Systematic Differences in Validity of Self-Reported Mammography Behavior: A Problem for Intergroup Comparisons?

Background. Prior studies of recall accuracy for screening mammogram behavior have examined relatively homogeneous groups. Data are limited on possible systematic group differences, so we evaluated womenʹs recall accuracy in two separate care systems in one city. Methods. Women 50 to 70 years old with and without screening mammograms 10 to 14 months prior were identified from fiscal, clinic, and radiology records at a military care system (MCS) and a county-funded system (CFS) for indigents. Mammogram status was verified through radiology records. Women were excluded if mammograms were diagnostic, done for other than annual screening, or had abnormal results. Interview ers blinded to mammogram status surveyed randomly selected eligible women. Results. For 62 screened/31 unscreened MCS women and 78 screened/61 unscreened CFS women, specificity was similar, at 65 and 62%, respectively. In contrast, sensitivity varied significantly: 95% versus 79% (P = 0.011). Primary ethonocultural groups were Euro-American (MCS—60%) and Mexican American (CFS—85%). Although not different in specificity of recall (67% versus 61%), these major subgroups significantly differed in sensitivity (97% versus 80%, P = 0.017), proportion of true negatives due to never having a mammogram (35% versus 57%, P = 0.003), and proportion with ≥high school education (78% versus 19%, P < 0.00001). Conclusion. Systematic differences in recall validity may exist and compromise the accuracy of intergroup comparisons.