||Composite Estimation Model for Producing Stabilized Health Rate Estimates for Small Areas Using Sample Surveys: Experience from Health Surveys in Ethiopia, India and Indonesia.
Manton, K. G. ;
Woodbury, M. A. ;
Stallard, E. ;
Dowd, J. E. ;
||Duke Univ., Durham, NC. ;World Health Organization, Geneva (Switzerland).;Health Effects Research Lab., Research Triangle Park, NC.
Bayes theorem ;
Probability density functions ;
Health surveys ;
Health planning ;
Delivery of health care ;
||Most EPA libraries have a fiche copy filed under the call number shown. Check with individual libraries about paper copy.
Data from population health surveys can be very important in planning service delivery for spatially defined subpopulations because they reflect the medical needs of the total subarea population--not just those utilizing services under the current service delivery system in each area. Unfortunately when modeling relatively rare health events at the subarea level, sample survey based rate estimates are often highly variable due to small numbers. To improve the stability of the local area rate estimates, an empirical Bayes strategy which produces posterior rate estimates that are a weighted composite (a) the rate estimates from the total survey population and (b) the rate estimates for each subarea is proposed. The weights, derived from a negative binomial regression analysis, reflect the distributional characteristics of the total assemble of local area rate estimates. The methodology is illustrated on health survey data from Ethiopia, India, and Indonesia, and some vital statistics data from the United States.