Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Jun 1.
Published in final edited form as: AIDS Behav. 2022 Oct 30;27(6):1741–1756. doi: 10.1007/s10461-022-03906-z

Intersecting Relationships of Psychosocial and Structural Syndemic Problems Among People with HIV in South Africa: Using Network Analysis to Identify Influential Problems

Jasper S Lee 1,2, Sierra A Bainter 3, Alexander C Tsai 4,5, Lena S Andersen 6, Amelia M Stanton 7, Jessica F Magidson 8, Ashraf Kagee 9, John A Joska 10, Conall O’Cleirigh 1,2, Steven A Safren 3
PMCID: PMC10148921  NIHMSID: NIHMS1885064  PMID: 36309936

Abstract

In South Africa, little is known about interrelationships between syndemic problems among people with HIV (PWH). A better understanding of syndemic problems may yield important information regarding factors amenable to mitigation. We surveyed 194 PWH in Khayelitsha, outside of Cape Town, South Africa. We used network analysis to examine the frequency of 10 syndemic problems and their interrelationships. Syndemic problems among PWH in South Africa were common; 159 (82.8%) participants reported at least 2 co-occurring syndemic problems and 90 (46.9%) endorsed 4 or more. Network analysis revealed seven statistically significant associations. The most central problems were depression, substance use, and food insecurity. Three clusters of syndemics were identified: mood and violence; structural factors; and behavioral factors. Depression, substance use, and food insecurity commonly co-occur among PWH in sub-Saharan Africa and interfere with HIV outcomes. Network analysis can identify intervention targets to potentially improve HIV treatment outcomes.

Keywords: Syndemics, HIV, Network analysis, Global Mental Health

Introduction

South Africa has the largest HIV-related disease burden of any country in the world, with approximately 7.7 million people with HIV (PWH), representing 20% of all PWH globally [1]. In addition to its large HIV burden, South Africa also evidences high rates of psychosocial and structural problems in general, and among PWH these problems greatly affect HIV outcomes. In South Africa, psychosocial problems (e.g., depression, post-traumatic stress, and alcohol and substance use) and structural problems (e.g., food insecurity, poverty, housing instability) are more prevalent among PWH than among those without HIV, and these problems have been shown to interfere with HIV outcomes [2–4]. For example, in South Africa specifically, depression, post-traumatic stress disorder (PTSD), and alcohol and other substance use are associated with decreased engagement in HIV care, poorer ART adherence, higher viral load over time, and lower CD4 cell counts [2, 5–13]. Similarly, food insecurity [3, 14–18], poverty [3, 14], housing instability [19], and other structural challenges [3, 20] have been associated with worse HIV outcomes. Poverty, for example, is a prevalent barrier to HIV care [3] and ART adherence [9, 21], and associated food insecurity is related to reduced ART adherence [22, 23].

Although each of these psychosocial and structural problems are independently associated with worse HIV outcomes, these problems rarely occur in isolation [4, 24]. Rather, they co-occur in a dynamic and complex interplay that is understudied among South African PWH. As such, it is not yet known how experiencing multiple psychosocial comorbidities and structural challenges affects HIV outcomes in this context.

One helpful framework for conceptualizing the interrelationships between psychosocial and structural problems is syndemic theory, which posits that psychosocial and structural problems interact synergistically to confer an increased burden of disease [25, 26]. This theory suggests that the individual psychosocial and structural problems that comprise an overall syndemic reinforce one another, causing poorer HIV outcomes (e.g., viral load, CD4, ART adherence, and treatment retention) than would be expected when considering the individual effects of each condition separately on HIV outcomes [25]. In addition to creating worse HIV outcomes, the interaction of syndemic problems with each other may decrease the effectiveness of interventions to support HIV treatment over time because their reinforcing effects may maintain the effects of a given disorder on HIV outcomes after intervention [27]. In other words, although there are many efficacious interventions to address individual psychosocial and/or structural factors among PWH, the mutually enhancing effects of syndemic problems may undermine the effectiveness of individual interventions. As such, more knowledge about the interrelationships among comorbid psychosocial and structural problems is needed, and the syndemic framework can help guide intervention development to create effective interventions that can address the most salient problem(s) related to poorer HIV outcomes.

Limited research has drawn on syndemic theory to study problems among PWH in sub-Saharan Africa generally, or South Africa specifically. The extant studies that have investigated the effects of syndemic conditions on HIV outcomes have mostly explored psychosocial problems among PWH, such as depression, alcohol use, substance use, CSA, IPV, PTSD, and anxiety, and have yet to examine a syndemic network comprised of psychosocial and structural problems [28–39]. Further research is needed to examine how psychosocial and structural conditions are associated with each other.

Network analysis is a theoretically motivated statistical approach that is capable of examining synergistic interrelationships and therefore may be useful in measuring associations between syndemic problems and HIV outcomes [40–42]. Network analysis has previously been used to examine the relationships among symptoms of a disorder [43–46], such as to better understand which symptoms of depression [47], substance use [46], or post-traumatic stress [48, 49] appear to be most important within each psychological problem. Recently, network analysis has also been employed to examine bidirectional correlations among psychosocial and structural problems that comprise a syndemic [40–42]. These studies examining syndemic problems using network analysis have done so in samples of 200 participants or less [40–42]. Although small in number, these studies provide initial evidence that this approach is a valid and useful first step in examining interrelationships among syndemic problems.

The current study sought to describe the frequency of psychosocial and structural syndemic problems and explore their interrelationships in PWH in the South African setting. We utilized network analysis to examine the interrelationships of syndemic problems reported.

Method

Participants and Procedures

Participants were recruited from six HIV clinics allocated by the City of Cape Town Health, all in the Khayelitsha periurban settlement outside of Cape Town.

Inclusion/Exclusion Criteria

Individuals with HIV who received HIV care from one of these six clinics in Khayelitsha were considered eligible. Those who were unable or unwilling to provide informed consent, unable to communicate in English or isiXhosa, and/or were less than 18 years of age were not eligible to participate. The sample utilized in the present study was a convenience sample of PWH attending clinics for HIV care. Inclusion criteria were designed to be broad to enhance the epidemiological aim of the present study.

Recruitment and Enrollment Procedures

Study staff approached potentially eligible participants enrolled in HIV care at their primary care clinic to assess for interest in participation and screening. PWH waiting for HIV care at their primary care clinics were informed of the study by research assistants. To avoid interfering with care, after completing their HIV care visit, interested individuals were taken to a private area to be assessed for eligibility. Eligible individuals completed informed consent procedures to participate in a cross-sectional survey and to complete participation procedures. Eligible individuals who completed the informed consent process were interviewed by study staff and provided self-report data. All study procedures were approved by the University of Miami institutional review board and the University of Cape Town ethics committee.

Assessment Procedures and Timing

The assessment was completed on the same day of the study participant’s HIV care visit to decrease the burden associated with study procedures. Data collection instruments were interviewer-administered. All de-identified data were collected via, and stored in, Research Electronic Data Capture (REDCap) software [50, 51].

Measures

We assessed 10 self-reported psychosocial and structural syndemic problems. All measures were translated to isiXhosa, and back-translated to English for review [52]. Participants indicated their responses in either isiXhosa or English.

Post-traumatic stress was assessed via a 4-item diagnostic screener that measures startle, physiological arousal, anxiety, and numbness since the criterion event occurred (SPAN) [53]. A score of 5 was used as a cutoff, such that a score of 5 or greater indicates clinically significant post-traumatic stress symptoms.

Depression was assessed via the 20-item Center for Epidemiologic Studies Depression Scale (CES-D) [54], which has previously been translated to isiXhosa and used successfully to measure depression in South African PWH [55]. A cutoff score of 16 or greater is suggestive of clinically significant depressive symptoms.

