Treatment initiation, substance use trajectories, and the social determinants of health in persons living with HIV seeking medication for opioid use disorder

Abstract

Background.

People living with HIV and opioid use disorder (OUD) are disproportionally affected by adverse socio-structural exposures negatively affecting health, which have shown inconsistent associations with uptake of medications for OUD (MOUD). This study aimed to determine whether social determinants of health (SDOH) were associated with MOUD uptake and trajectories of substance use in a clinical trial of people seeking treatment.

Methods.

Data are from a 2018–2019 randomized trial comparing the effectiveness of different MOUD to achieve viral suppression among people living with HIV and OUD. SDOH were defined by variables mapping to Healthy People 2030 domains: education (Education Access and Quality), income (Economic Stability), homelessness (Neighborhood and Built Environment), criminal justice involvement (Social and Community Context), and recent SUD care (Health Care Access and Quality). Associations between SDOH and MOUD initiation were assessed with Cox proportional hazards models, and SDOH and substance use over time with generalized estimating equation models.

Results.

Participants (N=114) averaged 47 years old, 63% were male, 56% were Black, and 12% Hispanic. Participants reported an average of 2.3 out of 5 positive SDOH indicators (SD=1.2). Stable housing was the most commonly reported SDOH (61%), followed by no recent criminal justice involvement (59%), having a high-school level education or greater (56%), income stability (45%), and recent SUD care (13%). Each additional favorable SDOH was associated with a 25% increase in the likelihood of MOUD initiation during the study period [adjusted HR=1.25, 95% CI=(1.01, 1.55), p=.044]. Positive SDOH were also associated with a decrease in the odds of baseline opioid use and a greater reduction in opioid use during subsequent weeks of the study (p<.001 for a joint test of baseline and slope differences).

Conclusions.

Positive social determinants of health, in aggregate, may increase the likelihood of MOUD treatment initiation among people living with HIV and OUD.

Background

The social determinants of health (SDOH), defined by the World Health Organization as “the conditions in which people are born, grow, work, live, and age, and the wider set of forces and systems shaping the conditions of daily life” ¹, influence a wide range of public health challenges, including the HIV and opioid epidemics. These fundamental non-medical contributors such as income, education, housing, and access to healthcare may be a useful framework for understanding the complex barriers that many persons living with HIV encounter when deciding to initiate treatment for opioid use disorder (OUD). HIV and OUD are highly comorbid ^2,3, overlapping public health crises, rooted in and amplifying the impact of socioeconomic and health disparities in communities. Low socioeconomic status, unemployment, criminal justice involvement, and housing instability are circumstances faced by many people living with both HIV and OUD ^4–6. Having OUD increases the risk of acquiring HIV and vice-versa ⁷, and addressing common risk factors and root causes such as SDOH is key to ending both epidemics ⁸.

FDA-approved medications for opioid use disorder (MOUD), including methadone, buprenorphine, and extended-release naltrexone, decrease opioid use and mortality while decreasing HIV risk behaviors ⁹ and increasing antiretroviral (ART) uptake, retention in HIV care and viral suppression in people living with HIV and OUD ^10–12. Despite these benefits, MOUD treatment initiation is often suboptimal. Nationally, only about one in four people with OUD receive MOUD ^13,14 and comorbid HIV reduces that likelihood by about half ², although uptake varies widely across populations. For example, although 65% of Medicaid beneficiaries living with HIV and OUD in New York state received MOUD ¹⁵, only 5% of veterans initiated medication treatment within thirty days of their first OUD treatment encounter, and patients with HIV were less likely than their peers to receive MOUD during that period ¹⁶. Treatment availability, medication type, and provider characteristics could partially explain some of these challenges, but even in OUD treatment studies where medication or referral was offered to all research participants, initiation rates ranged from 41% to 96% ^17–19.

Increased positive SDOH is associated with more positive HIV outcomes ²⁰; however, the role that the SDOH play in MOUD initiation is understudied, with mixed findings ^21,22. Homelessness and income-assistance have been positively and negatively correlated with engagement in MOUD treatment ^23–25, while lack of insurance, minority race/ethnicity, and the presence of medical comorbidities may negatively affect treatment uptake ^25–27. Among women living with HIV, adversities and economic hardship were correlated with greater drug use than their peers during HIV treatment ²⁸, which may reduce the likelihood of MOUD initiation ^29,30. Access to care, treatment setting, type of MOUD, methodology and rigor of research, and a multitude of other factors may explain heterogeneity in results, and further studies are warranted.

