Social inequalities in participation in cancer screening: does the mode of data collection matter? The CONSTANCES cohort.

Abstract

Self-reported data are prone to item non-response and misreporting. We investigated to what extent the use of self-reported data for participation in breast (BCS) and cervical cancer screening (CCS) impacted socioeconomic inequalities in cancer screening participation. We used data from a large population-based survey including information on cancer screening from self-reported questionnaire and administrative records (n = 14122 for BCS, n = 27120 CCS). For educational level, occupation class and household income per capita, we assessed the accuracy of self-reporting using sensitivity, specificity and both positive and negative predictive value. In addition, we estimated to what extent the use of self-reported data modified the magnitude of socioeconomic differences in BCS and CCS participation with age-adjusted non-screening rate difference, odds ratios and relative indices of inequality. Although women with a high socioeconomic position were more prone to report a date for BCS and CCS in questionnaires, they were also more prone to over-declare their participation in CCS if they had not undergone a screening test within the recommended time frame. The use of self-reported cancer screening data, when compared with administrative records, did not impact the magnitude of social differences in BCS participation but led to an overestimation of the social differences in CCS participation. This was due to misreporting rather than to item non-response. Women's socioeconomic position is associated with missingness and the accuracy of self-reported BCS and CCS participation. Social inequalities in cancer screening participation based on self-reports are likely to be overestimated for CCS.