Integrated survey designs: A framework for nonresponse bias reduction

Abstract

The quality and data content of household specific health surveys are often enhanced through integrated designs which include the conduct of follow back surveys to medical providers and facilities that have provided care to household respondents. In terms of data quality, household reported medical conditions can be evaluated for accuracy relative to provider specific records on medical conditions for the same patient and specific health events. With respect to health care expenditures collected from household respondents for their reported health care events, available linked medical provider level data is a more accurate source of information. The availability of such supplemental data on use and expenditures allows for the conduct of methodological studies to evaluate the accuracy of household reported data and informs adjustment strategies to household data in the absence of provider specific data to reduce bias attributable to response error. In this paper, the capacity of integrated survey designs to achieve reductions in bias attributable to survey nonresponse is discussed. Examples are drawn from the Medical Expenditure Panel Survey(MEPS), an ongoing longitudinal panel survey designed to produce estimates of health care utilization, expenditures, sources of payment, and insurance coverage of the US civilian non-institutionalized population.