Heterogeneity in the relationship between the time tradeoff and Short Form-36 for HIV-infected and primary care patients

OBJECTIVES. Evidence in the literature suggests that the overall correlation between descriptive and valuational measures of health are weak to moderate. In this study, the relationship between descriptive health status measures, obtained using the Short-Form 36, and health values, measured with the time tradeoff, was explored. METHODS. Two groups of patients matched for age and gender were interviewed. One group comprised 139 human immunodeficiency virus (HIV)-infected patients; the other group comprised 124 primary care patients. The relationship between the SF-36 and the time tradeoff was estimated, assuming homogeneity across patients, using multiple regression analysis. Subsequently, the relationship was examined assuming heterogeneity across patients and using the expectation maximization algorithm in a maximum likelihood context (latent class analysis). RESULTS. Four classes, representing 47%, 13%, 8%, and 32% of the population, respectively, were found. The overall percentage of variation explained under the assumption of a homogeneous relationship was only 33% as compared with 85% when heterogeneity was accounted for. Only three characteristics (educational level, employment status, and the SF-36 social functioning score) sufficed to generate a nearly perfect classification of the patients. CONCLUSIONS. Heterogeneity across subjects should be taken into account in describing the relationship between health values and health status dimensions. Key words: health status measure; health utility measure; latent class analysis; heterogeneity. (Med Care 1998;36:523-532)