Women's retrospective reports of childhood sexual abuse in cross-sectional household surveys: a multi-country secondary analysis of two data collection methods.

Abstract

Under-reporting and non-disclosure hinder the accurate collection of self-reported experiences of childhood sexual abuse. Although population-based household surveys often use direct (ie, face-to-face) interviews to assess childhood sexual abuse exposure retrospectively, some use confidential methods to encourage disclosure. Both methods show promise, but their comparative effectiveness remains underexplored. This study aimed to synthesise and compare childhood sexual abuse prevalence estimates derived from direct and confidential data collection methods across countries. For this secondary analysis, we used publicly available, cross-sectional report data from the WHO multi-country study on women's health and domestic violence against women and its subsequent adaptations (ie, national surveys on violence against women). Adaptation studies were included if their methods or questionnaires were adapted from the WHO multi-country study; they reported childhood sexual abuse case counts or prevalence estimates by the mode of data collection (ie, direct interview or confidential card); and they used a nationally representative sampling framework. After extracting prevalence estimates and sample sizes by collection method, we applied the Wilson score interval to calculate 95% uncertainty intervals (which were used as a proxy for statistical significance in the absence of individual-level data). Within each location, we then tested for significant differences in prevalence estimates derived from direct interviews, confidential self-report cards, and a combined approach. Using the direct interview data as the reference, we calculated the absolute and relative differences in prevalence between the data collection methods. We used representative data from 36 surveys and 30 countries and territories, all collected between 2000 and 2019. Overall, the prevalence of childhood sexual abuse measured by confidential cards was 5·7 percentage points greater than the prevalence measured by direct interviews. In 28 (85%) of the 33 studies that reported confidential card responses, the prevalence via confidential report was significantly higher than the direct interview. At the country level, the smallest observed absolute differences were in Viet Nam (1·0 percentage points in 2010) and Türkiye (1·9 percentage points in 2008), whereas the largest were in Jamaica (11·1 percentage points) and the Solomon Islands (22·1 percentage points). In relative terms, the pooled confidential estimate was 121·2% higher than the direct question, reflecting more than a two-fold increase in disclosure. Surveys should provide respondents with an opportunity to confidentially disclose childhood sexual abuse, which could increase disclosure, lead to more accurate prevalence estimates, and, in turn, inform expanded primary and secondary prevention efforts. Gates Foundation.