Social anxiety symptoms in the past week were assessed using the 3-item Social Phobia Inventory (Mini-SPIN) [56]. A score of 6 or greater indicated the presence of clinically significant symptoms of social anxiety.

Substance use was assessed via the Alcohol, Smoking, and Substance Involvement Screening Test (WHO-ASSIST) [57], which queried use of marijuana, cocaine, amphetamine-type stimulants, inhalants, sedatives, hallucinogens, opioids, and other substances in the past three months. To account for polysubstance use, a total score for each type of substance was calculated and summed together. Typically, a score of 3 or less indicates lower risk, 4–26 indicates moderate risk, and 27 or more suggests high risk [58]. However, given that few participants indicated problematic substance use (i.e., a total score of 4 or greater), substance use was dichotomized into low risk (0) for those with a total score of 3 or less, and moderate-to-high risk (1) for those with a total score of 4 or greater.

Alcohol use was assessed using the 10-item Alcohol Use Disorders Identification Test (AUDIT) [59]. A cut-off score of 8 was used, such that scores of 8 or higher indicate clinically significant alcohol use symptoms [60].

Lifetime intimate partner violence (IPV) was measured with a single item eliciting whether or not participants had ever experienced physical or sexual violence from an intimate partner after their 17th birthday.

Food insecurity in the prior 30 days was evaluated via the 9-item Household Food Insecurity Access Scale (HFIAS) [61]. In addition, categories of food insecurity were also calculated based on the frequency with which an individual endorsed experiencing certain items in the past 30 days. The four categories are food secure, and mildly, moderately, and severely food insecure. Participants were considered either food secure (0) or food insecure (1), such that individuals categorized as mildly, moderately, or severely food insecure were considered as experiencing food insecurity.

Housing instability was assessed with two items eliciting the number of times participants moved in the past 12 months, as well as the number of times they expected to move in the next 12 months. Participants who indicated they had moved or expected to move 2 or more times were considered as experiencing housing instability (1) and individuals who had moved or expected to move less than 2 times were not (0).

Structural barriers to HIV care were assessed via the structural barriers to care scales (SBS) [62]. The Structural Barriers to Clinic Attendance (SBCA) scale consists of 12 items that assess structural barriers to attending clinic appointments. The Structural Barriers to Medication Taking (SBMT) consists of 13 items that measure structural barriers to medication use in the past six months. Participants provided responses on 5-point Likert-type scales from 1 (never) to 5 (always) for both scales Total scores for the SBCA and the SBMT were calculated by summing their respective items, then a total score that combined the SBCA and SBMT was calculated. Items in the SBCA measure query constructs such as clinic distance, transportation cost, travel time, as well as potential for unintended disclosure of serostatus, negative interactions with staff, and lack of privacy in the clinic (e.g., “I do not attend my clinic appointments because there is no privacy at the clinic when I meet with the nurse”). The SBMT measures constructs including not having enough food and potential side effects of ART, forgetting, lack of reminder system, preference for traditional healing or religion, interference of alcohol, secondary gain because of governmental grants, lack of privacy at home when taking pills, hiding pills from family members and/or employers, and lack of social support (e.g., “I do not take my ART pills because I do not have someone to remind me to do so”).

Poverty was assessed via annual income. The poverty line in South Africa, set by the South African government, is an income of 14,196 Rand (~ $1025) per year (Statistics South Africa, 2018). Individuals who earned an annual income of less than the poverty line and who also did not identify as a student were categorized as experiencing poverty (1), and other individuals were coded as not experiencing poverty (0).

Statistical Analysis Plan

All analyses were conducted using R version 4.0.3 [63]. Prior to the network analyses, data were examined for outliers and assumptions of normality, and descriptive analyses were conducted.

First, we examined polychoric correlations between the variables of interest. Though continuous variables were used when possible, not all variables were continuous. Given the limited range of the housing instability variable (0–5), it was treated as an ordinal variable, whereas IPV, poverty, and substance use were treated as dichotomous variables. Despite being measured continuously, substance use was dichotomized because only a small number of participants (8.2%) endorsed moderate-to-high risk substance use. Depression, post-traumatic stress, social anxiety, food insecurity, structural barriers to care, and alcohol use were analyzed as continuous variables.

Network analysis is presented graphically and provides empirical results (i.e., centrality indices) regarding the amount of connectivity between problems in the network. The network analysis in the present study used a regularized, partial polychoric correlation matrix with the graphical LASSO (glasso) algorithm [64], and an Extended Bayesian Information Criteria (eBIC) hyperparameter of γ = 0.5. The eBIC hyperparameter of γ = 0.5 was chosen because it produces a more specific (i.e., yielding fewer spurious associations) and more sensitive (i.e., identifying more true associations) network than does a smaller hyperparameter [45]. This approach is advantageous in that it does not require latent factors to explain covariance. Rather than utilizing the underlying assumption that elements are measures of a construct, as is the case in a structural equation modeling (SEM) approach, network analysis assumes that elements within a network are aspects of a construct that reinforce and interact with one another [44, 65]. Network analysis accounts for potential confounding and type I error by applying a statistical penalty for multiple comparisons, biasing parameter estimates toward zero and increasing specificity to reduce the likelihood of estimating spurious relationships. Given that the current network was comprised of dichotomous, ordinal, and continuous data, polychoric correlations were estimated [66].

We also examined centrality indices produced in the network analysis. Centrality indices provide empirical measures of the associations between nodes in the network [46, 65]. Additionally, centrality indices can be used to identify the most central, and possibly most important, nodes within the general construct of interest. The three main centrality indices examined in this analysis were betweenness, closeness, and node strength [46, 67]. Nodes that are high in centrality, especially in node strength, may be considered optimal intervention targets, as such nodes are highly influential in that they are considered to be strongly linked to other nodes in the network. To determine whether any nodes were significantly more central than others, a difference test was conducted based on node strength.

We examined network density and the pattern of interrelationships between syndemic problems to determine whether any clusters of conditions emerged and, if so, their degree of interrelatedness. As potential clusters emerged, an empirical test for clustering was employed using the spinglass algorithm [68], which is appropriate for small (i.e., N < 1000) networks [69]. The spinglass algorithm was applied to the network 1,000 times to obtain the median cluster result.

Power Analysis.

Power to detect an effect in the network analysis [45] was estimated using the bootnet package [70] in R [63]. Results of the power analysis suggested that, for the current network (10 nodes), a sample size of N = 175 was required to achieve a correlation greater than 0.8 between the true network and the estimated network, which is needed for a stable network structure.

Results

Descriptive Statistics

Participants were 194 PWH receiving HIV care at primary care clinics in Khayelitsha, South Africa. The sample was predominantly Black (98.5%), isiXhosa-speaking, South African women (83%). The majority of participants identified as heterosexual (96.4%), and the average age of the sample was 41.25 (SD = 9.96) years. Additionally, most participants were unemployed (60.3%), some were seeking employment (22.7%), and fewer were employed part-time (15.5%) or full-time (13.9%). For housing, most participants lived in a shack (56.2%) or their own/family’s home (39.2%), and participants reported that an average of more than two people (M = 2.63, SD = 1.46) slept in the same room at night. Lastly, participants reported an average annual income of R34,908.93 (SD = R34,672.87), which was equivalent to $2323.47 (SD = $2,307.76) at the time the study was conducted. Additional participant characteristics are presented in Table 1.

Table 1.

