Does this activity target or include populations with disproportionate unmet health needs (DUHNs)?’, the following criteria must be met:  The narrative describes the DUHN population being targeted by describing the geographic parameters of the communities with DUHNs or the physical / mental disability faced by the targeted population OR  The narrative describes how the intervention, which may be targeted to the population-at- large, makes accommodations for DUHN populations It is important to note that, especially for interventions that address some specific mental or physical condition, the intervention would be considered targeted towards DUHN populations if it addresses the needs of that population outside the standard care for those populations. So, for example, treatment interventions that target those afflicted with substance use disorder wouldn’t be targeting a DUHN 136 population per se, but interventions that ensure those with substance use disorder have a primary care physician, or their food / security needs, would be considered targeting a DUHN population. DUHN Population Definitions: We also ask about how DUHN populations are defined, i.e., the criteria being used to define the barriers they face. Again, these criteria are from ASACB: (1) Financial barriers to access (e.g., uninsured, underinsured) (2) Language / cultural barriers to access (3) Documentation barriers (e.g., undocumented immigrants) (4) Lack of transportation (5) Physical disability / lack of physical mobility (6) Mental disability (7) Social isolation (e.g., seniors living alone) (8) Other Community Capacity Building: Community capacity building involves the strategic allocation of charitable resources (i.e., staffing, equipment, technical assistance, financial support, advocacy) to mobilize and build upon what is already in place in local communities. This approach will reinforce an ethic of shared accountability with community stakeholders, reduce duplication of effort, and increase the effectiveness and viability of community-based organizations. To answer ‘Yes’ to the question ‘Does this activity build upon existing community infrastructure?’, the following criteria must be met:  The implementing hospital is providing some measure of support to an existing community organization(s) and  The community organization plays a role in implementing the activity We also ask about the type of support being used to support community organizations: (1) Financial support (2) Technical assistance / education / training (3) Advocacy (4) Equipment or other material donation (5) In-kind support (e.g., use of hospital space) (6) Does not describe a provision of support 137 138 IV. Instrument Questions Form 1: Non-repeating form (answer these questions first) Question Response Type Possible Responses Q1: Coder (who was responsible for coding this report?) Drop-down list (one choice) (1) Hank; (2) Denise; (3) Other [NOTE: if ‘Other’, text box appears for coder to enter their name] Q2: Hospital State Drop-down list (one choice) (1) – (51): AL – WY + DC Q2a: Name of Hospital (STATE) Drop-down list (one choice) [LIST OF HOSPITALS FROM STATE] Q3: What year was the CHNA released? Drop-down list (one choice) (1) – (9): 2011-2019 Q4: What month was the CHNA released Drop-down list (one choice) (1) – (12): January - December Q5: Activity 1: Plan and define the scope of the CHNA (e.g., defining the community, identifying partners, etc) Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q5a; If ‘No’, proceed to Q6] Q5a: Types of organizations involved in Activity 1: Planning and defining scope of CHNA Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others 139 [Each checked box leads to Q5b and Q5c] Q5b: Activity 1: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q5c: Activity 1: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q6: Activity 2: Collect and analyze data Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q6a; If ‘No’, proceed to Q7] Q6a: Types of organizations involved in Activity 2: Collect and analyze data Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q6b and Q6c] Q6b: Activity 2: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q6c: Activity 2: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q7: Activity 3: Identify priority health issues Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q7a; If ‘No’, proceed to Q8] 140 Q7a: Types of organizations involved in Activity 3: Identify priority health issues Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q7b and Q7c] Q7b: Activity 3: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q7c: Activity 3: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q8: Activity 4: Communicate the CHNA results Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q8a; If ‘No’, proceed to Q9] Q8a: If yes to Activity 4, is report authorship the only evidence that this activity was completed? Yes / No (1) Yes; (0) No Q8b: Types of organizations involved in Activity 4: Communicate the CHNA results Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't 141 agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q8c and Q8d] Q8c: Activity 4: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q8d: Activity 4: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q9: Activity 5: Plan implementation strategy Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q9a; If ‘No’, proceed to Q10] Q9a: Types of organizations involved in Activity 5: Plan implementation strategy Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q9b and Q9c] 142 Q9b: Activity 5: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q9c: Activity 5: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q10: Activity 7: Evaluate strategies Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q10a; If ‘No’, proceed to Q11] Q10a: Types of organizations involved in Activity 1: Evaluate strategies Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q10b and Q10c] Q10b: Activity 7: what is the role of the [STAKEHOLDER]? Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q10c: Activity 7: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q11: Activity 6: Implement strategy Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q12; If ‘No’, survey is complete] Q12: Was the implementation plan Yes / No (1) Yes; (0) No [If ‘Yes’, proceed to Q13; If ‘No’, proceed to Q12a] 143 released on the same date as the CHNA? Q12a: If not, what year was the implementation plan released? Drop-down list (one choice) (1) – (9): 2011-2019 Q12b: And what month was the implementation plan released? Drop-down list (one choice) (1) – (12): January – December Q13: Does implementation plan provide sufficient details on planned activities? Yes / No (1) Yes; (0) No [If ‘Yes’, QXX-XX appear on implementation strategy repeating form; if ‘No’, they remain hidden] Q14: Do implementation strategy activities provide sufficient detail on stakeholders and their involvement in each activity? Yes / No (1) Yes; (0) No [If ‘Yes’, QXX-XX appear on implementation strategy repeating form; if ‘No’, they remain hidden and Q14a appears] Q14a: Does implementation strategy include a list of stakeholders who will participate in strategy, but without sufficient detail? Yes / No (1) Yes; (0) No [If ‘Yes’, Q15 – Option B appears, allowing you to fill in the list of stakeholders for all of Activity 6 (rather than for each implementation strategy activity); if ‘No’, survey is complete] The following questions are ONLY for when there is insufficient detail on stakeholders and their roles in the implementation strategy, but there IS a list of stakeholders that the reporting hospital cites as participants in their implementation strategy. Q11 (Non-repeating): Types of organizations involved in Activity 6: Implementing strategy Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; 144 (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community-based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q11a and Q11b] Q11a (Non-repeating): Activity 6: Implementing strategy Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q11b (Non-repeating): Activity 4: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] 145 Form 2: Repeating Form (for implementation strategy) – The following set of questions is repeated for EACH implementation strategy activity. After entering all of the information, select ‘Save & Go To Next Instance’ to open up the next repeating form. Once all implementation strategies have been entered, select ‘Save & Exit Form’ Question Response Type Possible Responses Q15: Copy and paste the first identifiable activity of the implementation strategy Notes [Copy and paste each implementation strategy activity in notes box] Q11 (Repeating): Types of organizations involved in Activity 6: Implementing strategy Checkbox (multiple choices) (1) Reporting Hospital; (2) Hospitals in same network / system; (3) Unaffiliated hospital; (4) Physician practices | 5, Mental health / substance abuse providers; (6) Community health centers / FQHCs; (7) Other provider organizations; (8) Private health insurers / managed care; (9) State health agencies; (10) Local health agencies; (11) Other state gov't agencies; (12) Other local gov't agencies; (13) Federal gov't agencies; (14) Faith communities; (15) Tribal organizations; (16) Criminal justice / police system; (17) Schools (daycare - 12); (18) Colleges / Universities; (19) Other community- based organizations and foundations; (20) Private businesses; (21) Private consultants; (22) Advocacy & professional associations; (23) Public; (24) Others [Checked boxes lead to Q11a and Q11b] Q11a (Repeating): Activity 6: Implementing strategy Drop-down list (one choice) (1) Primary responsibility; (2) Secondary responsibility; (3) Cannot tell involvement Q11b (Repeating): Activity 4: [STAKEHOLDER] Notes [Copy and paste name of stakeholder in notes box] Q16: What health priority is being addressed with this activity? Checkbox (multiple choices) (1) Access & availability of health care services; (2) Quality of care; (3) Aging & elderly health conditions; (4) Behavioral & mental health; (5) Cancer; (6) Cardiovascular disease; (7) Diabetes; (8) Disability; (9) Hypertension & stroke; (10) Infectious diseases; (11) Kidney disease; (12) Respiratory diseases; (13) Sexually transmitted diseases; (14) Obesity; (15) Chronic disease; (16) Tobacco use; (17) Diet & exercise; 146 (18) Alcohol & drug use; (19) Sexual activity; (20) Education; (21) Employment; (22) Income; (23) Community safety; (24) Food security; (25) Environment; (26) Housing / transportation; (27) Maternal & child health; (28) Oral health; (29) Emergency preparedness; (30) Other Q17: What social determinant is being addressed by this activity? Drop-down list (one choice) (1) Economic stability; (2) Education; (3) Social & community context; (4) Neighborhood & built environment; (5) Health & health care; (6) Other [If ‘Health & health care’ lead to Q17a; all other responses lead to Q18] Q17a: If addressing health & health care, what is the level of prevention for this activity? Drop-down list (one choice) (1) Primary prevention; (2) Secondary prevention; (3) Tertiary prevention; (4) Not preventative intervention (acute care); (5) Not preventative intervention (other) Q18: Does this activity describe a community- clinical linkage? Yes / No (1) Yes; (0) No Q19: Does this activity build upon existing community infrastructure? Yes / No (1) Yes; (0) No Q19a: What is the type of support being provided by the reporting hospital or health system for this activity? Checkbox (multiple choices) (1) Financial support; (2) Technical assistance / education / training; (3) Advocacy; (4) Equipment or other material donation; (5) In-kind support; (6) Does not describe a provision of support Q20: Does this activity target or include populations with disproportionate unmet health needs (DUHNs)? Drop-down list (one choice) (1) No, activity targets the community-at-large; (2) Yes, activity is targeted for the community-at-large, but describes provisions to include DUHN population(s); (3) Yes, activity targets DUHN populations; (4) Unable to tell who is the target population 147 Q20a: If yes, what DUHN population criteria does it meet (check all that apply)? Checkbox (multiple choices) (1) Financial barriers to access (e.g., uninsured, underinsured); (2) Language / cultural barriers to access; (3) Documentation barriers (e.g., undocumented immigrants); (4) Lack of transportation; (5) Physical disability / lack of physical mobility; (6) Mental disability; (7) Social isolation (e.g., seniors living alone); (8) Other