Browsing by Subject "Predictive Model"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Develop informatics solutions to deliver relevant information for clinical decision making that improve the management of cardiovascular disease risk of breast cancer patients
    (2025-01) Zhou, Sicheng
    In the realm of breast cancer treatment, balancing the efficacy of therapy against the risk of treatment toxicity particularly cardiotoxicity as one of the more morbid complications of treatment poses a significant clinical challenge. This thesis presents an innovative approach to addressing the issue of breat cancer-related cardiotoxicity by integrating health informatics with real world data from Electronic Health Records (EHRs) and other health information technology (e.g., imaging reports from cardiovascular information systems) to improve cardiovascular disease risk management in breast cancer patients. Our approach leverages deep learning algorithms and a range of natural language processing (NLP) approaches in order to extract, analyze, and interpret complex clinical data to enhance decision-making processes in healthcare settings for this vulnerable group of patients.At the core of this body of research is the development and application of transformer-based deep learning methods, which are specifically tailored to extract targeted information from clinical texts in EHRs. By creating a specialized cancer domain vocabulary, the study demonstrates the enhanced performance of these models in accurately identifying relevant clinical data, such as patient demographics, treatment details, and cancer phenotypes. This approach significantly advances the precision and reliability of extracting important clinical information, which often is a crucial step in developing robust predictive models. The thesis further explores the generalizability of these NLP algorithms across different healthcare institutions via external validations, a vital consideration given the varied nature of EHRs. Through cross-institutional evaluations, the research establishes the portability of these models, ensuring their effectiveness in diverse clinical environments. This aspect of the study is critical in validating the broader applicability of the developed methodologies in various healthcare settings. Central to the thesis is the creation of predictive models for assessing the risk of heart disease in breast cancer patients. Utilizing a deep learning approach, specifically LSTM-D models, the study effectively harnesses longitudinal EHR data to predict cardiovascular risks associated with cancer treatments. We find that these models outperform traditional methods, offering a more nuanced understanding of patient-specific risk factors and temporal patterns in treatment responses. The thesis' findings underscore the potential of integrating advanced data analysis tools in clinical decision-making, particularly in the context of breast cancer treatment. By providing a more detailed and personalized risk assessment, the research contributes significantly to the field of personalized medicine, enhancing the quality of patient care and treatment outcomes. Overall, this thesis bridges a critical gap in healthcare informatics by developing and validating innovative methodologies for extracting and analyzing EHR data. The research marks a significant step towards more informed and personalized breast cancer treatment, highlighting the transformative potential of health informatics in managing complex disease interactions and improving patient outcomes.
  • Thumbnail Image
    Item
    Model of international student persistence: factors influencing retention of international undergraduate students at two public statewide four-year university systems
    (2010-01) Kwai, Chee Khei (C.K.)
    The current global economy has created a new middle class around the world, making higher education more accessible to a wider population. The increasing diversity in U.S. higher education is not only the result of minority American students, but also due to the increasing enrollment of international students. This study examined the factors influencing retention from fall 2006 to fall 2007 of international undergraduate students (N = 454) in two public statewide four-year university systems. The model used in the study was based on a combination of retention models by Tinto (1975) and Astin (1970), and revisions made by Tierney (1992) and Pascarella and Terenzini (1980). The data in this study were analyzed using stepwise binomial logistic regression as the primary statistical technique. The findings of this study showed that the results were consistent with other retention studies where there was no single factor or model to predict the persistence of postsecondary students in U.S. higher education institutions. Results for most variables studied were either unclear or inconsistent. Only academic achievement was consistently shown to have a statistically significant and positive effect on persistence into the second year of international students in this study. The difference in the results of this study, in comparison to studies of factors affecting the retention of domestic students, is intriguing. In a way, this study raises more questions than it answers. In conclusion, this study indicated that variables, such as spring semester GPA, credit hours attempted, and on-campus employment have a positive effect on retention into the second year of international undergraduates.