Participant characteristics

Variable N (%) Variable Mean (SD) [observed range]
Gender Age (in years) 41.25 (9.96)
  Man 32 (16.5) [18–72]
  Woman 161 (83) Annual Income in ZAR 34,908.93 (34,672.87)
  Transgender man 1 (.5) [0–240,000]
Race
  Black 191 (98.5) Annual Income in USD 2,323.47 (2,307.76) [0–15,973.90]
  Coloured* 2 (1)
  Other 1 (.5) Number of people who sleep in the same room at night 2.63 (1.46)
Education [1–11]
  Grade 6 or below 13 (6.7)
  Grade 7 13 (6.7) Number of people who sleep in the same house at night 3.8 (2.33)
  Grade 8 9 (4.7) [1–15]
  Grade 9 21 (10.9)
  Grade 10 27 (14)
  Grade 11 68 (35.2)
  Grade 12 32 (16.6)
  Vocational training 6 (3.1) Variable N (%)
  University 4 (2.1) Antiretroviral Therapy Regimen
Sexual Orientation   First-line 135 (70.3)
  Heterosexual 187 (96.4)   Failed first-line 57 (29.7)
  Bisexual 5 (2.6)
  Gay 1 (0.5)
Employment Status†
  Unemployed 117 (60.3)
  Seeking employment 44 (22.7)
  Part-time employment 30 (15.5)
  Full-time employment 27 (13.9)
  Student 3 (1.5)
  Disabled 3 (1.5)
  Retired 3 (1.5)
  Homemaker 2 (1)
Housing Type
  None 1 (0.5)
  Shack 109 (56.2)
  Wendy house/backyard dwelling 9 (4.6)
  Own/family house 76 (39.2)
Open in a new tab
†

Totals for employment status may not equal 100% as participants could select more than one response option

*

In the race category, the term “Coloured” refers to a racial category during the apartheid era and remains relevant in describing present health disparities in South Africa

Most participants (81.4%) endorsed experiencing food insecurity in the past 30 days (M = 9.35, SD = 6.61). Approximately 8.2% endorsed moderate- or high-risk substance use (not including alcohol use) in the past three months (M = 1.29, SD = 6.22). Over one-third of the sample (35.8%) reported having current, clinically significant alcohol use (M = 7.45, SD = 9.86), approximately one-third (33.2%) endorsed experiencing depression or depressive symptoms in the past week (M = 13.1, SD = 13.31), and nearly one-third (31.1%) endorsed post-traumatic stress (M = 3.56, SD = 4.29). Most participants endorsed experiencing two or more syndemic problems: 39 (20.3%) participants gave responses consistent with 2 syndemic problems, 30 (15.6%) reported 3 problems, 38 (19.8%) reported 4 syndemic problems, 25 (13%) participants reported 5 problems, and 27 (14.1%) reported 6 or more problems. Complete descriptive statistics for the syndemic problems are presented in Table 2.

Table 2.

Syndemic problems descriptive statistics

Variable Mean (SD) [observed range] Above cut-off N (%)
Depression 13.1 (13.31) [0–55] Yes   64 (33.2)
No 129 (66.8)
Substance Use 1.29 (6.22) [0–60] Yes   16 (8.2)
No 178 (91.8)
Alcohol Use 7.45 (9.86) [0–36] Yes   69 (35.8)
No 124 (64.2)
Intimate Partner Violence (IPV) – Yes   58 (29.9)
No 136 (70.1)
Post-Traumatic Stress 3.56 (4.29) [0–16] Yes   60 (31.1)
No 133 (68.9)
Social Anxiety 3.74 (3.33) [0–12] Yes   66 (34)
No 128 (66)
Food Insecurity 9.35 (6.61) [0–27] Yes 158 (81.4)
No   36 (18.6)
Poverty – Yes   57 (29.5)
No 136 (70.5)
Housing Instability 0.4 (0.76) [0–5] Yes   14 (7.3)
No 179 (92.7)
Structural Barriers to Care 3.28 (5.43) [0–37] Yes   88 (46.1)
No 103 (53.9)
Additive syndemic sum score N (%) Binned syndemic sum score N (%)
0 13 (6.8) 0–1 33 (17.2)
1 21 (10.4) 2–3 69 (35.9)
2 39 (20.3) 4–5 63 (32.8)
3 30 (15.6) 6 or more 27 (14.1)
4 38 (19.8)
5 25 (13.0)
6 17 (8.9)
7   7 (3.7)
8   2 (1.0)
9   1 (0.5)
Open in a new tab

Preliminary Correlations

Inspection of the preliminary polychoric correlations (see Table 3) revealed that the psychosocial and structural syndemic problems included in the current study were correlated with one another, though the strengths of the associations varied between syndemic problems. Depression was the most consistently and strongly correlated with other problems. Food insecurity also showed a pattern of statistically significant associations with many other problems.

Table 3.

Polychoric correlation matrix

SUB DEP ALC IPV PTS SOC FIS POV HOU SBS
SUB 1 – – – – – – – – –
DEP − .01 1 – – – – – – – –
ALC .46* .17* 1 – – – – – – –
IPV .01 .32* .23* 1 – – – – – –
PTS .02 .52* .05 .39* 1 – – – – –
SOC .04 .39* .08 .25* .40* 1 – – – –
FIS .36* .34* .11 .20* .25* .08 1 – – –
POV .20 .15 − .02 − .02 .05 .02 .42* 1 – –
HOU .24 .17* .01 .06 .05 .03 .14 .32* 1 –
SBS .43* .26* .36* .22* .14 .13 .25* .003 .14 1
Open in a new tab

Full variable names are: substance use (SUB), depression (DEP), alcohol use (ALC), intimate partner violence (IPV), post-traumatic stress (PTS), social anxiety (SOC), food insecurity (FIS), poverty (POV), housing instability (HOU), and structural barriers to care (SBS)

*

p < .05

Network Analysis

The findings of the network analysis are presented graphically in Fig. 1, with edge weight parameter estimates presented in Fig. 2, measures of centrality in Fig. 3, and differences in strength centrality by node in Fig. 4. Of the 194 participants in the sample, 189 had no missing data on any syndemic problems and were therefore included in the network analysis. Overall, the results of the network analysis revealed 23 non-zero network edges (i.e., associations between nodes) estimated, out of a total possible 45. Visual inspection of these 23 edges (Fig. 1) suggested potential clusters among psychosocial and structural syndemic problems, given that nearly one-third of nodes (3 of 10) in the network had fewer than four non-zero edges. Additionally, the constellation of associations was such that three distinct triangular groupings of syndemic problems were present, suggesting that a cluster analysis may provide additional information.

Fig. 1.

Fig. 1

Open in a new tab

Network analysis of psychosocial and structural syndemic problems. Note. Circles, called nodes, represent psychosocial and structural syndemic problems. Lines, called edges, represent associations. Blue edges are positive associations, and red edges are negative associations between nodes. The thicker and darker an edge, the stronger the association between the nodes, compared to thinner and lighter edges. The * symbol indicates a significant pairwise association between nodes, such that the bootstrapped 95% confidence interval of the association did not contain zero. The size of the node corresponds with the Strength centrality measure for that node, such that larger node size indicated greater centrality within the network. Additionally, the background color of the nodes represents their cluster affiliation, such that nodes with the same background color are in the same cluster. The green background indicates affiliation with the mood- and violence-related syndemic problems cluster. The orange background indicates affiliation with the substance use, alcohol use, and structural barriers to care syndemic problems cluster. The yellow background indicates affiliation with the structural syndemic problems cluster. Full node names are: substance use (SUB), depression (DEP), alcohol use (ALC), intimate partner violence (IPV), post-traumatic stress (PTS), social anxiety (SOC), food insecurity (FIS), poverty (POV), housing instability (HOU), and structural barriers to care (SBS)

Fig. 2.

Fig. 2

Open in a new tab

Edge-weight parameter estimates, bootstrapped estimates, and 95% confidence intervals of edges in the network. Note. The red dots represent the parameter estimates of edge weights in the sample. Black dots represented the mean bootstrapped estimate of edge weights. The gray areas represent the bootstrapped 95% confidence intervals for the edge weights. Full node names are: substance use (SUB), depression (DEP), alcohol use (ALC), intimate partner violence (IPV), post-traumatic stress (PTS), social anxiety (SOC), food insecurity (FIS), poverty (POV), housing instability (HOU), and structural barriers to care (SBS)

Fig. 3.

