Predictors of VA and Non-VA Health Care Service Use by Homeless Veterans Residing in a Low-Demand Emergency Shelter

Separate logistic regression models predicted use of psychiatric and substance-abuse services with 3 models (VA, non-VA, or any service use) for each dependent variable from independent variables that reflected need and enabling factors (Tables 2 and 3). Demographic predisposing factors, which were not associated with service use, were not included as covariates in these models. For the model predicting the use of non-VA substance-abuse services, collinearity between the alcohol-abuse and cocaine-abuse variables required separate models for each of the 2 variables.

Medical sector service use predicted psychiatric sector service use in all models. In fact, VA medical service use was the only predictor of use of VA psychiatric services. Lower mental health functioning predicted the use of any (VA or non-VA) psychiatric service use. In addition to the use of medical services, 30-day cocaine use predicted non-VA psychiatric service use.

Any substance-related sector service use was predicted by lower mental health functioning, self-reported alcohol problem, and any medical services utilization. No independent variables included in the model predicted any VA substance-related service use. Non-VA substance abuse service use was predicted by non-VA psychiatric service use and alcohol abuse. In the separate analysis that replaced alcohol problems with 30-day cocaine use variable, only 30-day cocaine use predicted non-VA substance-related service utilization.

Discussion

This study examined the use of medical, psychiatric, and substance-abuse services by randomly sampled veterans from a low-demand emergency shelter. Random selection of the sample and its high (98%) participation rate virtually eliminated potential for bias within this sample. Another strength of this study is its focus on low-demand shelter users—a population that has not been well studied. This low-demand shelter-dwelling population of veterans experiencing homelessness is of interest because more substance-abuse problems and histories of incarceration seem to make them especially disadvantaged and challenged.

The limitation of the sample to users of a low-demand shelter at only 1 location may reduce generalizability to other veteran homeless populations and settings. The study also may not generalize to populations of female veterans experiencing homelessness. Another limitation of the study is that it did not use diagnostic assessments for psychiatric and substance use disorders and objective collateral information such as agency record data. Finally, although the limited size of the sample may have been insufficient to adequately test certain hypotheses, it was a relatively large sample of this population and was large enough to yield significant findings.

This study found that need-based factors predicted the use of some service sectors intended for those needs. For example, mental health functioning appropriately predicted any psychiatric service use, and presence of an alcohol problem appropriately predicted any substance abuse service use. Specifically for non-VA services, both cocaine use and presence of an alcohol problem in separate models predicted substance-abuse service use. However for VA substance-abuse services, neither cocaine use nor presence of an alcohol problem predicted service use. Despite the high need, very few veterans used substance-abuse services, and they rarely used VA substance-abuse services.

For 2 service sectors, need-based factors predicted the use of services intended for other needs. Cocaine use predicted non-VA psychiatric service use, and low mental health functioning predicted substance-abuse service use. One potential explanation for this finding could be that providers or patients incorrectly classified cocaine-related substance use problems as psychiatric. The VR-12 mental health functioning measure also may have incorrectly classified cocaine-related problems as psychiatric.

Three enabling factors predicted service use by sector and type. The first 2 are preference for VA-provided services and the geographic availability of services, which competed for veterans’ selection of service providers. When both VA and non-VA services were present in Fort Worth, a preference for VA-provided services was observed, with the exception of outpatient substance abuse services which were highly underutilized in general. No preference was observed for any non-VA services when both were present. When VA services were not present in Fort Worth, veterans used geographically available non-VA providers for some services, but for other services they used Dallas-based VA and Fort Worth-based non-VA providers equally (Table 3 and Table 4).

The third enabling factor influencing service use was through other service use as an enabling pathway. Those veterans who opted out of locally available services in favor of VA services in Dallas may have been prompted to do so by provider referrals, which were further facilitated by VA and public transportation between Fort Worth and Dallas. The most consistent enabling pathway was medical service use, which predicted all types of psychiatric service use (VA and non-VA combined, VA only, and non-VA only), and any substance-related service use. Psychiatric service use predicted substance abuse service use but only in non-VA settings; no pathways led from VA medical or psychiatric services to VA substance abuse services.