Moderating Effects of Social Support on the Relationship Between Substance Use Disorders and HIV Viral Load and Medication Adherence Among Black Women Living with HIV in the United States

Abstract

Substance use may predict lower rates of antiretroviral therapy (ART) adherence and viral suppression among Black Women Living with HIV (BWLWH). We assess how perceived social support (PSS) moderates the relationship between substance use disorder (SUD) and viral load (VL) and ART adherence. 119 BWLWH provided information on PSS, SUD, alcohol use disorder (AUD) and adherence (via Wisepill). Higher PSS from friends (β=−.263, p=.008) and significant others (β=−.219, p=.025) predicted lower VL. SUD predicted low Wisepill adherence (a) in the past week (β=−.273, p=.035) and past two weeks (β=−.273, p=.033) only for women low in PSS from friends, and (b) in the past two weeks only for women low in total PSS (β=−.294, p=.024). Ironically, for women high in total PSS, SUD predicted higher adherence in the past four weeks (β=.360, p=.006). Adherence and VL may be bolstered by understanding the role of various forms of PSS and SUDs.

Introduction

In 2016, Black women living with HIV (BWLWH) in the U.S. constituted 60% of all new diagnoses among women, which is related to high rates of poverty, unstable housing, and structural inequities (Bradley et al, 2019; Levy et al., 2014). Furthermore, substance use, which is often used to cope with adversity, has been common in individuals with trauma histories and poor social support systems (Bonn-Miller et al., 2007; Cross et al., 2015; Mitchell et al., 2016; Ullman et al., 2013). Substance use has also been shown to directly relate to antiretroviral therapy (ART) adherence, which is low among many ethnic/racial minority groups (Cohn et al., 2008; Dale et al., 2014; Mellins et al., 2009; Simoni et al., 2012). An understanding of the interconnected roles of substance use, trauma and social support among BWLWH, and its relation to ART adherence is key to reducing HIV-related health disparities.

Substance use is prevalent among people living with HIV (PLWH) and has been negatively associated with ART adherence among diverse samples (Aralis et al., 2018; Cohn et al., 2008; Dawson-Rose et al., 2017; Mellins et al., 2009). However, research on the relationship between substance use and ART adherence among BWLWH is limited to the specific substance of cocaine (Gonzalez et al., 2011; Sharpe et al., 2004). Research has shown positive associations between substance use and viral load; however, these findings apply to general PLWH and is not specific to BWLWH (Aralis et al., 2018; Blank et al., 2015; Coviello et al., 2018; Feldman et al., 2019; Sullivan et al., 2015). One study with Black PLWH has focused on injection drug use; therefore, the generalizability to other methods of drug use is unclear (Robinson & Knowlton, 2016). In light of significant health disparities experienced at the intersection of race and gender among BWLWH with histories of trauma/abuse, further research in this area is necessary (Althoff et al., 2014; Beer et al., 2016; Dale et al., 2019; Meditz et al., 2011).

Alcohol use has also been negatively associated with medication adherence (Azar et al., 2010; Conen et al., 2009; Parsons et al., 2007) and positively associated with viral load, though these findings differed based on the specific types of alcohol used (for example, liquor versus wine/beer) (Baum et al., 2010; Miguez-Burbano et al., 2009). These findings, which have included racially heterogenous samples, suggest a nuanced relationship between alcohol use and HIV health outcomes.

Conversely, the more adaptive coping resource of social support has been positively related to viral suppression as well as medication adherence (Anderson et al., 2018; Friedman et al., 2017). The association between social support and ART adherence also varies based on the source of social support (e.g. friends or family; Anderson et al., 2018; Dale & Safren, 2018a; Hamilton et al., 2007). The aforementioned studies were conducted with samples of men who have sex with men (MSM) and people living with and without HIV from diverse racial/ethnic backgrounds. There is scant evidence of the effects of social support from various sources on viral load and adherence among BWLWH with histories of trauma.