Fig. 3

Open in a new tab

Centrality measures by node. Note. Centrality estimates are presented here for strength, betweenness, and closeness, respectively. The x-axis presents z-scores. Full node names are: substance use (SUB), depression (DEP), alcohol use (ALC), intimate partner violence (IPV), post-traumatic stress (PTS), social anxiety (SOC), food insecurity (FIS), poverty (POV), housing instability (HOU), and structural barriers to care (SBS)

Fig. 4.

Fig. 4

Open in a new tab

Strength centrality difference plot. Note. Strength centrality sample estimates are presented along the diagonal in white squares. Black squares represent significant differences in strength centrality, such that one node is significantly more central than the other among the two to which the square corresponds. In the present network analysis, substance use was significantly more central than housing instability and social anxiety, and depression was significantly more central than social anxiety. Full node names are: substance use (SUB), depression (DEP), alcohol use (ALC), intimate partner violence (IPV), post-traumatic stress (PTS), social anxiety (SOC), food insecurity (FIS), poverty (POV), housing instability (HOU), and structural barriers to care (SBS)

Results of the spinglass algorithm examining the presence of clusters of syndemic problems suggested that depression, post-traumatic stress, social anxiety, and IPV formed one cluster of mood- and violence-related syndemic problems. Another cluster of structural syndemic problems was comprised of food insecurity, poverty, and housing instability. Lastly, substance use, alcohol use, and structural barriers to care grouped together in the substance use and structural barriers to care cluster.

The network analysis revealed seven significant associations between syndemic problems (i.e., the bootstrapped 95% confidence intervals [Fig. 2] of the edge weight parameter estimates did not contain zero): depression—post-traumatic stress (b = 0.33, SD = 0.09, 95%CI[0.16, 0.49]); depression—social anxiety (b = 0.19, SD = 0.08, 95%CI[0.02,0.32]); post-traumatic stress—social anxiety (b = 0.20, SD = 0.07, 95%CI[0.05, 0.34]); post-traumatic stress—IPV (b = 0.22, SD = 0.10, 95%CI[0.03, 0.46]); depression—food insecurity (b = 0.18, SD = 0.11, 95%CI[0.04, 0.53]); food insecurity—poverty (b = 0.30, SD = 0.12, 95%CI[0.01, 0.51]); and substance use—alcohol use (b = 0.30, SD = 0.16, 95%CI[0.09, 0.75]). The significant association between depression and food insecurity also connected the mood- and violence-related syndemic problems cluster to the structural syndemic problems cluster.

We also estimated strength, betweenness, and closeness centrality (Fig. 3). All three centrality indices were strongly correlated with one another, suggesting they provided similar information: strength and closeness (r = 0.85, p = 0.002), strength and betweenness (r = 0.81, p = 0.005), and betweenness and closeness (r = 0.93, p < 0.001). We focused on the strength index as the primary centrality index given that it provides more information on the influence of nodes in the network than do closeness or betweenness [71]. The most central node in the network was depression, followed by substance use, food insecurity, and post-traumatic stress. None of the four most central nodes were significantly more central than each other (Fig. 4). However, depression was significantly more central than social anxiety, and substance use was significantly more central than housing instability and social anxiety.

Discussion

The current study examined the prevalence and interrelationships of 10 psychosocial and structural syndemic problems among PWH receiving HIV care at primary care clinics in Khayelitsha, South Africa. Participants endorsed high rates of frequently co-occurring psychosocial and structural problems. Nearly half (46.9%) experienced four or more syndemic problems—a much higher rate compared with that estimated in a sample of MSM with HIV in Latin America [28]. The higher prevalence rate of numerous syndemic problems among PWH in South Africa compared to MSM with HIV in Latin America underscores the critical public health need to address these synergistic epidemics in South Africa—the country with the largest HIV-related disease burden in the world.

One crucial step toward addressing the effects of syndemic problems on HIV treatment outcomes is to better understand how syndemic problems are related, which can inform studies that would test which syndemic problems appear to confer central risk for poorer HIV outcomes. Although some researchers have advocated for the use of interactions effects approaches [72], it remains impractical to examine interaction effects among many syndemic problems without encountering the multiple comparisons problem [73].

In this study, we used network analysis to examine the correlations between syndemic problems and identify the most central problems that could potentially be targeted in interventions to improve HIV treatment outcomes. While this method has been used to estimate symptom networks [74], it has recently been extended to estimating networks of other types of variables, including psychosocial risk variables [40, 75]. In the network of psychosocial and structural problems among study participants in our sample, the three most central, and therefore potentially most influential, variables were depression, substance use, and food insecurity. This pattern of findings suggests that, when present, depression, substance use, and food insecurity are correlated with each other, as well as with other syndemic problems. If the estimated patterns of associations are causal, one potential implication of our findings is that treating depression may have beneficial spillover effects on other syndemic problems that propagate throughout the network, ultimately improving HIV treatment outcomes. Longitudinal data would be needed to confirm this hypothesis.

Within the mood- and violence-related cluster there were significant positive associations between depression and post-traumatic stress, depression and social anxiety, post-traumatic stress and social anxiety, and post-traumatic stress and IPV. It is not unexpected that depression, post-traumatic stress, and social anxiety were interrelated as several models exist to explain common underlying vulnerabilities [76–81]. One possible underlying mechanism common to depression, post-traumatic stress, and social anxiety is experiential avoidance [82], which could explain this pattern of relationships.

The mood- and violence-related syndemic problems cluster was connected to the structural syndemic problems cluster via a significant positive association between depression and food insecurity. This finding is consistent with other research on the relationship between food insecurity and depression [83], including among PWH in sub-Saharan Africa [84, 85]. One study examined this relationship among men and women with HIV in rural Uganda, and found an association between food insecurity and depression among women, but not among men [85]. This relationship between depression and food insecurity was also replicated among pregnant women in Cape Town, South Africa, who were not selected on the basis of HIV status [86]. In the current study, women comprised 83% of the sample, which may partially explain the relationship. Future research should examine the effects of gender roles on the association between food insecurity and depression. This significant association between food insecurity and depression may also result partially from nutritional deficiencies [87–91]. Improving nutrition via increased access to high-quality and nutritious foods could potentially alleviate some depressive symptoms. Importantly, nutritional status and food insecurity may affect the effectiveness of treatments for depression [87, 88]. The findings of the current study extend the association of food insecurity with depression and demonstrate the interrelationship of food insecurity and depression in a syndemic framework.

Within the structural syndemic problems cluster, food insecurity was significantly associated with poverty. Although one direct consequence of poverty is food insecurity, it is notable that food insecurity rates (81.4%) were much higher than poverty rates (29.5%) in the current sample. This discrepancy indicates that food insecurity cannot be explained solely by experiencing poverty, and, moreover, approaches that mitigate poverty, even at the systemic level, may not be sufficient to address food insecurity. Of all ten syndemic problems assessed, food insecurity had the highest prevalence rate and was one of the most potentially influential (i.e., central) variables in the network, indicating that food security interventions are urgently needed.

