This is the Final Report on the findings and methodology of a study that compared data-driven and a priori literature-based definitions of “elder abuse polyvictimization” (PV), with the goal of determining which definition best fits the data and whether it encompasses various types of abuse and mortality.
The study used 5 years of investigative data from the Texas Department of Family and Protective Services, Adult Protective Services Division, and the Center for Medicare Services. Structural equation modeling was used to examine the latent class definitions of PV, and machine-learning algorithms were used to maximize classification of participants into groups based on victim and perpetrator demographics and adult protective services (APS) investigation data. Logistic regressions were used to model the a priori defined PV types, the latent class analysis of classes, and the individual abuse types with Center for Medicare Services health outcomes, i.e., death, depression, dementia, and anxiety. The sample was majority female, White, and English-speaking. The findings from this study suggest that multiple types of PV are related to poor health, mental health, and mortality, which indicates the relevance of a broad and inclusive definition of PV. This study provides the first in-depth characterization study of elder abuse PV and relations with morbidity and mortality outcomes. It not only provides new insight into the reasoning regarding why PV should be defined broadly and inclusive of more than just co-occurring abuse; it also provides new evidence that links a priori PV categories to latent profiles that can be used to guide investigations and morbidity and mortality outcomes that should be considered when designing protective services plans for the current and future remediation and prevention of elder abuse. 3 tables, 2 figures, and 67 references