The complex relationships between substance use, social support and HIV health outcomes have been examined as a mediator and covariate in various studies that explore HIV viral load, HIV survival, and medication adherence as outcomes (Gonzalez et al., 2011; Ironson et al., 2016; Wilson et al., 2014). Of note, substance use and social support have not been examined as a moderator, while gender has been examined as a covariate and moderator in relation to HIV health outcomes (Robinson & Knowlton, 2016; Wilson et al., 2014). Current literature suggests the complex interplay of gender, social support and substance use in relation to HIV health outcomes. However, it remains to be seen whether these complex relationships are evident among BWLWH, as previous research included predominantly male and racially heterogenous samples.

The research evidence indicates that, among various populations living with HIV, there are connections between a) substance use and b) social support and HIV-related health outcomes of viral suppression and ART adherence. However, given the health-related disparities faced by BWLWH based on their race and gender, there is need for more research on a broader range of substances than previously explored and their relation to social support. Expanding this research may inform intervention strategies to bolster social support among BWLWH who have histories of trauma and substance use. The present study’s aim is to examine whether social support moderates the relationship between substance use and HIV viral load and medication adherence among BWLWH.

Methods

Participants

Data was collected between October 2017 and January 2019 as part of a behavioral medicine intervention development trial which sought to address trauma symptoms and improve coping strategies and medication adherence (Dale & Safren, 2018). Recruitment efforts involved visits to community outreach events and organizations (e.g., hospitals, community clinics, transitional housing and recovery programs, and colleges) by study staff members, where flyers were distributed and relationships were established and built with community partners. Interested individuals were screened via telephone (or in person if desired). Of 135 eligible participants, 3 declined participation after being screened, and 16 did not attend the baseline visit. Women were deemed eligible to attend a baseline interview in person if they met the following criteria: 1) Black or African American, 2) ≥ 18 years of age, 3) Cis-gender female, 4) English speaking, 5) Prescribed antiretroviral therapy for the past two months, at a minimum, 6) Had a history of trauma, and 7) Indicated the possibility of low ART adherence, detectable viral load within the past year and/or missed HIV-related medical visits within the past year.

Participants were invited to participate in baseline assessments that spanned two weeks/visits at the research institution. At these visits, participants gave written informed consent, completed self-report surveys using Research Electronic Data Capture ([REDCap], a secure web-based application) (Harris et al., 2009) and participated in a semi-structured clinical interview. Participants were reimbursed for their time and effort with a stipend of $25 for each baseline visit. These study procedures and measures were approved by the Institutional Review Board of the (University of Miami).

Measures

Sociodemographic variables.

These were measured by a sociodemographic survey that asked for participants’ demographic information including age, race, gender and HIV viral load (see Table 1).

Table 1.

Socio-Demographic, Clinical, and Health Characteristics of Participants (N=119)