People living with both HIV and OUD are a highly prevalent, especially marginalized, high transmission-risk group with outsized implications for ending the HIV and OUD epidemics. Therefore, we investigated associations between five supportive SDOH and MOUD initiation, opioid use, and other substance use over a six-month period in persons living with HIV enrolled in an OUD treatment randomized trial. We hypothesized that increased supportive SDOH would be associated with increased likelihood of MOUD initiation and reduced substance use over the trial period.

Methods

Study Sample

Data are from the NIDA Drug Abuse Treatment Clinical Trials Network (CTN) study “Comparing Treatments for HIV-Infected Opioid Users in an Integrated Care Effectiveness Scale-up Study” (CHOICES, CTN-0067, ClinicalTrials.gov NCT03275350) ¹⁸. CHOICES was a open-label, randomized, non-inferiority trial comparing a long-acting injectable MOUD (extended-release naltrexone) to office-based buprenorphine or methadone. The primary outcome was achievement of HIV viral suppression at 24 weeks; findings have been previously published ¹⁸. Advarra and Cook County Health Institutional Review Boards reviewed and approved the study ³¹. Enrollment took place between March 2018 and May 2019; eligible participants had unsuppressed HIV (HIV viral load ≥200 copies/mL), moderate to severe OUD, and had received no MOUD in the four weeks prior to enrollment. All participants were willing/interested in beginning MOUD treatment and were given equivalent access to MOUD as part of study participation (conditional on treatment assignment). The study enrolled 114 participants who were randomized to extended-release naltrexone (n=55) or treatment as usual (n=59) and were followed for six months. Extended-release naltrexone was administered by study clinicians in HIV clinics, while individuals in the treatment as usual group were offered medication by the study clinician (buprenorphine) or provided with a referral and linkage assistance (methadone or buprenorphine). Trained research staff administered computer-based questionnaires face-to-face or over the phone at all study visits, which occurred at least monthly. Participants provided a urine sample monthly, which was tested for opioids (excluding methadone and buprenorphine) and other drugs using a point of care urine drug screen panel. Detailed study recruitment strategies, methods, and main outcomes have been previously described ¹⁸.

Social Determinant of Health Indicators

CHOICES participants completed a series of standardized questionnaires at baseline. The study was not designed to measure SDOH, and therefore, all desired SDOH indicators were not available in our data set. However, we selected and created indicator variables relating to the U.S. Department of Health and Human Services’ Healthy People 2030 SDOH priority areas ³². The variables map to the following Healthy People 2030 domains: high school education or greater (Education Access and Quality), currently receiving income from employment or disability (Economic stability), no homelessness in the past 6 months (Neighborhood and Built Environment), no arrest, incarceration, probation or parole in past 3 months (Social and Community Context), and SUD care of any kind in the past 28 days, including inpatient, outpatient, or individual or group counseling, excluding 12-step programs, self-help, or detoxification-only inpatient stays (Health Care Access and Quality). For this analysis, we conceptualized SDOH as positive resources or supports and coded and labeled them as follows: education (completed high school or greater = 1), current economic stability (current employment or disability income = 1), housing stability (housed = 1), no criminal justice involvement (none = 1), and SUD treatment (any non-medication treatment, as defined above, in the 28 days prior to study baseline = 1). All SDOH data were participant self-reported.

