Abstract
Two studies were conducted to determine the reliability of measures of homelessness (Study 1), and to use these measures to predict homelessness from data collected over a 1O-year period (Study 2). For Study 1, to determine test-retest reliability, the Risk Behavior Assessment (REA; National Institute on Drug Abuse, 1991) was administered twice within a 48-hour period. Cohen’s kappa was calculated at .86 for the two REA measures of homelessness. For Study 2, to predict homelessness from data collected over a 10-year period, a logistic regression model was built. Earlier homelessness and illegal income were identified as risk factors; number of times tested for HIV and Black ethnicity were identified as protective against homelessness. Predictive accuracy was assessed with a reduced bias method approximating jackknifing techniques. These findings point to the need to provide enhanced services-including HIV testing opportunities andjob training- to the homeless.
Notes
* Ad hoc Editors for the review of this manuscript: Gilles Houle & Alison Munson