Characteristics	Mean (SD) / N (%)
Age	49.27 (10.81)
Race
African-American / Black	118 (99.2)
Asian	0
White	1 (.8)
Native Hawaiian or Other Pacific Islander	0
Native American	0
Other	0
I choose not to answer	0
Hispanic / Latino	3 (2.7)
Education
Eighth grade or Lower	6 (5)
Some high school	35 (29.4)
High school graduate/GED	41 (34.5)
Some college	29 (24.4)
College graduate	7 (5.9)
Some graduate school	1 (.8)
Income
Less than $5,000	43 (36.1)
$5,000 - $11,999	32 (26.9)
$12,000 - $15,999	8 (6.7)
$16,000 - $24,999	6 (5)
$25,000 - $34,999	3 (2.5)
$35,000 - $49,999	2 (1.7)
$50,000 and greater	3 (2.5)
Employment Status
Full time work	8 (6.7)
Part time work	8 (6.7)
Full time or part time in school	5 (4.2)
Neither work nor school	26 (21.8)
On Disability	69 (58)
Other	6 (5)
I choose not to answer	4 (3.4)
Relationship Status
Married	17 (14.5)
Cohabiting relationship, unmarried	18 (15.4)
Non-cohabiting relationship	15 (12.8)
Single	53 (45.3)
Divorced/Separated	11 (9.4)
Widow or Loss of Partner	3 (2.5)
Housing Arrangement
Renting home or apartment	80 (67.2)
Living in home or apartment owned by you or someone else in household	14 (11.8)
Publicly subsidized housing (Section 8)	13 (10.9)
A friend or relative’s home/apartment (pays little or no rent)	7 (5.9)
Homeless: sleeping in a shelter	2 (1.7)
Homeless: sleeping on the street, beach, car, etc.	1 (.8)
I choose not to answer
Living Situation
Self	64 (53.8)
Partner / Spouse	27 (22.7)
Roommates	6 (5)
Children	31 (26.1)
Group home / residential treatment
Other	22 (18.5)
I choose not to answer	3 (2.5)
Religion
Christian	32 (27.4)
Catholic	5 (4.3)
Baptist	62 (53)
None	9 (7.7)
Other	9 (7.7)
Sexual Orientation
Exclusively heterosexual (straight)	91 (79.8)
Heterosexual with some gay experience	11 (9.6)
Bisexual	8 (7)
Exclusively gay	4 (3.5)
Multidimensional Scale of Perceived Social Support	56.51 (19.05)
HIV Viral Load Log	2.01 (1.503)
Wisepill ART Adherence (%)
Past week	92.97 (18.84)
Past two weeks	92.56 (19.39)
Mini-International Neuropsychiatric Inventory – Substance Use
Diagnosis met	28 (23.9)
Diagnosis not met	89 (74.8)
Mini-International Neuropsychiatric Inventory – Alcohol Use
Diagnosis met	14 (11.8)
Diagnosis not met	103 (86.6)
Self-report ART Adherence
Past Week
Very Poor	3 (2.5)
Poor	1 (0.8)
Good	23 (19.3)
Very Good	24 (20.2)
Excellent	49 (41.2)
Fair	19 (16)
Past Two Weeks
Very Poor	2 (1.7)
Poor	2 (1.7)
Good	27 (22.7)
Very Good	30 (25.2)
Excellent	40 (33.6)
Fair	18 (15.1)
Past Four Weeks
Very Poor	4 (3.4)
Poor	6 (5)
Good	34 (28.6)
Very Good	26 (21.8)
Excellent	39 (32.8)
Fair	10 (8.4)

Perceived Social Support.

This was measured using the Multidimensional Scale of Perceived Social Support [MSPSS] (Zimet et al., 1988). This 12-item scale assesses levels of PSS from three sources (Significant Other, Family, and Friends) and Total PSS. The MSPSS uses a seven-point rating scale ranging from very strongly disagree (1) to very strongly agree (7). The MSPSS Total scale and subscales have demonstrated good internal consistency reliability (.85 < α < .91) and construct validity (r = −.25; Zimet et al., 1988).

Substance use.

Drug and alcohol use were measured in part via the Alcohol Use and Substance Use modules of the Mini-International Neuropsychiatric Inventory (M.I.N.I.). The M.I.N.I. is a clinician-administered structured diagnostic interview that assesses whether or not individuals meet criteria for 19 disorders, including alcohol use disorder (AUD) and substance use disorder (SUD) in the past year. Questions are based on the diagnostic criteria of the International Classification of Diseases (ICD-10; World Health Organization, 1992) and the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5; American Psychiatric Association, 2013). The M.I.N.I. Alcohol Use and Substance Use modules have shown good inter-rater and test-retest reliability (.88 < κ < .94 and .85 < κ < .96 respectively; Sheehan et al., 1998).

HIV viral load.

This was measured via blood specimens, which were collected from participants at the baseline visit or taken from medical records for participants who had their HIV blood work done within one month of the baseline assessment.

ART Adherence.

This was measured via self-report and the Wisepill device. Self-report adherence was assessed by a brief measure of participants’ adherence in the past week, past two weeks, and past four weeks (Lu et al., 2008). Wisepill adherence in the past week and past two weeks was measured using the Wisepill electronic medication adherence device. The Wisepill device holds up to 30 large pills or 60 small pills and data are sent to the research team via a web-based server in real time each time the device is opened (Haberer et al., 2010). Participants were asked to use the Wisepill for one HIV medication that they used most frequently or found most difficult to take. Participants’ Wisepill adherence was captured for two weeks between baseline visits one and two.

