Abstract
<p>Both HIV/AIDS and STD’s represent significant financial and public health burdens in the United States. Substance use and sexual risk behaviors play major roles in HIV/AIDS transmission, as well as in the spread of sexually transmitted infections (STIs). In this dissertation, we aim to 1) describe sexual risk and IDU behaviors and examine associations between viral suppression and HIV risk behaviors among hospitalized HIV-positive patients; 2) develop a new approach to examine the count of sexual acts using a fully disaggregated repeated measures design and show how this new approach can be used to evaluate the impact of substance use on sexual risk behaviors; 3) perform a novel multivariate analysis of sexually transmitted infections using random forest (RF) to identify replicable relationships among substance use, sexual risk behaviors, and STIs using the new sexual behavior measures created in 2). The first manuscript used screening data from the Project HOPE study which included 2,291 HIV-infected patients from 11 hospitals in the U.S. Among those screened, 765 (33.4%) identified as women, 1116 (48.8%) as heterosexual men, and 408 (17.8%) as men who have sex with men. Overall, viral suppression (viral load < or = 200 copies/ml) was 25.2%. Among virally suppressed patients, 40.4% reported a history of IDU, compared to 27.2% of virally unsuppressed patients (χ2=23.0, df=1, p<.0001). Viral suppression did not impact IDU in the last 12 months or the sharing of needles/injection paraphernalia. Virally suppressed patients were less likely to engage in sex acts while high on drugs or alcohol (15.7% vs. 22.1%; OR=0.64, 95% CI=0.46-0.88, p=.007), and were less likely to have sex with individuals whose HIV status was negative or unknown (19.6% vs. 26.3%; OR=0.67, 95% CI=0.49-0.90, p=.008) compared to those who were virally unsuppressed. In the second manuscript, we create a more sensitive approach to assess how the relative characteristics of the sex acts may determine the level of risk in which an individual choses to engage. Particularly, we (1) describe a new approach to examine the count of sexual acts using a disaggregated repeated measures design, and (2) show how this new repeated measures approach can be used to evaluate interactions among different categories of sexual risk behaviors and other predictors of interest (such as gender/sexual orientation). Profiles of different subtypes of sexual acts in the past 6 months were assessed. Potential interactions of the characteristics associated with each subtype of sex acts which resulted up to 48 distinct sub-types of sexual risk behaviors—with a primary/non-primary partner; partner’s HIV status; vaginal/anal; condom use, and substance use before or during sex act—can be examined. Specifically, we chose condom use, and primary and non-primary status of partner as an application in this paper to illustrate our method. There are significantly more condomless sex acts (M=23, s.e.=0.9) and sex acts with primary partners (M=27.1, s.e.=0.9) compared to sex acts with condom use (M=10.9, s.e.=0.4, IRR=2.10, 95% CI=1.91-2.32, p<.001) and sex acts with non-primary partner (M=10.9, s.e.=0.5, IRR=2.5, 95% CI=2.33-2.78, p<.001). In addition, there are significant differences for the count of sexual risk behaviors among women who have sex with men (WSM), men who have sex with women (MSW) and men who have sex with men (MSM) for sex acts with and without condom use, primary and non-primary partner, and their interaction (ps=.03, <.0001 and .001, respectively). In the third manuscript, we aim to how substance use and sexual risk behaviors are related to sexually transmitted infections (STIs) may help to tailor STI/HIV risk reduction interventions. Random forests, a machine learning technique, provides a principled approach to explore a large number of effects including interactions to identify replicable sets of predictive factors. We used data from Project Aware, a randomized clinical trial conducted among 5012 patients in 9 sexually transmitted disease clinics in the US. Predictive models for both prevalence and incidence of STIs were created, respectively. Socio-demographics, substance use, sexual risk behaviors, and characteristics of sexual networks were assessed and examined using a random forest machine learning approach. High accuracies in prediction were achieved for both STI prevalence and incidence outcomes. Particularly, socio-demographics such as ethnicity, gender, age, education, incarceration history, site of recruitment, employment status and previous STI histories are top ranked variables that related to STI prevalence. Several different types of count of sexual risk behaviors, including sex acts specificities (condomless, substance use, insertive/receptive) and partner specificities (sex with primary or non-primary partners, multiple sex partners) showed stronger associations with STI incidence. Conclusions: Although significantly more HIV virally suppressed patients reported ever being an IDU compared to virally unsuppressed patients, IDU prevalence in the past 12 months was relatively low and not significantly different between the groups. Patients who were virally unsuppressed reported more sexual risk behaviors than suppressed individuals. Developing interventions targeting sexual behaviors among virally unsuppressed patients has the potential to improve individuals’ and the public’s health. The repeated measure approach can extend our understanding of how people make choices among sexual risk behaviors and may be useful in future research on disaggregated characteristics of sex acts. As for the machine learning analysis, results provide initial support for use of random forests to predict STI. A challenge with these methods is the lack of a statistical test for significance of individual variables; nevertheless, these methods are useful for exploratory, model-building in substance abuse and sexual health research.</p>