Outcomes

MOUD treatment data were abstracted approximately monthly for six months from clinical trial treatment logs, covering the entire study period for each participant. MOUD initiation date was defined as the first day of the first abstraction period where a prescription for buprenorphine, an injection of extended-release naltrexone, or a dose of methadone was received (the exact dates of MOUD receipt were not recorded, only whether or not MOUD was received during each abstraction period). Participants who did not initiate MOUD were censored at the last day of their final abstraction period. Participants provided urine drug screens (UDS) and completed timeline follow-back assessments of substance use monthly for six months. Opioid use at each timepoint was defined as having either a positive UDS for opioids (excluding methadone and buprenorphine) or any self-reported use of heroin, prescription opioids, fentanyl, methadone, or buprenorphine to get high during the assessment window. In this study, “other drugs” included methamphetamine, cocaine, benzodiazepines, or heavy alcohol use (5+/4+ drinks on 5+ occasions in one month for men/women). Similar to opioids, participants were classified as having used other drugs if they were UDS positive for or self-reported any use of methamphetamine, cocaine, or benzodiazepines or self-reported heavy alcohol use during an assessment window. Missing opioid/other drug use data points were imputed as positive, as in the parent clinical trial ¹⁸; rates of missing data averaged 17% at each timepoint and were highest at the 20-week assessment (26%). We also conducted sensitivity analyses omitting missing drug use datapoints (treating missing data as completely at random).

Statistical Analyses

Participant characteristics and rates of SDOH were summarized with descriptive statistics. Cumulative incidences of MOUD initiation by SDOH were described using Kaplan-Meier curves and analyzed with Cox proportional hazards models, controlling for age, sex, race, ethnicity, baseline pain score, history of serious psychiatric conditions, injecting opioids in 30 days prior to baseline assessment (binary), and treatment arm (extended-release naltrexone vs. treatment as usual). SDOH were treated individually, as both binary indicator variables and also a continuous variable representing the sum total of SDOH supports. Prior to analysis, the proportional hazards assumption was verified by examining and testing for an association between scaled Schoenfeld residuals and (transformed) study time. The assumption was violated for the analysis of recent SUD care; based on graphical examination, two hazard ratios were estimated, one for the first 30 days of the study (i.e., the first abstraction period for MOUD prescription data) and another for subsequent days.

Associations between SDOH and opioid use over time were analyzed using logistic GEE models with autoregressive working correlation structures (chosen by quasi-information criteria), controlling for the same covariate set as above. Prior to model fitting, plots of observed proportions of opioid use over time were examined. A number of potential specifications of the time effect were considered, and a linear spline model with a single knot at 4 weeks (a “bent-line” model) was chosen as the optimal compromise between interpretability and fit to the observed data. As the data come from an opioid use treatment trial and many participants initiated MOUD early in the study, there is a strong theoretical rationale for estimating different rates of change in opioid use early and later in the study. Associations between SDOH and other drug use were examined similarly, except with time effects modeled using cubic B-splines with a single knot at the median assessment time (12 weeks). Single or joint hypothesis tests of time effect parameters were conducted to determine whether changes in opioid/other drug use differed between SDOH groups. All analyses were conducted using R v.3.6.2 with the ‘survival’, ‘geepack’, and ‘splines’ packages at a two-tailed level of significance of .05.

Results

Participant Characteristics and SDOH

Participants (N = 114) averaged 47 years old (SD = 11.1); 63% were male; 56% were Black; and 12% were Hispanic. Out of five possible, participants reported an average of 2.3 supportive SDOH (SD = 1.2), with 6 participants reporting zero SDOH (5%), 23 reporting one (20%), 32 two (28%), 33 three (29%), 19 four (17%), and a single participant reporting all five (1%). Stable housing was the most commonly reported supportive SDOH (n = 70, 61%), followed by no recent criminal justice involvement (n=67, 59%) and a high-school level education or greater (n=64, 56%). Less than half of participants reported economic stability (n=51, 45%) and few had any kind of recent SUD care (n=15, 13%). Table 1 provides further detail on participant characteristics and SDOH.

Table 1.

Participant baseline characteristics and social determinants of health, N = 114 CHOICES participants

Characteristic	n(%)	Mean # SDOH (SD)
Age group
≤30	12 (10.5%)	1.5 (1.2)
31–40	22 (19.3%)	1.7 (1.2)
41–50	32 (28.1%)	2.2 (1.2)
51+	48 (42.1%)	2.8 (1)
Gender
Male	71 (62.3%)	2.3 (1.2)
Female	43 (37.7%)	2.3 (1.2)
Race
Black	64 (56.2%)	2.6 (1)
White	42 (36.8%)	1.7 (1.3)
Other	8 (7%)	2.2 (1.3)
Ethnicity
Hispanic	14 (12.3%)	1.8 (1.1)
Non-Hispanic	100 (87.7%)	2.3 (1.2)
SDOH	n(%)	Mean # other SDOH (SD)
Economic stability
Current income from work or disability	51 (44.7%)	2.1 (0.8)
Neighborhood/Physical environment
Stably housed in past 6 months	70 (61.4%)	1.8 (0.9)
Education
High school or more	64 (56.1%)	1.6 (1.1)
Community and social context
No criminal justice involvement in past 3 months	67 (58.8%)	1.9 (0.9)
Health care system (SUD Care)
Received SUD treatment in past 28 days	15 (13.2%)	2.6 (0.7)