Analyses

Statistical analyses were conducted using SPSS version 24. Hierarchical linear regressions were conducted to test for associations between AUD, SUD, PSS, viral load, self-report adherence, and Wisepill adherence. Covariates of age and education were entered in block 1, predictors of substance use and/or social support were entered in block 2, interaction terms were entered in block 3, and viral load linear, viral load log, or ART adherence were entered as outcome variables. All analyses included both log and linear transformed viral load data in order to facilitate meaningful interpretations of this key outcome variable given the limited interpretability of log transformed data (see Feng et al., 2014 for an exposition).

Results

One hundred and nineteen BWLWH in an urban city in the Southeastern United States participated in this study. The average age among participants was 49.27 (range=22–67), 34.5% had completed at least a high school level education, 36.1% earned less than $5000 per year, and 58% were on Disability (see Table 1 for demographic characteristics).

Perceived Social Support, Viral Load, and Medication Adherence

Higher PSS from friends (β=−.263, t=−2.728, p=.008) and significant others (β=−.219, t=−2.270, p=.025) predicted lower viral load log. Of note, there was a trend such that higher PSS from family predicted lower viral load log (β=−.191, t=−1.897, p=.061). However, total PSS did not significantly predict viral load log. In addition, higher PSS from significant others predicted lower viral load linear (β=−.200, t=−2.025, p=.046). PSS total, from friends and from family did not significantly predict viral load linear. In terms of Wisepill and self-report adherence, there were three trends, as both PSS from friends (β=.162, t=1.717, p=.089) and from significant others (β=.170, t=1.828, p=.070) predicted self-report adherence in the past week and PSS from friends predicted self-report adherence in the past four weeks (β=.168, t=1.786, p=.077). However, PSS total did not significantly predict self-report or Wisepill adherence (See Table 2).

Table 2.

Main effects of Perceived Social Support on HIV-related Health Outcomes

Predictor	Outcome	B	Standard Error	Beta	t	p
PSS - Total	Viral Load Log	−.006	.196	−.003	−.028	.977
PSS from Friends	Viral Load Log	−.392	.144	−.263	−2.728	.008
PSS from Significant Others	Viral Load Log	−.325	.143	−.219	−2.270	.025
PSS from Family	Viral Load Log	−.287	.151	−.191	−1.897	.061
PSS - Total	Viral Load Linear	51.932	2568.153	.002	.020	.984
PSS from Friends	Viral Load Linear			−.161	−1.586	.116
PSS from Significant Others	Viral Load Linear	−.3851.663	1901.971	−.200	−2.025	.046
PSS from Family	Viral Load Linear	−2795.351	2016.821	−.145	−1.386	.169
PSS - Total	Wisepill adherence – 2 weeks	−1.384	1.830	−.072	−.756	.451
PSS from Friends	Wisepill adherence – 2 weeks	1.104	1.852	.056	.596	.552
PSS from Significant Others	Wisepill adherence – 2 weeks	.418	1.838	.021	.228	.820
PSS from Family	Wisepill adherence – 2 weeks	−.449	1.901	−.023	−.236	.814
PSS - Total	Wisepill adherence – 1 week	−1.270	1.778	−.068	−.714	.476
PSS from Friends	Wisepill adherence – 1 week	1.577	1.796	.083	.878	.382
PSS from Significant Others	Wisepill adherence – 1 week	.064	1.787	.003	.036	.972
PSS from Family	Wisepill adherence – 1 week	−.380	1.847	−.020	−.206	.837
PSS - Total	Self-report adherence – 4 weeks	−.190	.115	−.157	−1.660	.100
PSS from Friends	Self-report adherence – 4 weeks	.202	.113	.168	1.786	.077
PSS from Significant Others	Self-report adherence – 4 weeks	.151	.113	.125	1.341	.183
PSS from Family	Self-report adherence – 4 weeks	.097	.117	.081	.834	.406
PSS - Total	Self-report adherence – 2 weeks	−.137	.109	−.121	−1.261	.210
PSS from Friends	Self-report adherence – 2 weeks	.128	.108	.113	1.181	.240
PSS from Significant Others	Self-report adherence – 2 weeks	.075	.107	.066	.702	.484
PSS from Family	Self-report adherence – 2 weeks	−.002	.111	−.002	−.017	.986
PSS - Total	Self-report adherence – 1 week	−.085	.110	−.074	−.779	.437
PSS from Friends	Self-report adherence – 1 week	.186	.108	.162	1.717	.089
PSS from Significant Others	Self-report adherence – 1 week	.195	.106	.170	1.828	.070
PSS from Family	Self-report adherence – 1 week	.103	.112	.089	.923	.358

