Injuries/Violence
Understanding Discordance of Self-Reported Child Abuse in the Health & Retirement Study Sophie Selbe* Sophie Selbe Rachel Slimovitch Jeanine I. Nasser Michelle Flesaker Sarah E. Weber Virginia Cafferky Casey DeMarsico Anthony J. Rosellini Jaimie L. Gradus
Prior studies have assessed the validity of retrospective self-reporting of child abuse in research by examining individual characteristics and life events that may explain recall bias. Yet, none have evaluated potential predictors of reporting discordance in older adults. Data were from the Health and Retirement Study (HRS), a nationally representative longitudinal panel study that began in 1992 comprised of adults ages 50+ in the U.S. (n = 9,101). Child abuse was asked at two timepoints, 4 years apart, from 2006-2012, and was defined as any physical abuse by parents before age 18. Reporting discordance was defined as a changed response across timepoints. We evaluated possible predictors of reporting discordance at baseline including demographic, socioeconomic, physical and mental health, and other trauma-related factors. We further examined changes from 2006-2012 in socioeconomic, mental and physical health factors. We calculated crude RRs and 95% CIs to examine associations between potential predictors and child abuse reporting discordance. Two sets of analyses were conducted: one that examined participants who changed from yes to no (compared to concordant yes responses) and a second that examined participants who changed from no to yes (compared to concordant no responses). Among those who changed responses from no to yes, associations were found for depression (RR=2.3, 95% CI: 1.7, 3.0), improved depression between timepoints (RR=2.1, 95% CI: 1.4, 3.0), new psychological diagnosis (RR=2.5, 95% CI: 1.4, 4.3), family substance use problems (RR=2.6, 95% CI: 2.0, 3.5), and being a victim of a physical attack (RR=2.9, 95% CI: 2.0, 4.3). Among those who changed responses from yes to no, associations were found for Black/African American race (RR=1.6, 95% CI: 1.2, 2.1), Hispanic ethnicity (RR=1.4, 95% CI: 1.0, 2.0), and worse cognition between timepoints (RR=1.3, 95% CI: 1.0, 1.8). Results from machine learning random forests prediction models will also be presented.