The third and final cluster was comprised of substance use, alcohol use, and other structural barriers to care. The only significant association among variables in this cluster was between substance use and alcohol use. Despite being the second most central variable in the network, substance use was endorsed by few (8.2%) participants, and approximately three-quarters of whom also used alcohol (i.e., 6.3% of the total sample engaged in substance and alcohol use concurrently), further indicating that substance use is strongly associated with alcohol use. The third cluster also contained the structural barriers to care variable (a composite score of the SBCA and SBMT), which includes some items that, though rooted in structural problems, measure constructs related to behaviors, stigma, alcohol use, and social support. One could conceptualize the structural barriers to care measure as measuring behavior driven by structural inequality. Because the other two variables in the cluster (substance use and alcohol use) are also behaviors, this cluster seems to reflect behavioral syndemic problems, whereas the structural syndemic problems cluster reflects the experience of systemic problems. Additionally, some items of the SBCA and SBMT relate to stigma and to social support, and individuals who use substances and/or alcohol experience greater levels of stigma [92, 93], and may have worse access to social support than those who do not [94]. Social support is associated with improved ART adherence in South Africa [95] and is thought to assist PWH in overcoming structural barriers to care [96]. For individuals with problematic substance or alcohol use, decreased social support and increased stigma may be related to structural barriers to care.

Demographically, participants in the present study were similar to those in prior research among PWH in Khayelitsha. In a large longitudinal study of virologic response to ART in PWH in Khayelitsha (N = 8058), participants were approximately 74% female, with an average age of 39 years (IQR 34–45) [97]. Approximately 70% of participants in the present study were on first-line ART regimen. Although only approximately 4% of PWH accessing ART were on second-line in 2016 in South Africa [98], the greater number of individuals on second-line ART in the present study may be representative of high psychosocial and structural burden.

The prevalence of psychosocial and structural problems in the present study are generally similar to prevalence estimates among PWH in prior research. Depression has an estimated 14–45% among PWH in South Africa, depending on cut-off scores utilized [4]. Alcohol use and substance use have been estimated to have a prevalence of 32% [99] and 13% [6], respectively, which are similar to the prevalence in the present study. The prevalence of post-traumatic stress has also been estimated at approximately 21%, which is somewhat lower than that of 31% found in the present study. However, the present study found a prevalence of approximately 30% for IPV, which is consistent with previous research suggesting a 23–56% prevalence of IPV among PWH in SA [100, 101]. It is also notable that although the prevalence of food insecurity in the present study (approximately 80%) was generally consistent with that of prior research also suggesting 80% [102], the prevalence of poverty was rather low in the present study. Prior research has suggested that approximately 75% of PWH in Cape Town experience poverty [4], whereas the prevalence of poverty in the present study was approximately 30%. It is possible that sampling biases led to a sample experiencing less poverty than those who are not able to attend primary care clinics for HIV care as frequently. In the current study, participants endorsed low rates of experiencing housing instability (7.3%), which may suggest that in the peri-urban settlement of Khayelitsha, housing may be relatively stable among PWH, even in the context of experiencing poverty. However, the majority of participants (60.8%) indicated living in either a shack or a Wendy house/backyard dwelling. Examining the prevalence of syndemic problems utilized in the network analysis is important as dichotomized variables in the network with low endorsement rates may not behave as expected and differences in prevalence of psychosocial and structural problems endorsed in the present study compared to those found in extant research may result in biases in network estimation.

The current study was not without limitations, and caution is warranted in interpreting these findings. One major limitation was the sampling procedure. This study was partially designed to capture epidemiological data on patients receiving HIV care at public primary care clinics in Khayelitsha. As such, the sample is a convenience sample, and individuals were not recruited for study participation based on difficulties with any aspects of HIV care or wellbeing. It is possible and likely that PWH who attended the clinic more regularly had a higher likelihood of enrollment, compared to individuals with irregular attendance and individuals who are not currently accessing HIV care, due to the effects of syndemic problems and/or other reasons. Therefore, the findings may be affected by selection bias, which may result in an underestimate of the centrality of syndemic problems that interfere with clinic attendance altogether, and results may not be generalizable to other populations experiencing syndemic problems. The use of self-reported measures of syndemic problems is an additional limitation.

Conclusions

The prevalence of syndemic problems in this population of PWH in Khayelitsha, South Africa was high, with food insecurity reported by most participants. The majority of the sample experienced syndemic problems, and approximately half reported experiencing four or more syndemic problems. The results of the network analysis indicate that the most central and potentially influential syndemic problems were depression, substance use, and food insecurity. Each of these syndemic problems was located in different clusters: the mood- and violence-related syndemic problems cluster; the structural syndemic problems cluster; and the substance use, alcohol use, and other structural barriers to care cluster (the behavioral cluster). The centrality indices, as well as the cluster analysis, may inform future analyses that examine interaction effects approaches to assessing syndemic problems and statistically examining the model of syndemic synergy, and may inform future intervention development utilizing a syndemic care approach [103].

Acknowledgements

Funding for this project came from the National Institute of Mental Health F31MH122279 (Lee) and R01MH103770 (Safren, O’Cleirigh). Some additional support was from P30MH116867 (Safren), K24DA040489 (Safren), and Dr. Lee’s time on this manuscript was supported by T32MH116140 (Henderson, Fricchione). Dr. Magidson’s time on this manuscript was supported by K23DA041901 (Magidson). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Mental Health, the National Institute of Drug Abuse, or the National Institutes of Health, or any of the other funders. Data were collected in public primary care clinics in Khayelitsha outside of Cape Town, South Africa. We would like to thank the City of Cape Town Health Department for allowing us access to their clinics and for their ongoing support. We would also like to thank the clinic staff and study participants for their time spent on the project, as well as the study team who worked tirelessly on this project. Special thanks to Nicolas Cardenas for translating the Abstract into Spanish.

Funding

Funding for this project came from the National Institute of Mental Health F31MH122279 (Lee) and R01MH103770 (Safren, O’Cleirigh). Some additional support was from P30MH116867 (Safren), K24DA040489 (Safren), and T32MH116140 (Henderson, Fricchione). Dr. Magidson’s time on this manuscript was supported by K23DA041901 (Magidson). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Mental Health, the National Institute of Drug Abuse, or the National Institutes of Health, or any of the other funders.

Footnotes

Code availability Code is available upon request to the corresponding author.

Ethical approval and consent to participate All study procedures were approved by the University of Miami institutional review board and the University of Cape Town ethics committee. All participants completed the informed consent process prior to participation.

Consent for publication Not applicable.

Competing interests Dr. Safren receives royalties for books published by Oxford University Press, Springer/Humana Press, and Guilford Publications.

Availability of data and materials

Data are available upon request to the corresponding author.