Substance Use, Viral Load and Medication Adherence

A diagnosis of SUD significantly predicted self-report adherence in the past four weeks (β=.213, t=2.345, p= .021). However, SUD and AUD did not predict self-report adherence in the past two weeks or past week. Also, SUD and AUD did not predict Wisepill adherence or viral load (See Table 3).

Table 3.

Main effects of Substance Use and HIV-related Health Outcomes

Predictor	Outcome	B	Standard Error	Beta	t	p
Substance use (12 months)	Viral Load Linear	38.046	2031.890	.002	.019	.985
Alcohol use (12 months)	Viral Load Linear	−1160.234	1395.489	−.083	−.831	.408
Substance use (12 months)	Viral Load Log	.089	.152	.057	.585	.560
Alcohol use (12 months)	Viral Load Log	−.086	.146	−.057	−.591	.556
Substance use (12 months)	Wisepill adherence – 2 weeks	−2.885	1.782	−.149	−1.619	.108
Alcohol use (12 months)	Wisepill adherence – 2 weeks	1.786	1.783	.092	1.001	.319
Substance use (12 months)	Wisepill adherence – 1 week	−2.764	1.732	−.147	−1.596	.113
Alcohol use (12 months)	Wisepill adherence – 1 week	1.660	1.733	.088	.958	.340
Substance use (12 months)	Self-report adherence – 4 weeks	.260	.111	.213	2.345	.021
Alcohol use (12 months)	Self-report adherence – 4 weeks	.128	.112	.106	1.146	.254
Substance use (12 months)	Self-report adherence – 2 weeks	.143	.106	.125	1.345	.181
Alcohol use (12 months)	Self-report adherence – 2 weeks	.010	.106	.009	.093	.926
Substance use (12 months)	Self-report adherence – 1 week	.126	.107	.109	1.184	.239
Alcohol use (12 months)	Self-report adherence – 1 week	.068	.107	059	.636	.526

Perceived Social Support and Substance Use

The effects of PSS on substance use were also examined via multivariable linear regressions, controlling for age and education. Total PSS as well as PSS from friends, significant others and family did not predict SUD or AUD (See Table 4).

Table 4.

Main effects of Perceived Social Support on Substance Use

Predictor	Outcome	B	Standard Error	Beta	t	p
PSS - Total	Substance use	−.042	.041	−.098	−1.007	.316
PSS from Friends	Substance use	.002	.041	.005	.056	.955
PSS from Significant Others	Substance use	.057	.040	.135	1.435	.154
PSS from Family	Substance use	.020	.042	.047	.477	.634
PSS - Total	Alcohol use	−.051	.031	−.159	−1.648	.102
PSS from Friends	Alcohol use	−.015	.031	−.046	−.472	.638
PSS from Significant Others	Alcohol use	.006	.031	.017	.180	.857
PSS from Family	Alcohol use	−.020	.032	−.060	−.610	.543

Perceived Social Support and Substance Use Interactions Predicting Outcomes

Multivariable linear regressions were conducted to examine the effects of the interaction between PSS and SUD on viral load and ART adherence. The Total PSS x SUD interaction was significantly related to viral load linear (β=.945, t=2.060, p=.042). The Total PSS x AUD interaction was also significantly related to viral load linear (β=1.020, t=2.042, p=.044). However, the interaction between specific types of PSS (e.g. from significant others, family and friends) and AUD as well as SUD did not significantly predict viral load linear.

