To examine the hypothesis that a demonstration clinic integrating homeless, primary care, and mental health services for homeless veterans with serious mental illness or substance abuse would improve medical health care access and physical health status. A quasi-experimental design comparing a 'usual VA care' group before the demonstration clinic opened (N = 130) and the 'integrated care' group (N = 130). Regression models indicated that the integrated care group was more rapidly enrolled in primary care, received more prevention services and primary care visits, and fewer emergency department visits, and was not different in inpatient utilization or in physical health status over 18 months. The demonstration clinic improved access to primary care services and reduced emergency services but did not improve perceived physical health status over 18 months. Further research is needed to determine generalizability and longer term effects.