Practical issues in conducting small-area variation analysis

This paper provides a detailed examination of the conduct of a small-area variation analysis (SAVA) health services research project to aid other researchers in performing and interpreting such projects.

SAVA was performed on hospital admissions with suspected acute cardiac ischemia in the state of Michigan, and the details of each of the steps are presented. The study team consisted of the principal investigator, a biostatistician, a health economist, a medical geographer, and a systems analyst. Data were obtained from the state hospital association and from a commercial demographics source. Small areas for analysis were defined by a patient origin clustering method. Crude area rates were adjusted for age and gender composition of small areas. Multivariate Poisson regression that included sociodemographic variables was performed on adjusted rates. Issues in interpretation of SAVA and practical barriers to primary care SAVA are discussed.