Similarly, Total PSS x SUD significantly predicted viral load log (β=1.241, t=2.851, p=.005). Total PSS x AUD also significantly predicted viral load log (β=1.309, t=2.771, p=.007). PSS from friends x SUD significantly predicted Wisepill adherence in the past week (β=.312, t=3.527, p=.001) and past two weeks (β=.298, t=3.347, p=.001). However, the interactions between PSS from significant others, family and friends and AUD, and the interactions between these types of PSS and SUD did not significantly predict viral load log. The interactions of PSS from significant others and family with SUD and the interactions of PSS from significant others and family with AUD did not significantly predict Wisepill adherence in the past week or two weeks. Furthermore, the interactions between all four types of PSS and AUD and SUD did not significantly predict self-report adherence in the past week, two weeks or four weeks (See Table 5).

Table 5.

Results of Linear Regression Analyses assessing Interactions between Perceived Social Support and Substance Use and HIV-related Health Outcomes

Interaction term	Outcome variable	B	Standard error	Beta	t	p
Total PSS x Substance use	Viral Load Linear	13739.538	6668.327	.945	2.060	.042
PSS SO x Substance use		1023.165	2450.536	.045	.418	.677
PSS Fam x Substance use		2890.662	2352.513	.134	1.229	.222
PSS Friends x Substance use		−1887.460	1938.042	−.098	−.974	.333
Total PSS x Alcohol use		6975.652	3416.313	1.020	2.042	.044
PSS SO x Alcohol use		714.700	1537.050	.047	.465	.643
PSS Fam x Alcohol use		186.183	1620.900	.012	.115	.909
PSS Friends x Alcohol use		1040.999	1519.320	.070	.685	.495
Total PSS x Substance use	Viral Load Log	1.416	.497	1.241	2.851	.005
PSS SO x Substance use		.136	.175	.079	.439	−.211
PSS Fam x Substance use		.054	.170	.033	.319	.750
PSS Friends x Substance use		−.022	.143	−.015	−.152	.880
Total PSS x Alcohol use		.969	.350	1.309	2.771	.007
PSS SO x Alcohol use		.040	.160	.025	.250	.803
PSS Fam x Alcohol use		.082	.169	.048	.489	.626
PSS Friends x Alcohol use		.051	.160	.031	.321	.749
Total PSS x Substance use	Self-report Adherence – Past 4 weeks	.000	.096	.000	−.002	.999
PSS SO x Substance use		.107	.137	.077	.783	.435
PSS Fam x Substance use		.013	.118	.010	.107	.915
PSS Friends x Substance use		.089	.111	.073	.797	.427
Total PSS x Alcohol use		−.020	.074	−.032	−.272	.786
PSS SO x Alcohol use		.044	.123	.034	.359	.720
PSS Fam x Alcohol use		.082	.120	.065	.683	.496
PSS Friends x Alcohol use		−.039	.119	−.032	−.333	.740
Total PSS x Substance use	Self-report Adherence – Past 2 weeks	−.029	.092	−.033	−.315	.753
PSS SO x Substance use		.176	.131	.135	1.343	.182
PSS Fam x Substance use		−.083	.114	−.071	−.728	.468
PSS Friends x Substance use		.033	.108	.029	.303	.763
Total PSS x Alcohol use		−.041	.071	−.070	−585	.560
PSS SO x Alcohol use		.044	.118	.036	.370	.712
PSS Fam x Alcohol use		.088	.115	.074	.767	.445
PSS Friends x Alcohol use		−.073	.114	−.062	−.639	.524
Total PSS x Substance use	Self-report Adherence – Past week	−.045	.093	−.052	−.488	.626
PSS SO x Substance use		.160	.131	.121	1.227	.222
PSS Fam x Substance use		−.209	.113	−.176	−1.851	.067
PSS Friends x Substance use		−.036	.108	−.032	−.336	.738
Total PSS x Alcohol use		−.042	.071	−.070	−.590	.556
PSS SO x Alcohol use		−.004	.117	−.003	−.033	.974
PSS Fam x Alcohol use		.014	.116	.011	.119	.905
PSS Friends x Alcohol use		−.059	.114	−.050	−.518	.606
Total PSS x Substance use	Wisepill Adherence – Past 2 weeks	.603	1.545	.041	.390	.697
PSS SO x Substance use		3.135	2.205	.138	1.421	.158
PSS Fam x Substance use		2.463	1.909	.122	1.290	.200
PSS Friends x Substance use		5.846	1.747	.298	3.347	.001
Total PSS x Alcohol use		.754	1.191	.074	.633	.528
PSS SO x Alcohol use		.175	2.008	.008	.087	.931
PSS Fam x Alcohol use		−.015	1.947	−.001	−.008	.994
PSS Friends x Alcohol use		.802	1.948	.039	.412	.681
Total PSS x Substance use	Wisepill Adherence – Past week	.567	1.502	.040	.377	.707
PSS SO x Substance use		2.171	2.155	.099	1.007	.316
PSS Fam x Substance use		2.195	1.858	.112	1.181	.240
PSS Friends x Substance use		5.946	1.686	.312	3.527	.001
Total PSS x Alcohol use		.701	1.158	.071	.606	.546
PSS SO x Alcohol use		.015	1.952	.001	.008	.994
PSS Fam x Alcohol use		−.320	1.892	−.016	−.169	.866
PSS Friends x Alcohol use		.585	1.891	.029	.309	.758

