We applied the Gelberg-Andersen Behavioral Model for Vulnerable Populations to predict health services utilization (HSU) in 875 homeless US women. Structural models assessed the impact of predisposing (demographics, psychological distress, alcohol/drug problems, homelessness severity), enabling (health insurance, source of care, barriers) and need (illness) variables on HSU (preventive care, outpatient visits, hospitalizations). Homelessness severity predicted illness, barriers and less insurance. Distress predicted more barriers, illness and less outpatient HSU. Drug problems predicted hospitalizations. Barriers predicted more illness and less outpatient HSU. Health and homelessness indicators were worse for White women. Better housing, access to care and insurance would encourage appropriate HSU.