Community-Based Risk Assessment of Elder Mistreatment and Self-Neglect: Evidence of Construct Validity and Measurement Invariance Across Gender and Ethnicity

Mistreatment and self-neglect among older adults are linked to significantly high mortality rates, but research in this area is hindered by a lack of standardized assessments to identify these circumstances. No available assessment has been tested for construct validity (CV) or measurement invariance (MI) across gender or race/ethnicities. This study presents the CV and MI properties of a comprehensive assessment of elder mistreatment (EM) and self-neglect (SN). The 57-item Texas Adult Protective Services Client Assessment and Risk Evaluation (CARE) tool was completed on 7,580 substantiated cases of EM and SN. Confirmatory factor analysis (CFA) and cross-validation techniques were used to assess CV, and MI techniques were used to assess item and domain bias across genders and race/ethnicities. Model fit was estimated using robust estimators. A 43-item CARE tool with 5 domains reached good fit and cross-validated (X-2 DIFF TEST, p = 0.69). Overall, the construct reliability ranged from 0.78 (physical/medical) to 0.93 (living conditions). The same model was measurement invariant across gender (X-2 DIFFTEST, p = 0.02). The financial status domain could not be used for the Hispanic population, but a 4-factor CARE tool model was measurement invariant across African Americans, Caucasians, and Hispanics (X-2 DIFFTEST, p = 0.05). The constructs of the CARE tool appear to be valid and reliable assessments of EM and SN. Overall, the use of this assessment to identify EM and SN does not result in health or social service disparities among older adults. Further research on this topic is needed.