  • Thumbnail Image
    Item
    A simplified risk prediction model using electronic medical record data for pediatric and adult patients with congenital heart disease undergoing cardiac surgery
    (2013-05) Golden, Adele Wen
    Background: Vasoactive-inotrope score (VIS) has recently been proposed as a surrogate marker of illness severity after cardiac surgery for pediatric patients with congenital heart disease (CHD). However, it has not been validated in an exclusively pediatric population as a robust outcome predictor in the early postoperative period. Furthermore, as a result of advances in the treatment of CHD, the majority of these children now survive to adulthood when they will require additional surgical intervention. However, there are no risk prediction tools for these adult patients with CHD; and pediatric and adult non-CHD cardiac risk scores perform poorly in this population. A simple yet robust risk prediction tool is crucial to support clinical decision making and optimize quantity and quality of life for both pediatric and adult CHD patients undergoing cardiac surgery. Objectives: This research aims to 1) externally validate VIS risk predictive performance of early outcome in pediatric CHD patients after cardiac surgery; 2) propose a simplified VIS Index model with robust predictive performance of early postoperative mortality and morbidity by incorporating both the magnitude and duration of inotrope support required for pediatric CHD patients after cardiac surgery; 3) evaluate whether the proposed VIS Index has strong discriminative performance of early mortality and morbidity outcome for adult CHD patients after cardiac surgery. Methods: Automated data capture of the electronic medical record (EMR) system was utilized in conjunction with retrospective clinical chart review. A total number of 244 infant CHD patients and 243 adult CHD patients undergoing cardiac surgery at the Mayo Clinic Rochester, MN were included in the study. Inotrope and vasoactive dose values were collected at 15-minute intervals for the first 96 hours after cardiac Intensive Care Unit (ICU) admission. Demographic and clinical data were collected from both Mayo Clinic institutional Society of Thoracic Surgeons database and clinical chart review. Maximum vasoactive inotrope support (maxVIS) values were calculated and VIS postoperative temporal characteristics were further assessed to evaluate their relationship with early mortality and morbidity. The logistic regression model with generalized estimating equation methodology was applied to address the correlated outcomes from the same patient. The maxVIS model was validated on pediatric CHD patients. A simplified VIS index model incorporating both the magnitude and duration of inotrope support was developed with superior predictive performance of early mortality and morbidity for both pediatric and adult CHD patients following cardiac surgery. The area under the curve (AUC) of the receiver operating characteristic (ROC) curves was used to evaluate the discriminative performance; Hosmer-Lemeshow (H-L) test was used to assess the goodness of fit of the model. Results: The maxVIS model proposed by recent research was externally validated in our institution to exhibit good predictive ability (H-L test, P = 0.791) and discriminate reasonably well between CHD patients with high- and low-risk for early mortality and morbidity (AUC = 0.77, 95% CI: 0.72 to 0.82). The new VIS index risk prediction model shows superior discriminative performance over the existing maxVIS model for pediatric CHD patients undergoing cardiac surgery (AUC = 0.84, 95% CI: 0.78 to 0.88; H-L tests, P = 0.725). A high VIS index is strongly associated with a poor clinical outcome compared to a low VIS index. Patients with a VIS index of 6 have an estimated risk of 98% (95% CI: 85% to 100%) of having a poor outcome after cardiac surgery, compared with a risk of 20% (95% CI: 11% to 34%) for patients with a VIS index of 1. Furthermore, both maxVIS model and VIS index model presents robust predictive performance for adult CHD patients after cardiac surgery with the VIS index model consistently showing superior discriminative performance over the maxVIS model for early postoperative mortality and morbidity (MaxVIS model AUC = 0.82, 95% CI: 0.76 to 0.88; VIS index model AUC = 0.88, 95% CI: 0.82 to 0.93). Adult CHD patients with a VIS index of 6 have an estimated risk of 95% (95% CI: 72 % to 99%) of experiencing a poor clinical outcome during early postoperative period, compared with a risk of 6% (95% CI: 3% to 11%) for patients with a VIS index of 1. Conclusions: The maxVIS model is a strong predictive tool of early mortality and morbidity for CHD patients undergoing cardiac surgery. The VIS index we proposed is a more robust, yet much simpler tool to predict early postoperative mortality and morbidity for both pediatric and adult CHD patients after cardiac surgery. More importantly, this is the first analysis evaluating the correlation between VIS and poor clinical outcomes in adult CHD patients undergoing cardiac surgery. The findings of this research will facilitate earlier detection of high risk patients to direct clinical interventions and preventative measures that will improve outcome for pediatric and adult CHD patients after cardiac surgery.