Measuring adherence to mammography screening recommendations among low-income women

OBJECTIVES: (1) To describe the accuracy of self-report and administrative claims for measuring mammography adherence among low-income women; (2) to determine whether the accuracy of self-report differed between telephone interview and mailed questionnaire; and (3) to explore whether the method of measuring adherence affected associations between mammography adherence and participant sociodemographic characteristics. DESIGN: Retrospective cohort study of women receiving care from a Philadelphia Medicaid Managed Care Organization (MCO). PARTICIPANTS: Three hundred and ninety-nine low-income women eligible for screening mammography of whom 64% were African American, 14% Caucasian, 13% Hispanic, and 8% Asian American. MEASURES: Self-reported use of mammography screening, administrative mammography claims data, and dates of mammograms from radiology facility records. The "gold standard" categorized women as having undergone screening if they had either a billing claim or facility record for a mammogram in the past 12 months. MAIN RESULTS: Two hundred and eighty-three of the 399 women reported having had a mammogram in the past 12 months. The sensitivity of self-report was 0.93, specificity was 0.54, positive predictive value was 0.70, and negative predictive value was 0.86. One hundred and seventy-nine of the 399 women had a claim for a mammogram in the past 12 months. The sensitivity of claims data was 0.83 with a negative predictive value of 0.84. The sensitivity of self-report was higher with telephone data collection (0.98) than with mailed data collection (0.82), while the specificity was higher with mail (0.64) than telephone (0.50). African American race was associated with adherence to screening recommendations when mammography use was measured by self-report (RR 1.31, P = 0.002) but not when it was measured by claims or facility validation (RR 1.03, P = 0.56, and RR 1.12, P = 0.15, respectively). CONCLUSIONS: Accurate measurement of adherence to mammography screening among low-income women is difficult. Self-report substantially overestimates adherence (particularly when collected through telephone interviews), while also misclassifying some women who underwent screening as not having been screened (particularly when collected through mailed questionnaires). In contrast, administrative claims data substantially underestimates adherence. Inaccurate measurement of mammography adherence can lead to a biased understanding of the factors associated with adherence.