Ethnic Differentials in Responses to Self-Reported Health: An Application of Item Response Theory (IRT)

Rania Tfaily, University of Pennsylvania
Beth J. Soldo, University of Pennsylvania

Despite its utility as an overall summary indicator of health, self-reported health (SRH) has limited value in comparative studies of health inequalities. The SRH variable suffers from problems in measurement comparability, due to linguistic differences, cultural differences, variations in interpretation of the question, and the response set. Item response theory (IRT) offers a method to test the meaningfulness of the self-reported health measure and to compare its categorical cut-points across different groups. In this paper, we apply a Graded Response Model (GRM), a generalization of the two-parameter logistic IRT model, to study the measurement of self-reported health using data from the 1998 Health and Retirement Study (HRS) and the 2001 Mexican Health and Aging Study (MHAS). Category response curves for each response level of self-reported health are drawn for: White-Americans, African-Americans, Mexican-Americans, and Mexicans with and without controls for age, gender and education.