Multilevel regression models were used to examine change over time for SSIS

Participants’ responses to the ASSIST during the SBIRT intervention procedure were not used to determine this outcome.Baseline demographic characteristics and SSIS were summarized with descriptive statistics.Major advantages of multilevel regression over traditional repeated measures analysis include the fact that cases are not dropped due to missing observations on the dependent variable at any assessment, numeric as well as categorical predictors can be used, and methods for non-normal outcomes are available. Estimation was carried out in Stata/SE Release 14.1 using maximum likelihood and the Expectation–Maximization algorithm. Models were estimated to examine unconditional change over time for unit-increases over the assessment months; differences in the change trajectories as a function of baseline risk of use for each substance, and differences in the change trajectories due to the intervention by initial risk interaction. Due to strong right-skewness in many of the substance use scores , estimation for the multilevel regression models was carried out with a non-parametric bootstrap with 5000 repetitions to obtain bias-corrected confidence intervals, unaffected by either extreme values or skewness. Te study sample size for the overall study was calculated based on the results of a prior clinic waiting room survey that measured current substance use in 33% of clinic patients . Te sample size estimates were large enough to detect a 15% difference in alcohol use between the Computer and Clinician groups,dry racks with 80% power and 95% confidence.A total of 225 people living with HIV were assessed for eligibility to participate in the study . Of these, seven were excluded because they did not meet eligibility criteria, and 10 were excluded because they did not complete the baseline survey.

Te remaining 208 individuals were enrolled in the study and randomized. These 208 participants were primarily male and largely African American , with a mean age of 45.4 years . Te majority had a high school education or less , were unemployed , and reported substance use in the past 3 months . Te mean time since HIV diagnosis was 12.4 years. Mean Specific Substance Involvement Scores were in the moderate risk range for all substances except inhalants, hallucinogens, and other substances. SSIS were highest for tobacco, alcohol, cannabis, cocaine, and amphetamine. Of the 208 individuals enrolled in the study, 134 individuals completed the baseline assessment visit and also received an SBIRT intervention. Of the 134, follow-up assessment rates were: 123 at 1-month, 106 at 3-month and 109 at 6-month; 92 completed all four study assessments and the intervention. Ninety-five participants with high SSIS accepted referrals to the clinic social worker, but only four met with the social worker. There were no significant baseline differences in socio-demographic characteristics or mean SSIS between those who received the intervention and those who did not . Similarly, we found no differences between SBIRT treatment modality in our outcome measures of interest over time . For all substances, mean SSIS increased over time among those initially in the lower risk groups. Te increase was statistically significant for all substances except amphetamines and sedatives . However, among those individuals with moderate–high risk at baseline, mean SSIS for all substances decreased at 6 months. Te decrease was statistically significant for all substances except tobacco and cannabis. For all substances, the decrease in mean SSIS for the moderate– high risk group differed significantly from the increase in mean SSIS for the lower risk group.We conducted a screening, brief intervention, and referral to treatment intervention in an urban safety-net HIV primary care clinic and detected a high prevalence of self-reported alcohol, tobacco, cannabis, cocaine, amphetamine, sedatives, and opioid use at enrollment. For all substances, the mean SSIS score for participants whose baseline substance use risk was moderate–high and who received the SBIRT intervention declined over the 6 months following the intervention, and this decrease was significant when compared to those at baseline lower risk.

While active substance use was not one of the inclusion criteria, 92% of study participants reported any substanceuse in the prior 3 months. This finding is consistent with the known higher prevalence of substance use for PLHIV compared to the general U.S. national population. Our results also show that, in an HIV primary care population, while mean SSISs were in the moderate range for most substances, a number of individuals were in the high risk range, as indicated by the large standard deviations for each substance. These indicators of the severity of self-reported substance use, underscore the opportunity for detection and intervention in HIV primary care settings. We measured a significant reduction over time in the mean SSIS for alcohol − 1.59 among participants who scored in the medium high risk categories. Several other studies that measured self-report of substance use before and following SBIRT implementation in clinical settings have been conducted and allow for a comparison with the findings of our analysis. One of the first studies to determine the effect of SBIRT in diverse clinic populations found SBI to be associated with a decrease in self-reported alcohol use at follow-up. Other studies evaluating measures of alcohol use severity before and after participating in SBIRT show similar results. A more recent study among PLHIV, however, found that although alcohol use declined over time, the decline was not associated with receipt of a brief intervention. We also measured moderate but statistically significant decreases in mean SSISs for illicit drugs, including reductions in cocaine − 0.82 , amphetamines − 0.69 , sedatives − 1.58 and opioids − 1.31 . Other studies have shown mixed results of the impact of SBIRT on illicit drug use following participation in SBIRT. Te ASPIRE study , a 3-group randomized controlled trial for unhealthy drug use among adults from an urban primary care setting, did not demonstrate a decrease in unhealthy drug use following receipt of a primary care based SBIRT intervention. Other studies have shown similar negative results of the effects of SBIRT on illicit substance use. In contrast, Humeniuk and colleagues found significantly reduced SSISs among participants receiving a brief intervention compared to control participants, for all substances except opioids. And Bernstein and colleagues found reductions in cocaine and heroin use among individuals receiving SBIRT. In our study, we saw a reduction in mean SSIS for tobacco use among participants at moderate–high risk at baseline.