Perceived Social Support as a moderator in the relationship between Substance Use and HIV Health Outcomes

To tease apart significant interactions noted above, moderational analyses were conducted via multivariable hierarchical linear regressions to examine the potential differences between women with high or low PSS in the relationships between substance use and outcomes of adherence and viral load. In accordance with guidelines suggested by Holmbeck, scores that fell one standard deviation above and below the mean for PSS were used to group women into high and low PSS categories respectively (Holmbeck, 2002; Holmbeck, 1997). Results showed that SUD significantly predicted low Wisepill adherence in the past week (β=−.651. t=−3.478, p=.003) and past two weeks (β=−.690, t=−3.933, p=.001) only for women low in PSS from friends. Among women who reported high PSS from friends, these relationships were not significant (see Table 6).

Table 6.

Results of regression analyses using high and low perceived social support (derived from standard deviation) as predictors

Substance use	Social Support	Outcome variable	Beta	t	p
Substance Use disorder	Low PSS from Friends	Wisepill Adherence – Past Week	−.651	−3.478	.003
Substance Use disorder	High PSS from Friends	Wisepill Adherence – Past Week	−.064	−.261	.798
Substance Use disorder	Low PSS from Friends	Wisepill Adherence – Past 2 Weeks	−.690	−3.933	.001
Substance Use disorder	High PSS from Friends	Wisepill Adherence – Past 2 Weeks	−.065	−.270	.791

In addition, the median split method was used to generate high and low PSS scores to better preserve the sample size, which was significantly reduced when utilizing the standard deviation. Results indicated that SUD significantly predicted low Wisepill adherence in the past week (β=−.273, t=−2.167, p=.035) and past two weeks (β=−.273, t=−2.190; p=.033) only for women low in PSS from friends. SUD also significantly predicted low Wisepill adherence in the past two weeks only for women who were low in total PSS (β=−.294, t=−2.328, p=.024). Finally, SUD was a significant predictor of more frequent Wisepill adherence in the past four weeks only for women who were high in total PSS (β=.360, t=2.833, p=.006; see Table 7).

Table 7.

