Mammography use by elderly women: a methodological comparison of two national data sources

Ann Epidemiol. 1998 Oct;8(7):439-44. doi: 10.1016/s1047-2797(98)00010-6.


Purpose: Estimates of mammography utilization vary considerably, depending on the data source. Among women aged 65 years and older, recent estimates of annual mammography derived from the 1992 National Health Interview Survey (NHIS) were 50% higher than estimates from Medicare claims. We investigated possible reasons for the different estimates.

Methods: We identified differences in the populations covered by the two data sources and made appropriate adjustments. Differences due to age were addressed by age restriction and age adjustment. Women in health maintenance organizations were eliminated from the NHIS sample so it more closely resembled the Medicare database, and estimates of mammography utilization by noninstitutionalized women were derived for Medicare to increase comparability with NHIS. By using the Medicare Current Beneficiary Survey to obtain individual-level comparisons between self-report and claims, we explored potential biases in self-reported data and missing claims.

Results: Differences between the sample populations accounted for more than one-fourth of the rate difference. About half of the difference could be attributed to erroneous self-reports, biases in self-reported dates (forward and reverse telescoping) and missing Medicare claims.

Conclusions: Most of the discrepancy between the two data sources can be plausibly explained. However, caution must be used in using either data source alone, or both together, to represent the "true" mammography rate.

Publication types

  • Comparative Study
  • Research Support, U.S. Gov't, P.H.S.

MeSH terms

  • Aged
  • Aged, 80 and over
  • Bias
  • Data Collection / methods*
  • Female
  • Health Care Surveys / methods*
  • Health Maintenance Organizations / statistics & numerical data
  • Humans
  • Insurance Claim Reporting / statistics & numerical data*
  • Mammography / statistics & numerical data*
  • Medicare / statistics & numerical data*
  • Reproducibility of Results
  • United States