partners with community studies in sub-Saharan Africa that are part of the INDEPTH network (the International Network of field sites with continuous Demographic Evaluation of Populations and Their Health in developing countries); they provide the opportunity to replicate this study inexpensively and reasonably quickly in African contexts and in a comparative framework.
In this paper we consider five measures of health—three self-reported and two objectively observed—and address three questions:
How interrelated are these indicators? Can the same information on individual health be obtained with a smaller set of questions or observations, thereby reducing costs?
How well does each serve as a predictor of mortality? Here we assume that the poorer the health of an individual, the greater the risk of dying in a defined period (in this case, 5 years) after health is measured. The greater the predictive power, under this assumption, the better the measure is as a gauge of individual health.
Does the predictive power of the indices vary by age or gender? Here we ask whether, for example, self-reported poor health is related to subsequent mortality similarly across age and gender.
The study is based on health measures collected in the 1996 Matlab Health and Socioeconomic Survey (MHSS). The MHSS was carried out with funding from the U.S. National Institute on Aging as a collaboration among researchers based at the International Centre for Health and Population Research in Dhaka, Bangladesh (ICDDR,B), and at several institutions in the united States. ICDDR,B began its Health and Demographic Surveillance System (HDSS) in Matlab in the early 1960s and, for over 40 years, has visited every household at least monthly to collect accurate information on vital events. The HDSS served as the sampling frame for the MHSS (Rahman et al., 1999). Information on MHSS respondents was therefore automatically collected in the HDSS each month of the 5 years subsequent to the survey. We first consider, using MHSS data alone, the extent to which each of the health measures provides information independent of the others. Next, the combined MHSS/HDSS data are used to test how well each health indicator serves as a predictor of mortality in the follow-up period. We then rank these measures in terms of mortality predictive power and discuss the prevalence of the poor health by age and gender they reveal. Finally, we discuss two issues regarding the usefulness of these simple and inexpensive to collect measures: international comparability and comparability of the indices to more specific information on health obtainable through other means, such as biomarkers. As part of this discussion, we propose a research agenda for measuring health and evaluating health mea-