SDOH = social determinants of health; SD = standard deviation; SUD = substance use disorder;

SDOH and MOUD Initiation

Overall, 81 out of 114 participants initiated MOUD during the study (71%). Participants randomized to the extended-release naltrexone arm were less likely to initiate their assigned medication than those randomized to treatment as usual (47% vs. 73%), and twelve participants randomized to extended-release naltrexone started buprenorphine during the study period. Kaplan-Meier cumulative incidences of MOUD initiation by SDOH are presented in Figure 1. When examining the cumulative effect of SDOH as a linear covariate in Cox regression, each additional SDOH was associated with a 25% increase in the likelihood of MOUD initiation during the study period [adjusted HR = 1.25, 95% CI = (1.01, 1.55), p = .044]. There was no evidence that the association between SDOH and MOUD initiation was dependent on treatment assignment (p for interaction = .65) or that the impact of cumulative SDOH was nonlinear (p for quadratic effect = .15, cubic effect = .45).

Figure 1.

Kaplan-Meier cumulative incidence of medication for opioid use disorder (MOUD) initiation by social determinants of health. CJI = criminal justice involvement; SUD = substance use disorder.

Recent non-medication SUD care was associated with 3.7 times the likelihood of MOUD initiation within the first 30 days of study participation [aHR = 3.72 (1.76, 7.78), p < .001], but not during the remainder of follow-up. There was some evidence that participants who were stably housed [aHR = 1.53 (0.86, 2.73), p = .145] and those with at least a high-school education [aHR = 1.6, (0.9, 2.6), p = .07] were also more likely to initiate MOUD, although those effects did not reach the threshold of statistical significance. There was little evidence that recent criminal justice involvement or economic stability affected treatment initiation. Full Cox regression results are presented in Table 2.

Table 2.

Associations between supportive SDOH and MOUD initiation, N = 114 CHOICES participants

SDOH	Number initiating MOUD/N (%)	aHR (95% CI)	p
Linear effect of +1 SDOH		1.25 (1.01, 1.55)	.044
Economic stability
Current income from work or disability	37/51 (73%)	1.10 (0.66, 1.84)	.709
No stable income	44/63 (70%)	Ref
Stable housing
Stably housed in past 6 months	52/70 (74%)	1.53 (0.86, 2.73)	.145
Homeless in past 6 months	29/44 (66%)	Ref
Education
High school or more	51/64 (80%)	1.59 (0.96, 2.63)	.072
Less than high school	30/50 (60%)	Ref
Criminal justice involvement
No CJI in past 3 months	48/67 (72%)	1.01 (0.60, 1.64)	.967
CJI in past 3 months	33/47 (70%)	Ref
Substance use disorder treatment
MOUD initiation in first 28 days in study
Received non-medication SUD treatment in 28 days prior to baseline	10/15 (67%)	3.72 (1.77, 7.85)	<.001
No SUD treatment in past 28 days	35/99 (35%)	Ref
MOUD initiation during subsequent days
Received non-medication SUD treatment in 28 days prior to baseline	1/5 (20%)	0.32 (0.04, 2.42)	.271
No SUD treatment in past 28 days	35/64 (55%)	Ref

SDOH = social determinants of health; MOUD = medication for opioid use disorder; aHR = adjusted hazard ratio; CI = confidence interval; CJI = criminal justice involvement; SUD = substance use disorder

SDOH and Opioid Use

Observed and model-fitted probabilities of opioid use over time by SDOH are depicted in Figure 2. Estimated slopes (on the log-odds scale) and hypothesis test results are presented in Table S1. When considering the cumulative effect of SDOH as a linear covariate, each additional SDOH support was associated with a substantial decrease in the log-odds of baseline opioid use and a greater reduction in opioid use during subsequent weeks of the study [p < .001 for a test of overall difference (Figure 2) and p < .05 for each individual effect (Table S1)]. Stable housing (p = .043), recent SUD care (p .002), and having a stable income (p = .051) were associated with, or trended towards association with, lower odds of opioid use early in the study. However, those differences attenuated over time (Figure 2). No statistically significant relationships between education or criminal justice involvement and opioid use were observed. Sensitivity analyses treating missing opioid use datapoints as completely at random did not show substantial changes in patterns of use over time (Table S2, Figure S1).

