The Behavioral Model for Vulnerable Populations supported our main hypothesis that predisposing variables such as homelessness and recent drug use would affect the clinical outcome of detectable viral loads in a sample of HIV sero-positive men recruited into a larger study in Los Angeles. Persons who were homeless or reported recent drug use were more than two times more likely to have detectable viral loads compared to those persons not homeless or reporting recent drug use. However, when introducing enabling variables representing factors that facilitate HIV health care utilization (insurance, income and seeing an HIV specialist), the negative association of homelessness and VL was attenuated. Further, when introducing perceived and observed need for health care utilization variables such as taking medications for HIV; the association of homelessness with detectable viral load is further weakened. In our model, although homelessness itself can be predictive of having a detectable viral load, the provision of health care evidenced by having the ability to pay for care, to have an HIV specialist and to take HIV medications can diminish the association of homelessness with the HIV clinical outcome of viral load. This finding is similar to reports from San Francisco showing a high penetration of antiretroviral medical care among the homeless (Riley et al., 2005
) and is also consistent with findings that persons who reported access to HIV specialists and adherence to antiretroviral medications were more likely to have better biological outcomes (Riley et al., 2005
In contrast, participants who reported recent drug use were consistently less likely to have undetectable viral loads even when health care utilization enabling or need variables were introduced into the model. Paradoxically, the negative association of recent drug use on viral load appears to strengthen as enabling and need variables are brought in.
These analyses have several limitations. First, our sample was recruited using respondent-driven sampling (RDS). While RDS has been proposed as a bias-free sampling methodology to access “hidden” populations, such as MSM and drug users (Ramirez-Valles et al.
, 2005), its ability to produce representative samples is unclear (Martin et al., 2003
). Second, potential self-report bias may have led to an under-reporting of behaviors that are stigmatized or unpopular such as admitting recent drug use or non-adherence to HIV medication regimens. Third, the single biological outcome variable of viral load, while correlated with control of viral replication, is only one aspect of biological control of HIV disease overall. Use of CD4 counts, which is reflective of overall immune status, would undoubtedly broaden the assessment of biological status for these participants. Fourth, although individuals in the sample who reported taking HIV medications were approximately 92% less likely to have detectable viral loads, we did not have a standardized measure of adherence. There are subpopulations of mutated viruses in which the virus is slow to replicate (Pastori et al., 2006
), therefore, persons infected with the delta CCR5 mutation would have lower viral loads, with or without taking medications. However, it is highly improbable that all 107 individuals would have this mutation and is more likely an indirect measure of adherence to medications. Additional analyses of the entire sample from the cooperative agreement (SATH-CAP) need to occur to answer these questions.
Despite these limitations, the Behavioral Model for Vulnerable Populations provided a useful framework in analyzing the association of health care utilization factors on viral load among HIV positive men recruited by RDS sampling. Predisposing factors such as homelessness and reported recent drug use negatively influenced viral load. Enabling factors and need factors such as insurance, visits to an HIV specialist and taking medications had a positive association on viral load for persons who were homeless and HIV positive, but not on those persons who were HIV positive and reported recent substance abuse.
These findings are crucial to understanding factors that correlate with one of the primary biomarkers of HIV disease status in a population with significant healthcare disparities. The use of the Andersen/Gelberg health care utilization model clarified associations between key variables to isolate the relative contribution of homelessness, recent drug use and HIV medication taking on a key marker of HIV medical care. . Findings from this project illustrate the need for policy directed toward emphasizing drug use reduction strategies in groups of HIV-positive, individuals with healthcare disparities to improve HIV and general health outcomes. Further confirmatory studies to analyze the associations between drug use variables and HIV viral load are needed.