Validity of women's self-reports of cancer screening test utilization in a managed care population

Abstract

This study was undertaken to examine the validity of self-reported data on breast and cervical cancer screening behavior. An abbreviated version of the Behavioral Risk Factor Surveillance System telephone survey, including questions on mammography, clinical breast examination (CBE), and Papanicolaou test utilization, was administered to a sample of 480 women aged 40-74 years, enrolled in Kaiser Permanente Colorado for at least 5 years. Screening information reported in the telephone interview was compared with that abstracted from respondents' medical records. The vast majority of women had a mammogram, CBE, and Pap test according to both self-report and medical record. Sensitivity for determining whether her last test was within 2 years (3 years for Pap test) exceeded 95% for all, whereas specificities were <55%. The percentage of overall agreement between self-reported and recorded information was 88.4% (kappa = 0.62) for mammography, 87.9% (kappa = 0.45) for CBE, and 87.2% (kappa = 0.54) for Pap test. Pearson correlations between self-reported and recorded information for specific time interval since most recent mammogram, CBE, and Pap test were 0.72, 0.58, and 0.65, respectively. Correlation increased greatly when time interval was allowed to vary by +/-1 year. In most cases of disagreement, the self-report underestimated the time since last screening. These results suggest that self-reporting of breast and cervical cancer screening is fairly accurate in this managed care population, although women tend to underestimate the time since their last screening.