References

  • 1.UNAIDS. South Africa [Internet]. 2018. [cited 2018 Nov 6]. Available from: http://www.unaids.org/en/regionscountries/countries/southafrica
  • 2.Cichowitz C, Maraba N, Hamilton R, Charalambous S, Hoffmann CJ. Depression and alcohol use disorder at antiretroviral therapy initiation led to disengagement from care in South Africa. PLoS ONE. 2017;12(12):e0189820. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Kagee A, Remien RH, Berkman A, Hoffman S, Campos L, Swartz L. Structural barriers to ART adherence in Southern Africa: challenges and potential ways forward. Glob Public Health. 2011;6(1):83–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Myer L, Smit J, Roux LL, Parker S, Stein DJ, Seedat S. Common mental disorders among HIV-infected individuals in South Africa: prevalence, predictors, and validation of brief psychiatric rating scales. AIDS Patient Care STDs. 2008;22(2):147–58. [DOI] [PubMed] [Google Scholar]
  • 5.Dahab M, Charalambous S, Hamilton R, Fielding K, Kielmann K, Churchyard GJ, et al. “That is why I stopped the ART”: patients’ & providers’ perspectives on barriers to and enablers of HIV treatment adherence in a South African workplace programme. BMC Public Health. 2008;8(1):63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Kader R, Seedat S, Govender R, Koch JR, Parry CD. Hazardous and harmful use of alcohol and/or other drugs and health status among south African patients attending HIV clinics. AIDS Behav. 2014;18(3):525–34. [DOI] [PubMed] [Google Scholar]
  • 7.Magidson JF, Saal W, Nel A, Remmert JE, Kagee A. Relationship between depressive symptoms, alcohol use, and antiretroviral therapy adherence among HIV-infected, clinic-attending patients in South Africa. J Health Psychol. 2017;22(11):1426–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Mills EJ, Nachega JB, Bangsberg DR, Singh S, Rachlis B, Wu P, et al. Adherence to HAART: a systematic review of developed and developing nation patient-reported barriers and facilitators. PLoS Med. 2006;3(11): e438. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Naidoo P, Peltzer K, Louw J, Matseke G, Mchunu G, Tutshana B. Predictors of tuberculosis (TB) and antiretroviral (ARV) medication non-adherence in public primary care patients in South Africa: a cross sectional study. BMC Public Health. 2013;13(1):396. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Nel A, Kagee A. Common mental health problems and antiretroviral therapy adherence. AIDS Care. 2011;23(11):1360–5. [DOI] [PubMed] [Google Scholar]
  • 11.Nel A, Kagee A. The relationship between depression, anxiety and medication adherence among patients receiving antiretroviral treatment in South Africa. AIDS Care. 2013;25(8):948–55. [DOI] [PubMed] [Google Scholar]
  • 12.Rose AL, Belus JM, Ma T, Lee JS, Wan C, De Los RA, et al. The relationship between harmful alcohol use and antiretroviral non-adherence in people accessing HIV treatment in Cape Town, South Africa: an event-level analysis. AIDS Behav. 2022;26(6):2055–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Sikkema KJ, Mulawa MI, Robertson C, Watt MH, Ciya N, Stein DJ, et al. Improving AIDS Care After Trauma (ImpACT): pilot outcomes of a coping intervention among HIV-infected women with sexual trauma in South Africa. AIDS Behav. 2018;22(3):1039–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Biadgilign S, Deribew A, Amberbir A, Deribe K. Barriers and facilitators to antiretroviral medication adherence among HIV-infected paediatric patients in Ethiopia: a qualitative study. SAHARA-J J Soc Asp HIVAIDS. 2009;6(4). Available from: https://www.ajol.info/index.php/saharaj/article/view/50639 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Himmelgreen DA, Romero-Daza N, Turkon D, Watson S, Okello-Uma I, Sellen D. Addressing the HIV/AIDS—food insecurity syndemic in sub-Saharan Africa. Afr J AIDS Res. 2009;8(4):401–12. [DOI] [PubMed] [Google Scholar]
  • 16.Pienaar M, van Rooyen FC, Walsh CM. Household food security and HIV status in rural and urban communities in the Free State province, South Africa. SAHARA J J Soc Asp HIVAIDS Res Alliance. 2017;14(1):118–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Senkomago V, Guwatudde D, Breda M, Khoshnood K. Barriers to antiretroviral adherence in HIV-positive patients receiving free medication in Kayunga, Uganda. AIDS Care. 2011;23(10):1246–53. [DOI] [PubMed] [Google Scholar]
  • 18.Singer M Toward a critical biosocial model of ecohealth in Southern Africa: the HIV/AIDS and nutrition insecurity syndemic. Ann Anthropol Pract. 2011;35(1):8–27. [Google Scholar]
  • 19.Drain PK, Losina E, Parker G, Giddy J, Ross D, Katz JN, et al. Risk factors for late-stage HIV disease presentation at initial HIV diagnosis in Durban, South Africa. PLoS ONE. 2013;8(1):e55305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Kagee A, Steel H, Coetzee B. The relationship between structural barriers to adherence to antiretroviral therapy, psychological distress, and health-related quality of life. S Afr J Psychol. 2014;44(2):170–9. [Google Scholar]
  • 21.Mepham S, Zondi Z, Mbuyazi A, Mkhwanazi N, Newell ML. Challenges in PMTCT antiretroviral adherence in northern Kwa-Zulu-Natal, South Africa. AIDS Care. 2011;23(6):741–7. [DOI] [PubMed] [Google Scholar]
  • 22.Coetzee B, Kagee A, Vermeulen N. Structural barriers to adherence to antiretroviral therapy in a resource-constrained setting: the perspectives of health care providers. AIDS Care. 2011;23(2):146–51. [DOI] [PubMed] [Google Scholar]
  • 23.Goudge J, Ngoma B. Exploring antiretroviral treatment adherence in an urban setting in South Africa. J Public Health Policy. 2011;32(1):S52–64. [DOI] [PubMed] [Google Scholar]
  • 24.Pitpitan EV, Kalichman SC, Eaton LA, Cain D, Sikkema KJ, Watt MH, et al. Co-occurring psychosocial problems and HIV risk among women attending drinking venues in a South African township: a syndemic approach. Ann Behav Med. 2013;45(2):153–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Singer M A dose of drugs, a touch of violence, a case of AIDS: conceptualizing the Sava Syndemic. Free Inq Creat Sociol. 1996;24(2):99–110. [Google Scholar]
  • 26.Stall R, Friedman M, Catania JA. Interacting epidemics and gay men’s health: a theory of syndemic production among urban gay men. In: Wolitski RJ, Stall R, Valdiserri RO, editors. Unequal opportunity: health disparities affecting gay and bisexual men in the United States. New York: Oxford University Press; 2008. p. 251–74. [Google Scholar]
  • 27.Fried EI, Cramer AOJ. Moving forward: challenges and directions for psychopathological network theory and methodology. Perspect Psychol Sci. 2017;12(6):999–1020. [DOI] [PubMed] [Google Scholar]
  • 28.Biello KB, Oldenburg CE, Safren SA, Rosenberger JG, Novak DS, Mayer KH, et al. Multiple syndemic psychosocial factors are associated with reduced engagement in HIV care among a multinational, online sample of HIV-infected MSM in Latin America. AIDS Care. 2016;28(sup1):84–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Blashill AJ, Bedoya CA, Mayer KH, O’Cleirigh C, Pinkston MM, Remmert JE, et al. Psychosocial syndemics are additively associated with worse ART adherence in HIV-infected individuals. AIDS Behav. 2015;19(6):981–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Friedman MR, Stall R, Silvestre AJ, Wei C, Shoptaw S, Herrick A, et al. Effects of syndemics on HIV viral load and medication adherence in the multicentre AIDS cohort study. AIDS. 2015;29(9):1087–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Glynn TR, Safren SA, Carrico AW, Mendez NA, Duthely LM, Dale SK, et al. High levels of syndemics and their association with adherence, viral non-suppression, and biobehavioral transmission risk in Miami, a US City with an HIV/AIDS epidemic. AIDS Behav. 2019;23(11):2956–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Harkness A, Bainter SA, O’Cleirigh C, Mendez NA, Mayer KH, Safren SA. Longitudinal effects of syndemics on ART non-adherence among sexual minority men. AIDS Behav. 