Injuries/Violence
Addressing measurement error in IPV self-report data using multiple overimputation and multidimensional quantitative bias analysis Irina Bergenfeld* Irina Bergenfeld Cari Jo Clark Robin A. Richardson Regine Haardörfer Alexandria R. Hadd Charis Wiltshire Timothy L. Lash Angela M. Bengtson
Intimate partner violence (IPV) is an important global health issue for which measurement error, including underreporting, limits public health action. Although most national IPV prevalence estimates come from general health surveys like the Demographic and Health Surveys (DHS), such data are presumed to underestimate prevalence compared to violence-focused surveys (VFS). Using VFS conducted in the same country and year (±1) as validation data, we explored two methods of bias adjustment to address measurement error in DHS prevalence estimates. In multidimensional bias analysis (MBA), we directly adjusted summary prevalence estimates, using a range of possible sensitivities (10%-100%) and specificities (95%-100%) to elucidate their reasonable bounds. In multiple overimputation (MO), we probabilistically overimputed all IPV observations, incorporating prior information on measurement error, and combined prevalence estimates over 50 MO iterations. MBA revealed that an assumption of 95% specificity resulted in negative prevalence estimates in some cases, confirming that false positives are likely negligible (Figure). Reasonable sensitivities varied considerably across countries and IPV types, likely due to differences in the number of items used to assess IPV. MO-adjusted estimates were similar to VFS estimates, except when unadjusted DHS estimates were <5% and highly discrepant. In surveys assessing physical IPV using several items, DHS and VFS estimates were comparable, while DHS estimates for sexual and emotional IPV were lower. Past-year estimates were less discrepant than lifetime estimates, suggesting that recall bias may be a factor in underreporting. This study examines the nature and scope of measurement error due to IPV underreporting in specific contexts where external information exists, highlighting the need for more accurate IPV assessment using multiple items per domain and for internal validation studies to be incorporated into large-scale surveys.