Cropsey and colleagues also found that PLHIV who smoked at least five cigarettes per day significantly reduced their smoking over time following an SBIRT intervention that included a counseling session, nicotine replacement therapy, and follow-up visits, compared to those in usual care. In a pilot study of 30 women living with HIV, those who received a motivational interviewing session reported significant reductions in the mean number of cigarettes smoked, compared to those who did not receive the MI intervention. Surprisingly, we found that mean SSIS scores for participants who scored in the lower risk range at baseline increased over the 6 months for all substances at the same time that use dropped for those in the moderate–high risk group . It is possible that the Brief Intervention that was given to those in the moderate–high risk groups had an important effect on reducing SSIS scores, but that the minimal intervention given to those in the lower risk group was not fully effective at keeping risk levels low. This was particularly the case for tobacco and cannabis; for both of these substances, mean SSISs in the lower risk group increased more than the scores decreased in the moderate–high risk group . In some studies, ASSIST scores were reported as lower,drying racks moderate or high risk. Reporting and analyzing SSIS by risk categories is important because individuals who fall into the lower and moderate risk group may derive different benefits, and because brief interventions have previously been shown to be more effective among people with less severe substance use problems. In contrast to other studies, our outcome observation is based on a mean change score and on dichotomized risk groups, which may or may not be clinically useful distinctions. This area needs further study and exploring more effective interventions for people at lower risk for substance use related problems is an important area for future research.While the levels of substance use self-reported among this cohort of PLHIV patients is higher than in the general U.S. national population, it can be difficult to make comparisons between studies because of the variability of substance use measures. For example, in a study with a safety-net primary care population, participants received an intervention based on reports of problem drug use via the Addiction Severity Index-Lite measure.

Because of differences between the Addiction Severity Index and the ASSIST, similar participants in each study may have been assessed at different levels of risk, and therefore may have differed in whether they qualified to receive an SBIRT intervention or not. Such differences make comparisons difficult. Our study has several limitations. First, the data reported here were all self-reported, which could be biased due to participant recall or social desirability. Second, while these findings suggest that SBIRT delivery in HIV care settings may be associated with a decrease in the mean SSIS scores for moderate–high risk substance use, we do not have a good understanding of the clinical significance of these changes in mean scores. A decrease in the SSIS for any substance is a change in the right direction when our goal is to address substance use in HIV clinical settings. However, in the absence of a non-treatment control group it is possible that the decrease in SSIS scores across both arms of the study could be due to regression to the mean and not the intervention. Third, for this analysis, which examined only those participants who received the intervention, our analytic sample may have been under powered, despite the fact that we enrolled a sufficient number of participants into the study based on our a priori sample size calculations for a randomized trial. Notwithstanding the smaller analytic sample size, we did detect a statistically significant decrease on the moderate–high risk mean SSIS of those who received the SBIRT intervention when compared to those with lower risk scores. Fourth, while the ASSIST measure includes many drugs and does not solely capture the level of use for the substance of concern to the participant, the BI that was delivered by each modality was based on the substance of most concern to the patient. Nonetheless, our use of the ASSIST allowed us to gain a more expansive understanding of the number and types of substances used by this HIV primary care sample and this may be one benefit for using the ASSIST. Use of the ASSIST in clinical settings could have the advantage of giving providers screening and assessment information for multiple substances. Fifth, as part of our study procedures we did not adequately document the number of brief intervention visits either with a clinician or through the computer portal so we were unable to capture meaningful information about the dose of the exposure to the intervention to allow for dose–response analyses. Sixth, our findings may not be generalizable to PLHIV who are not engaged in primary care or to patients of HIV clinics that do not serve an urban safety-net population. Further, while we collected baseline substance use and follow up data on 208 participants, and all participants were assigned to one of two SBIRT modalities, we observed a significant drop of between assessment and participation in the intervention; only 64% of participants who completed the baseline study visit actually received the intervention by either modality, leading to our decision to present an “as treated” analysis. This was particularly the case for those assigned to the computer group, which may indicate difficulties or discomfort with accessing computers and the Internet, or concern about the privacy of data entered into a computer linked to the Internet . In addition, very few of those referred to treatment actually met with the social worker as indicated, possibly indicating that they were not ready to take the initiative to seek out treatment for themselves, and that a more immediate and supported linkage might be needed.