Figure 2.

Observed (dots) and fitted (lines) probability of opioid use over time by social determinants of health (SDOH), N = 114 CHOICES participants. p values are from hypothesis tests of any difference in rate of change over time by SDOH. CJI = criminal justice involvement; SUD = substance use disorder.

SDOH and Other Drug Use

Proportions of participants using other drugs by SDOH are presented in Figure 3. When analyzed as a cumulative linear covariate, participants with more positive SDOH were more likely to not change or decrease other drug use early in the study, while those with fewer positive SDOH were more likely to increase. Regression to the mean was noted in all groups in the second half of the study, resulting in groups being quite similar at 24 weeks. Substantially less use of other drugs at baseline was observed among participants with recent SUD care. No other differences in baseline use of other drugs or changes in other drug use over time were associated with SDOH (Figure 3). Sensitivity analyses treating missing other drug use datapoints as completely at random did not show substantial changes in patterns of use over time (Figure S2).

Figure 3.

Observed (dots) and fitted (lines) probability of other drug use over time by social determinants of health (SDOH), N = 114 CHOICES participants. CJI = criminal justice involvement; SUD = substance use disorder.

Discussion

These data suggest that positive social determinants of health, in aggregate, may increase the likelihood of MOUD treatment initiation among people living with HIV and OUD. Recent non-medication SUD care, which we included as a marker of healthcare access, positively impacted early MOUD initiation. People with more positive SDOH had substantially lower baseline opioid use, a steeper reduction in use during the first month of the trial, but an attenuation of effects in months two through six compared to those with fewer SDOH supports. Use of other substances over time showed fewer differences associated with SDOH. Our study contributes novel, longitudinal data suggesting that positive SDOH contribute to increased MOUD treatment uptake and a stronger treatment effect, at least initially.

People living with HIV and OUD are some of the most marginalized in the US, and are disproportionately affected by lack of jobs, education, income, access to healthcare and increased involvement in the criminal-legal system ^33–35. The COVID-19 pandemic exacerbated these disparities, which may have amplified inequalities in healthcare access, reduced treatment engagement, and resulted in substantially worse COVID, HIV, and substance use outcomes ^36–38. Our results emphasize the value of targeting these and other “upstream” societal and structural determinants of OUD outcomes in this vulnerable population. Addressing root causes such as the SDOH to improve OUD outcomes may require large and complex interventions, if not major societal shifts, but promise has been shown by supportive housing programs ³⁹, employment assistance and skills training ⁴⁰, integration of OUD care into the criminal-legal system ⁴¹, large-scale community opioid education programs ⁴², collegiate recovery communities ⁴³, and multifactorial initiatives such as Massachusetts’s “Access to Recovery” program ⁴⁴.

Inequalities in SDOH often result in healthcare access gaps, which may be a principal mediator of poor outcomes ⁴⁵. Importantly, in this clinical trial, all participants had relatively equivalent access to MOUD (conditional on their randomized treatment assignment) and all expressed willingness to initiate treatment (inclusion criteria). Still, 29% of randomized participants did not receive medication during the six-month trial, and differences between initiators and non-initiators could not be attributable to lack of access. Our finding that greater cumulative SDOH were linked to increased likelihood of engaging in evidence-based, lifesaving treatment supports holistic and intersectional models of health disparities. These findings further suggest that access is not enough; rather, improving a broad range of SDOH is key to reducing disparities ^46–48.