Results of Linear Regression Analysis between Substance use, High and Low Social Support (derived from the Median) and HIV-related Health Outcomes

Substance use variable	Moderator	Outcome variable	Beta	t	p
Substance use Disorder	Low PSS from Friends	Wisepill Adherence - Past Week	−.273	−2.167	.035
Substance use Disorder	High PSS from Friends	Wisepill Adherence - Past Week	−.024	−.178	.859
Substance use Disorder	Low PSS Total	Wisepill Adherence - Past 2 Weeks	−.294	−2.328	.024
Substance use Disorder	High PSS Total	Wisepill Adherence - Past 2 Weeks	−.036	−.264	.792
Substance use Disorder	Low PSS from Friends	Wisepill Adherence - Past 2 Weeks	−.273	−2.190	.033
Substance use Disorder	High PSS from Friends	Wisepill Adherence - Past 2 Weeks	−.029	−212	.833
Substance use Disorder	High PSS Total	Self-report Adherence - Past 4 Weeks	.360	2.833	.006
Substance use Disorder	Low PSS Total	Self-report Adherence - Past 4 Weeks	.060	.456	.650

Discussion

Findings from this study are the first, to our knowledge, that has pointed to the moderating role of PSS in the relationship between SUD/AUD and adherence and viral load among BWLWH. Results suggest that PSS from friends and significant others affects viral load and a diagnosis of SUD affects self-report adherence in the past four weeks. Total PSS x SUD and Total PSS x AUD significantly predicted higher viral load and PSS from friends x SUD significantly predicted higher Wisepill adherence in the past week and 2 weeks. This indicates that the effects of substance use on VL and adherence may vary depending on social support. Additional findings showed that SUD significantly predicted low Wisepill adherence in the past week and in the past two weeks only for women low in PSS from friends. Similarly, a diagnosis of SUD also significantly predicted low Wisepill adherence in the past two weeks only for women who were low in total PSS. Ironically, a diagnosis of SUD was a significant predictor of more frequent Wisepill adherence in the past four weeks only for women who were high in total PSS. In addition to the novelty of the moderations, these findings are consistent with existing literature, which has shown a) relationships between PSS and HIV health outcomes and b) that social support offers increased accountability for adherence, which in turn has implications for viral suppression (Dale & Safren, 2018; Burgoyne, 2005; Cox, 2002; Edwards, 2006; Friedman et al., 2017). Furthermore, the current study broadened the scope of research by assessing the use of multiple substances (recall that previous work has focused on cocaine use; Gonzalez et al., 2011).

Substance use may have been related to viral load among women with low PSS from friends because of the protective role of social support from friends (e.g. in the context of support groups or programs such as Alcoholics Anonymous or Narcotics Anonymous) in maintaining health enhancing behaviors such as medication adherence and attending medical appointments, both of which directly predict viral load. The moderating role of total PSS also makes conceptual sense because, perhaps in the context of having high social support across multiple relationships (e.g. friends and family), BWLWH who struggled with SUD were still able to adhere to their HIV medication. Unlike the current study, previous research may have found significant effects of injection drug use on viral load due to the direct biological link between injection drug use and virologic outcomes (National Institute on Drug Abuse, 2019) as well as the pathways via which the effects of intravenous drug use may affect cognitive functioning and social factors that, in turn, reduce ART adherence (Davis et al., 2018; Robinson & Knowlton, 2016). Taken together, our results highlight the importance of social support in relation to ART adherence and viral load among BWLWH in the context of substance use.

The findings of this study should be considered in light of its limitations. First, the use of self-report measures of medication adherence, substance use and PSS may have resulted in social desirability and recall bias. Second, our study utilized a convenience sample and enrolled BWLWH in the Southeastern U.S, which may lead to sampling bias and limit the generalizability of findings. Third, the cross-sectional design of this study does not allow for inferences about causation among variables.

In sum, the present study offers important insights into two key aspects of psychosocial functioning - social support and substance use - that affect BWLWH, especially those with histories of trauma/abuse. More specifically, it highlights the nuanced ways in which sources of social support (especially from friends and significant others) may differentially relate to substance use, medication adherence and viral load for BWLWH. These findings may inform the development of interventions for BWLWH, which may include content that explores the role of social support networks and ways to compensate for lack of support in the context of substance use. Such interventions may not only help to improve quality of life, but also improve HIV medication adherence and viral load among BWLWH.