2018;22(8):2564–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Harkness A, Bainter SA, O’Cleirigh C, Albright C, Mayer KH, Safren SA. Longitudinal effects of syndemics on HIV-positive sexual minority men’s sexual health behaviors. Arch Sex Behav. 2019;48(4):1159–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Kuhns LM, Hotton AL, Garofalo R, Muldoon AL, Jaffe K, Bouris A, et al. An index of multiple psychosocial, syndemic conditions is associated with antiretroviral medication adherence among HIV-positive youth. AIDS Patient Care STDs. 2016;30(4):185–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Mizuno Y, Purcell DW, Knowlton AR, Wilkinson JD, Gourevitch MN, Knight KR. Syndemic vulnerability, sexual and injection risk behaviors, and HIV continuum of care outcomes in HIVpositive injection drug users. AIDS Behav. 2015;19(4):684–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Pantalone DW, Valentine SE, Woodward EN, O’Cleirigh C. Syndemic indicators predict poor medication adherence and increased healthcare utilization for urban HIV-positive men who have sex with men. J Gay Lesbian Ment Health. 2018;22(1):71–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Sullivan KA, Messer LC, Quinlivan EB. Substance abuse, violence, and HIV/AIDS (SAVA) syndemic effects on viral suppression among HIV positive women of color. AIDS Patient Care STDs. 2015;29(S1):S42–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Walters SM, Braksmajer A, Coston B, Yoon I, Grov C, Downing MJ, et al. A syndemic model of exchange sex among HIV-positive men who have sex with men. Arch Sex Behav. 2020. 10.1007/s10508-020-01628-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Wawrzyniak AJ, Rodríguez AE, Falcon AE, Chakrabarti A, Parra A, Park J, et al. Association of individual and systemic barriers to optimal medical care in people living with HIV/AIDS in Miami-Dade County. JAIDS J Acquir Immune Defic Syndr. 2015;1(69):S63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Choi KW, Smit JA, Coleman JN, Mosery N, Bangsberg DR, Safren SA, et al. Mapping a syndemic of psychosocial risks during pregnancy using network analysis. Int J Behav Med. 2019. http://search.ebscohost.com/login.aspx?direct=true&db=psyh&AN=2019-11144-001&site=ehost-live [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Lee JS, Safren SA, Bainter SA, Rodríguez-Díaz CE, Horvath KJ, Blashill AJ. Examining a syndemics network among Young Latino men who have sex with men. Int J Behav Med. 2020;27(1):39–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Lee JS, Bainter SA, Carrico AW, Glynn TR, Rogers BG, Albright C, et al. Connecting the dots: a comparison of network analysis and exploratory factor analysis to examine psychosocial syndemic indicators among HIV-negative sexual minority men. J Behav Med. 2020;43(6):1026–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Choi KW, Batchelder AW, Ehlinger PP, Safren SA, O’Cleirigh C. Applying network analysis to psychological comorbidity and health behavior: depression, PTSD, and sexual risk in sexual minority men with trauma histories. J Consult Clin Psychol. 2017;85(12):1158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Cramer AOJ, Waldorp LJ, van der Maas HLJ, Borsboom D. Comorbidity: a network perspective. Behav Brain Sci. 2010;33:137–93. [DOI] [PubMed] [Google Scholar]
  • 45.Epskamp S, Fried EI. A tutorial on regularized partial correlation networks. Psychol Methods. 2018;23:617. [DOI] [PubMed] [Google Scholar]
  • 46.Rhemtulla M, Fried EI, Aggen SH, Tuerlinckx F, Kendler KS, Borsboom D. Network analysis of substance abuse and dependence symptoms. Drug Alcohol Depend. 2016;161:230–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Fried EI, Epskamp S, Nesse RM, Tuerlinckx F, Borsboom D. What are “good” depression symptoms? Comparing the centrality of DSM and non-DSM symptoms of depression in a network analysis. J Affect Disord. 2016;1(189):314–20. [DOI] [PubMed] [Google Scholar]
  • 48.Armour C, Fried EI, Deserno MK, Tsai J, Pietrzak RH. A network analysis of DSM-5 posttraumatic stress disorder symptoms and correlates in U.S. military veterans. J Anxiety Disord. 2017;45:49–59. [DOI] [PubMed] [Google Scholar]
  • 49.Spiller TR, Schick M, Schnyder U, Bryant RA, Nickerson A, Morina N. Symptoms of posttraumatic stress disorder in a clinical sample of refugees: a network analysis. Eur J Psychotraumatol. 2017;8(sup3):1318032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Harris PA, Taylor R, Minor BL, Elliott V, Fernandez M, O’Neal L, et al. The REDCap consortium: building an international community of software platform partners. J Biomed Inform. 2019;1(95): 103208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Brislin RW. Back-translation for cross-cultural research. J Cross-Cult Psychol. 1970;1(3):185–216. [Google Scholar]
  • 53.Meltzer-Brody S, Churchill E, Davidson JRT. Derivation of the SPAN, a brief diagnostic screening test for post-traumatic stress disorder. Psychiatry Res. 1999;88(1):63–70. [DOI] [PubMed] [Google Scholar]
  • 54.Radloff LS. The CES-D scale a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1(3):385–401. [Google Scholar]
  • 55.Andersen LS, Magidson JF, O’Cleirigh C, Remmert JE, Kagee A, Leaver M, et al. A pilot study of a nurse-delivered cognitive behavioral therapy intervention (Ziphamandla) for adherence and depression in HIV in South Africa. J Health Psychol. 2018;23(6):776–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Connor KM, Kobak KA, Churchill LE, Katzelnick D, Davidson JR. Mini-SPIN: a brief screening assessment for generalized social anxiety disorder. Depress Anxiety. 2001;14(2):137–40. [DOI] [PubMed] [Google Scholar]
  • 57.WHO Assist Working Group. The alcohol, smoking and substance involvement screening test (ASSIST): development, reliability and feasibility. Addiction. 2002;97(9):1183–94. [DOI] [PubMed] [Google Scholar]
  • 58.Humeniuk R, World Health Organization. The Alcohol, smoking and substance involvement screening test (ASSIST): manual for use in primary care. Geneva: World Health Organization; 2010. [Google Scholar]
  • 59.Saunders JB, Aasland OG, Babor TF, De la Fuente JR, Grant M. Development of the alcohol use disorders identification test (AUDIT). WHO collaborative project on early detection of persons with harmful alcohol consumption-II. Addict-ABINGDON. 1993;88:791–791. [DOI] [PubMed] [Google Scholar]
  • 60.Conigrave KM, Hall WD, Saunders JB. The AUDIT questionnaire: choosing a cut-off score. Alcohol use disorder identification test. Addict Abingdon Engl. 1995;90(10):1349–56. [DOI] [PubMed] [Google Scholar]
  • 61.Coates J, Swindale A, Bilinsky P. Household Food Insecurity Access Scale (HFIAS) for Measurement of Food Access: Indicator Guide: Version 3. Washington, DC: Food and Nutrition Technical Assistance Project, Academy for Educational Development; 2007. 10.1037/e576842013-001 [DOI] [Google Scholar]
  • 62.Coetzee B, Kagee A. The development of an inventory to assess the structural barriers to clinic attendance and pill-taking amongst users of antiretroviral therapy. AIDS Behav. 2013;17(1):319–28. [DOI] [PubMed] [Google Scholar]
  • 63.R Development Core Team. R: a language and environment for statistical computing [Internet]. Vienna, Austria: R Foundation for Statistical Computing; 2016. Available from: http://www.R-project.org [Google Scholar]
  • 64.Mazumder R, Hastie T. The graphical lasso: new insights and alternatives. Electron J Stat. 2012;9(6):2125–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Borsboom D, Cramer AO. Network analysis: an integrative approach to the structure of psychopathology. Annu Rev Clin Psychol. 2013;9:91–121. [DOI] [PubMed] [Google Scholar]
  • 66.Holgado-Tello FP, Chacón-Moscoso S, Barbero-García I, Vila-Abad E. Polychoric versus Pearson correlations in exploratory and confirmatory factor analysis of ordinal variables. Qual Quant. 2010;44(1):153. [Google Scholar]
