A Perceived Health Index for use in persons with advanced HIV disease: derivation, reliability, and validity

Acceptance of health status as an outcome in clinical trials of new treatments for HIV disease has been hampered by the lack of valid and clinically relevant means of summarizing differences across multiple dimensions. We formed a summary Perceived Health Index from health status measures adapted from RAND Medical Outcomes Study scale and contained in the HIV-PARSE survey instrument, which had been administered to a large number of participants in clinical trials for advanced HIV disease. The psychometric properties of the included health status scales were assessed using multitrait scaling and test-retest stability. Weights for the index were derived from regressions of Current Health Perceptions on the domain-specific health status scales. The effect of participant characteristics on weights was tested. Finally, the reliability and known-clinical groups validity of the index was assessed. Data were obtained from 1,862 clinical trial participants who provided a total of 7,352 observations. The mean CD4 count was 131. The internal consistency reliability of the multi-item scales ranged from 0.86 to 0.90, and items demonstrated excellent discrimination across scales. The domain-specific scales explained 59% of the variation in the Current Health Perceptions scale (P < 0.00001). The resulting Perceived Health Index was equal to 0.20 Physical Functioning + 0.15 Pain + 0.41 Energy/Fatigue + 0.10 Emotional Well-being + 0.05 Social Functioning + 0.09 Role Functioning. The strong positive bivariate relationship between the Cognitive Function/Distress scale and the Current Health Perceptions scale was subsumed by the combination of the other domain-specific scales in multiple regressions, so it does not appear independently in the index. The proportional weights used in the index were insensitive to variations in demographics. The reliability of the index was conservatively estimated to be 0.94. Patients with index scores in the lowest quartile had a 2- to 11-fold higher probability than those in the highest quartile of reporting various specific clinical events, and the index correlated significantly more highly with the number of such events than did the current health perceptions scale. The modified MOS health status scales included in the HIV-PARSE are reliable and valid in patients with advanced HIV disease. The Perceived Health Index provides a reliable and valid means of summarizing self-reported current health, correlates strongly with clinical indicators, and should be useful as a outcome measure in patients enrolling into clinical trials of therapies for advanced HIV disease. Regression based weights are a useful means of summarizing multidimensional measures.