Measurement properties of quality-adjusted life year (QALY) measures among older adults with chronic neck pain

Abstract

Background Quality-adjusted life year (QALY) measures are an important outcome for assessing the cost-effectiveness of healthcare interventions. Ideally, the choice of QALY measures will be informed by the measurement properties within the population of interest. Currently, the EQ-5D and SF-6D are the most commonly used QALY measures within cost-effectiveness analyses for spine pain. A number of studies have assessed the measurement properties of QALY measures for individuals with spine pain, but primarily within surgical populations. The psychometric properties of QALY measures may vary substantially within non-surgical populations. The primary aim of this thesis is to assess the psychometric properties (reliability, validity, and responsiveness) of commonly used QALY measures (SF-6D, EQ-5D, EQ Visual Analog Scale) among older U.S. adults with chronic mechanical neck pain managed non-surgically. The secondary aim of the thesis is to assess differences in the psychometric properties of QALY measures derived from the same instrument (SF-6D), but using different valuation methods. Methods Data for the study was collected within a randomized clinical trial comparing different combinations of non-invasive interventions (home exercise and advice, supervised exercise therapy, spinal manipulation) for the management of chronic neck pain in older adults. Quality-adjusted life years (QALYs) were measured with the 1) SF-6D, 2) EQ-5D, and 3) Euroqol visual analogue scale (EQ VAS) using U.S. population values for the primary aim. Test-retest reliability was determined using intraclass correlation coefficients (ICCs). The Bland-Altman method for limits of agreement and the smallest detectable change (SDC) were used to assess agreement. The longitudinal known-group validity and responsiveness of QALY measures was estimated using four external criteria: 1) global perceived change in health; 2) global improvement in neck symptoms; 3) neck pain; and 4) neck disability. Known-group validity was assessed by calculating mean QALY changes for each category of global perceived change in health and neck symptoms in addition to quintiles of neck pain and disability improvement. The relative responsiveness of QALY changes was estimated using correlation and area under the receiver operating characteristic (ROC) curve analyses. Results The SF-6D demonstrated better test-retest reliability (ICC = 0.81; 95% CI 0.77 to 0.85) relative to the EQ-5D (ICC = 0.44; 95% CI 0.33 to 0.53) and EQ VAS (ICC = 0.68; 95% CI 0.61 to 0.75). In addition, the smallest detectable change was lowest for the SF-6D (0.16; 95% CI 0.14 to 0.17), followed by the EQ-5D (0.18; 95% CI 0.16 to 0.20), and EQ VAS (0.22; 95% CI 0.20 to 0.25). Differences in QALYs during the one-week baseline period were evenly spread over the range of mean QALYs for the SF-6D, but not the EQ-5D or EQ VAS. The SF-6D and EQ VAS demonstrated better longitudinal known-group validity relative to the EQ-5D. Mean SF-6D and EQ VAS QALY changes were monotonically decreasing across levels of improvement for three of the four external criteria. All three QALY measures demonstrated similar responsiveness to change. Correlations between QALY measures and three of the external criteria were similar and very low to low in strength (-0.233 to -0.391). Correlations with neck disability were low to moderate in strength with the SF-6D demonstrating the strongest association (-0.596; p-values for differences with EQ-5D and EQ VAS = 0.01). There were no significant differences among the QALY measures when measuring responsiveness with area under the ROC curve. SF-6D based QALY measures had similar reliability, agreement, validity, and responsiveness. Conclusions There were minor differences between U.S. QALY measures in terms of responsiveness; however, the SF-6D was more reliable and demonstrated less measurement error relative to the EQ-5D and EQ VAS, in addition to better known-group validity relative to the EQ-5D. The different methods for obtaining QALY values from the same instrument (SF-6D) had little to no impact on the psychometric properties.