Abstract:
|
The Medical Expenditure Panel Survey (MEPS) is a nationally representative survey conducted annually by the Agency for Health Research and Quality (AHRQ). Respondents to the Household Component (HC) of MEPS provide detailed information on health care use and expenditures, as well as health insurance coverage, access to care, demographic and socioeconomic characteristics. For a subset of respondents, medical providers associated with medical events reported by the household are contacted to obtain more precise expenditure information. While the primary motivation for conducting this follow-back survey, called the Medical Provider Component (MPC), is to collect data to improve the quality and completeness of expenditure data, we leverage data in the MPC to determine the extent to which HC respondents may be mis-reporting the number of medical events. We treat the MPC as a validation data set and use machine learning methods to build a model to identify correlates of reporting accuracy. We use this model predict reporting accuracy for those respondents for whom provider data were not collected.
|