  • 67.Boccaletti S, Latora V, Moreno Y, Chavez M, Hwang DU. Complex networks: structure and dynamics. Phys Rep. 2006;424(4):175–308. [Google Scholar]
  • 68.Csardi G, Nepsuz T. Igraph: network analysis and visualization. Int J. 2006. Available from: https://igraph.org [Google Scholar]
  • 69.Yang Z, Algesheimer R, Tessone CJ. A comparative analysis of community detection algorithms on artificial networks. Sci Rep. 2016;6(1):30750. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Epskamp S, Fried EI. bootnet: bootstrap methods for various network estimation routines. 2017. Available from: https://CRAN.R-project.org/package=bootnet
  • 71.Opsahl T, Agneessens F, Skvoretz J. Node centrality in weighted networks: generalizing degree and shortest paths. Soc Netw. 2010;32(3):245–51. [Google Scholar]
  • 72.Tsai AC, Mendenhall E, Trostle JA, Kawachi I. Co-occurring epidemics, syndemics, and population health. The Lancet. 2017;389(10072):978–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 73.Columb MO, Sagadai S. Multiple comparisons. Curr Anaesth Crit Care. 2006;17(3):233–6. [Google Scholar]
  • 74.Bringmann LF, Vissers N, Wichers M, Geschwind N, Kuppens P, Peeters F, et al. A network approach to psychopathology: new insights into clinical longitudinal data. PLoS ONE. 2013;8(4): e60188. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.Elovainio M, Hakulinen C, Pulkki-Råback L, Juonala M, Raitakari OT. A network approach to the analysis of psychosocial risk factors and their association with health. J Health Psychol. 2020;25(10–11):1587–600. [DOI] [PubMed] [Google Scholar]
  • 76.Beck AT. Cognitive models of depression. Clin Adv Cogn Psychother Theory Appl. 2002;14(1):29–61. [Google Scholar]
  • 77.Clark DA, Beck AT. Cognitive theory and therapy of anxiety and depression: convergence with neurobiological findings. Trends Cogn Sci. 2010;14(9):418–24. [DOI] [PubMed] [Google Scholar]
  • 78.Elwood LS, Hahn KS, Olatunji BO, Williams NL. Cognitive vulnerabilities to the development of PTSD: a review of four vulnerabilities and the proposal of an integrative vulnerability model. Clin Psychol Rev. 2009;29(1):87–100. [DOI] [PubMed] [Google Scholar]
  • 79.Foa EB, Steketee G, Rothbaum BO. Behavioral/cognitive conceptualizations of post-traumatic stress disorder. Behav Ther. 1989;20(2):155–76. [Google Scholar]
  • 80.Friedman MJ, Yehuda R. Post-traumatic stress disorder and comorbidity: psychobiological approaches to differential diagnosis. In: Neurobiological and clinical consequences of stress: From normal adaptation to post-traumatic stress disorder. Philadelphia, PA, US: Lippincott Williams & Wilkins Publishers; 1995. p. 429–45. [Google Scholar]
  • 81.Ingram RE, Price JM. Vulnerability to psychopathology: risk across the lifespan. New York: Guilford Press; 2010. [Google Scholar]
  • 82.Tull MT, Gratz KL, Salters K, Roemer L. The role of experiential avoidance in posttraumatic stress symptoms and symptoms of depression, anxiety, and somatization. J Nerv Ment Dis. 2004;192(11):754–61. [DOI] [PubMed] [Google Scholar]
  • 83.Leung CW, Epel ES, Willett WC, Rimm EB, Laraia BA. Household food insecurity is positively associated with depression among low-income supplemental nutrition assistance program participants and income-eligible nonparticipants. J Nutr. 2015;145(3):622–7. [DOI] [PubMed] [Google Scholar]
  • 84.Boateng GO, Workman CL, Miller JD, Onono M, Neilands TB, Young SL. The syndemic effects of food insecurity, water insecurity, and HIV on depressive symptomatology among Kenyan women. Soc Sci Med. 2020;15: 113043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 85.Tsai AC, Bangsberg DR, Frongillo EA, Hunt PW, Muzoora C, Martin JN, et al. Food insecurity, depression and the modifying role of social support among people living with HIV/AIDS in rural Uganda. Soc Sci Med. 2012;74(12):2012–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 86.Tsai AC, Tomlinson M, Comulada WS, Rotheram-Borus MJ. Food insufficiency, depression, and the modifying role of social support: evidence from a population-based, prospective cohort of pregnant women in Peri-Urban South Africa. Soc Sci Med. 2016;1982(151):69–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 87.Alpert JE, Fava M. Nutrition and depression: the role of folate. Nutr Rev. 1997;55(5):145–9. [DOI] [PubMed] [Google Scholar]
  • 88.Friedman M. Analysis, nutrition, and health benefits of tryptophan. Int J Tryptophan Res. 2018;1(11):1178646918802282. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 89.German L, Kahana C, Rosenfeld V, Zabrowsky I, Wiezer Z, Fraser D, et al. Depressive symptoms are associated with food insufficiency and nutritional deficiencies in poor community-dwelling elderly people. J Nutr. 2011;15(1):6. [DOI] [PubMed] [Google Scholar]
  • 90.Popa T, Ladea M. Nutrition and depression at the forefront of progress. J Med Life. 2012;5(4):414–9. [PMC free article] [PubMed] [Google Scholar]
  • 91.Racagni G, Popoli M. Cellular and molecular mechanisms in the long-term action of antidepressants. Dialogues Clin Neurosci. 2008;10(4):385–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 92.Batchelder AW, Foley JD, Wirtz MR, Mayer K, O’Cleirigh C. Substance use stigma, avoidance coping, and missed HIV appointments among MSM who use substances. AIDS Behav. 2021;25(5):1454–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 93.Hammarlund R, Crapanzano K, Luce L, Mulligan L, Ward K. Review of the effects of self-stigma and perceived social stigma on the treatment-seeking decisions of individuals with drug- and alcohol-use disorders. Subst Abuse Rehabil. 2018;23(9):115–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 94.Peirce R, Frone M, Russell M, Cooper M, Mudar P. A longitudinal model of social contact, social support, depression, and alcohol use. Health Psychol. 2000;1(19):28–38. [DOI] [PubMed] [Google Scholar]
  • 95.Psaros C, Smit JA, Mosery N, Bennett K, Coleman JN, Bangsberg DR, et al. PMTCT adherence in pregnant South African Women: the role of depression, social support, stigma, and structural barriers to care. Ann Behav Med. 2020;54(9):626–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 96.Bangsberg DR, Deeks SG. Spending more to save more: interventions to promote adherence. Ann Intern Med. 2010;152(1):54–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 97.Kehoe K, Boulle A, Tsondai PR, Euvrard J, Davies MA, Cornell M. Long-term virologic responses to antiretroviral therapy among HIV-positive patients entering adherence clubs in Khayelitsha, Cape Town, South Africa: a longitudinal analysis. J Int AIDS Soc. 2020;23(5): e25476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 98.Venter WF, Kaiser B, Pillay Y, Conradie F, Gomez GB, Clayden P, et al. Cutting the cost of South African antiretroviral therapy using newer, safer drugs. South Afr Med J Suid-Afr Tydskr Vir Geneeskd. 2016;107(1):28–30. [DOI] [PubMed] [Google Scholar]
  • 99.Magidson JF, Fatch R, Orrell C, Amanyire G, Haberer JE, Hahn JA, et al. Biomarker-measured unhealthy alcohol use in relation to CD4 count among individuals starting ART in Sub-Saharan Africa. AIDS Behav. 2018. 10.1007/s10461-018-2364-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 100.Mitchell J, Wight M, Van Heerden A, Rochat TJ. Intimate partner violence, HIV, and mental health: a triple epidemic of global proportions. Int Rev Psychiatry. 2016;28(5):452–63. [DOI] [PubMed] [Google Scholar]
  • 101.Matseke G, Rodriguez VJ, Peltzer K, Jones D. Intimate partner violence among HIV positive pregnant women in South Africa. J Psychol Afr. 2016. 10.1080/14330237.2016.1185912. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 102.Battersby J Urban food insecurity in Cape Town, South Africa: an alternative approach to food access. Dev South Afr. 2011;28(4):545–61. [Google Scholar]
  • 103.Mendenhall E, Kohrt BA, Logie CH, Tsai AC. Syndemics and clinical science. Nat Med. 2022;28(7):1359–62. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Data are available upon request to the corresponding author.

RESOURCES