Measurement innovation for sensitive behaviours: applying direct and social network-based estimation approaches to intimate partner violence in Burkina Faso and the Democratic Republic of the Congo using cross-sectional data

Objectives To explore the feasibility of the confidante methodology to measure past-year intimate partner violence (IPV) experiences in Burkina Faso and the Democratic Republic of the Congo (DRC) through (1) comparison of direct assessment with indirect estimation via the confidante method and (2) assessment of the performance of each confidante method assumption. Design Cross-sectional study with nationally and subnationally representative data collected from December 2020 to March 2021 in Burkina Faso (national) and from December 2021 to April 2022 in Kinshasa and Kongo Central, DRC (subnational). Setting Burkina Faso; Kinshasa, DRC; Kongo Central, DRC. Participants Partnered women (married or cohabiting) aged 15–49 in Burkina Faso (N=3047), Kinshasa, DRC (N=702) and Kongo Central, DRC (N=688) and their partnered confidantes aged 15–49 (N=2064 in Burkina Faso, N=304 in Kinshasa, DRC, N=393 women in Kongo Central, DRC). Primary and secondary outcome measures Past-year IPV (emotional, physical, sexual, any) comparing differences in prevalence between the direct respondent sample and the indirect confidante sample, adjusting for confidante method assumptions. Results The confidante method produced comparable IPV estimates to respondents’ direct reports across sites (35.3% respondent vs 36.1% confidante in Kinshasa, DRC; 29.7% respondent vs 39.0% confidante in Kongo Central, DRC; 25.7% respondent vs 26.0% confidante in Burkina Faso, differences not statistically significant). Of note, there were differences in IPV estimates between respondents and confidantes by IPV subtype, with physical IPV consistently lower among respondents across sites and sexual IPV lower among confidantes in Kinshasa, DRC and Burkina Faso, though generally not statistically significant. Conclusions The confidante methodology did not afford advantages over standard, direct assessment for IPV. Overall, findings indicate the reliability of population-based surveys with direct IPV questions when implemented under recommended ethical guidelines, though direct reports are likely undercounts.