Using Data from the National Health Interview Survey (NHIS) To Assess the Effectiveness of Nonresponse Adjustment in the Medical Expenditure Panel Survey (MEPS)

• Activity Number: 96
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2005 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Lap-Ming Wun*+ and Trena M. Ezzati-Rice and Ralph DiGaetano and Huseyin Goksel and Hongsheng Hao
• Companies: Agency for Healthcare Research and Quality and Agency for Healthcare Research and Quality and Westat and Westat and Westat
• Keywords: bias ; nonresponse ; logistic regression ; interaction ; weight adjustment ; propensity score

Abstract

The Medical Expenditure Panel Survey (MEPS) is a large, complex survey sponsored by the Agency for Healthcare Research and Quality (AHRQ). The MEPS, like most sample surveys, experiences unit nonresponse. Adjusting weights to compensate for unit nonresponse is a common practice in large surveys. A major goal of the adjustment is to reduce bias in survey estimates. Generally, the "true" value of variables of interest is not known, so it is not possible to assess the effectiveness of the nonresponse adjustment in reducing potential bias. Each year, a new sample for the MEPS is drawn from respondents to the previous year's National Health Interview Survey (NHIS). A sample weight from NHIS is the initial base weight for the MEPS sample. Adjustment for MEPS dwelling unit nonresponse is made to this initial weight. Through the linkage of the NHIS and MEPS, we can determine the "true" value for variables of interest for all NHIS households sampled for the MEPS (i.e., prior to incurring nonresponse). Then, after adjustments for MEPS nonresponse, we can examine which adjustments came closest to the "true" values. In this paper, we investigate alternative adjustment methods.