The only individual SDOH predicting MOUD initiation was having a recent history of non-medication SUD treatment, an association supported by previous studies ^24,49. There was some evidence that higher levels of education and stable housing positively impacted treatment engagement, although these findings did not reach the threshold for statistical significance. Significant barriers impede access to OUD care for those who lack stable housing ^25,50. Our study suggests that there may be additional benefits associated with stable housing once access is established; furthermore, housing interventions have shown numerous benefits to HIV and OUD outcomes ^39,51,52.

Analyses of opioid use over time showed a steeper reduction in early use associated with more positive SDOH, generally. This likely follows from increased rates of MOUD initiation among those with more SDOH support. Reductions in opioid use attenuated over time, possibly as participants discontinued MOUD or were lost to follow-up. Retention on MOUD remains a major challenge; less than half of people who initiate MOUD are expected to continue beyond 6 months ¹³. Poor SDOH are associated with reduced MOUD retention ^53,54. Despite engagement in OUD treatment, we observed that other drug use including stimulants, benzodiazepines, heavy alcohol consumption, and other drugs (cannabis excluded) was frequent and relatively stable, consistent with previous research ^55,56. Baseline use of other substances tended to be lower among people with more positive SDOH, but we found no relationship between the number of SDOH and changes in other substance use throughout the study. This is not surprising given that the focus of this clinical trial was treatment of OUD. Especially considering methamphetamine’s role in HIV transmission ⁵⁷ and the high degree of co-use of opioids and methamphetamine ⁵⁸, our data support increasing calls to better address polysubstance use during OUD care ⁵⁶. Medications such as combined bupropion/extended-release naltrexone for methamphetamine use disorder ⁵⁹ and interventions such as contingency management ⁶⁰, integrated harm reduction ⁶¹, and peer support services to increase retention ⁶² are potential strategies that could be incorporated.

The major limitation to our analyses, especially of opioid and other drug use over time, is that SDOH were not measured longitudinally. Some SDOH, such as housing status or criminal justice involvement, may fluctuate rapidly in this population. By examining only baseline levels of these variables, we risk missing important relationships between time-varying SDOH, MOUD treatment, and longitudinal outcomes. Including longitudinal and time-to-event outcome data is a strength, but a more complete evaluation of the SDOH should also include multiple exposure measurements to better understand how these factors change, and how those changes play a role in OUD treatment. We also acknowledge that, as a secondary analysis of a clinical trial, we were unable to measure several important SDOH and characteristics which may be relevant to treatment initiation and outcomes. We also caution that these results come from a single clinical trial. While clinical trial participants are not usually representative of “real-world” OUD patients ⁶³, this particular trial included historically disenfranchised groups: mostly Black and Latino participants, a substantial number of patients without stable housing, and many individuals with other substance use. These groups are disproportionally impacted both by HIV and OUD. We also recognize that imputing missing UDS data as positive is a suboptimal, but common practice in SUD treatment trials ⁶⁴. We chose this strategy to be consistent with the analytic strategy utilized in the parent clinical trial and completed sensitivity analyses treating missing data as completely at random. Finally, we recognize that this was a small clinical trial, and tests to detect nonlinear relationships between SDOH and MOUD initiation as well as associations between individual SDOH and outcomes may have been underpowered. The true burden of health disparities caused by cumulative, or intersectional, marginalizing factors is likely to be greater than suggested by a linear relationship.

Conclusions

Interventions targeting SDOH are increasingly recognized by the National Institute on Drug Abuse and others as vital to mitigating harms of the opioid epidemic and preventing future use disorders ⁶⁵. Our study suggests that modifiable social determinants of health, including income, education, housing, criminal justice involvement, and engagement in SUD treatment may impact MOUD initiation and opioid use outcomes among people living with HIV and OUD. These findings, when added to the body of literature showing similar associations, provide possible intervention targets for future experimental studies. Especially following the steep rise in inequalities and overdose deaths during the COVID-19 pandemic, which disproportionately affected people of color, people experiencing homelessness, and those with co-occurring conditions including HIV, results highlight the need to address SDOH as part of OUD care ⁴⁸.

Highlights:

• Among people with opioid use disorder and HIV, positive social determinants of health may increase the likelihood of medication treatment initiation.

• Medication for opioid use disorder may be more effective, at least early on, for those with more positive social determinants of health.

• Results support calls to address root causes of health inequality, such as the social determinants of health, as part of opioid use disorder and HIV care.