Do Official Hospitalization Records Predict Self-Reported Medical Vulnerablity?: A Test of the Post-Dictive Validity of the Vulnerability Index
Methods: This study relied on secondary data from individuals receiving homeless services in a southwestern city (N = 97, male = 53.3%, African American = 57.5%). Rates of hospitalizations were computed from past-year official hospitalization records and ranged from 0 – 0.02 (M = 0.001, s.d. = 0.003). The prevalence of substance use (78%) and mental health (72%) were based on individuals’ self-reports on the VI. Physical healthwas the sum of individuals’ self-reports on ten items inquiring about the prevalence of chronic diseases (e.g., HIV+/AIDS). Responses ranged from 0 – 6 (mode = 1). Bivariate correlations were tested initially. Then substance use, mental health, and physical health were regressed on official rates of past-year hospitalizations using three regression models (binary logistic and linear regression, depending on the outcome), controlling for gender and race.
Results: Self-reported substance use and mental health were positively correlated (r = .23, p < 0.01). At both the bivariate and multivariate levels, none of the trimorbid conditions predicted official hospitalization rates. The model for mental health was significant (x2 = 17.27, df = 3, p < 0.01), and females and Whites were more likely than males and African Americans to report mental health problems. The physical health model was also significant (x2 = 9.02, df = 3, p= 0.02), and Whites were more likely to report physical health problems.
Implications: Official hospitalizations did not predict self-reported trimorbid conditions when measured separately – a finding inconsistent with prior research showing a relationship among homelessness, trimorbidity, and hospital use. Moreover, the prevalence of self-reported substance use and mental health conditions was high in this sample suggesting that they individuals were at risk for hospital use. These findings may mean that either: (1) many homeless individuals, with self-reported health conditions, are not using health services, or (2) self-reports are not valid predictors of medical vulnerability. Currently, practitioners may want to collect additional information, e.g., social supports and risk behaviors and non-self-report sources, rather than relying solely on the VI to determine medical vulnerability. Considering the increasing popularity of this instrument, future research assessing the psychometric properties of the VI, e.g., discriminant validity, reliability, measurement invariance, etc., is critical.