Abstract
We examined the effects of length of recall period and written records on the accuracy of household reports of health care use of Medicare beneficiaries in the Medical Expenditure Panel Survey (MEPS), a widely used nationally-representative federal survey. Our sample contained 1,375 Medicare beneficiaries with 5 complete rounds of MEPS interviews during 2001–2003 who were matched to their Medicare claims records. Household respondents systematically underreported any emergency department use per round (mean of 0.07 vs. 0.09 in the claims, p< 0.001) and number of ambulatory visits per 90 days (2.3 vs. 2.7, p < 0.001). In logistic regressions, recall periods less than 2 months and, to a lesser extent, 3 to 4 months were associated with lower odds of underreporting visits and higher overall accuracy. Calendars and bills, insurance statements, and other records of health care events kept by some MEPS respondents also were associated with lower odds of underreporting.