Secondary Logo

Journal Logo

Prevention Research

A Social Network Analysis of HIV Treatment Partners and Patient Viral Suppression in Botswana

Bogart, Laura M. PhD*; Mosepele, Mosepele MD, MSc†,‡; Phaladze, Nthabiseng PhD§; Lekoko, Bright RN; Klein, David J. MS*; MacCarthy, Sarah ScD*; Green, Harold D. Jr PhD*,‖

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: June 1, 2018 - Volume 78 - Issue 2 - p 183-192
doi: 10.1097/QAI.0000000000001661
  • Free



Botswana has an HIV prevalence of 21.9% among those aged 15–49 years,1 and 90% of people living with HIV (PLWH) on treatment are virally suppressed.2 To eradicate HIV and maximize individual health benefits of antiretroviral therapy (ART), innovative, scalable, and low-cost interventions are needed to improve adherence among the remaining individuals who are not suppressed.

Community-based interventions hold promise for supporting PLWH with adherence, especially in resource-limited settings.3 One intervention used in at least 20 countries is treatment partners (aka treatment/adherence buddies/supporters), who tend to be trusted family members or friends selected by patients.4–13 Treatment partners can motivate adherence and engagement in care and accompany patients to appointments.14 According to 2016 Botswana guidelines,15 “Before initiating ART ensure…that adherence partners are present for psycho-social support.” Although treatment partners are still strongly recommended, recent guidelines clarified that ART initiation should not be postponed for patients without a treatment partner. However, patients are given little guidance about treatment partner selection (aside from choosing someone to whom they have disclosed), and treatment partners are given little guidance about how to support patients.

Randomized controlled trials (RCTs) of treatment partners show mixed findings.16–19 A network meta-analysis of global health studies indicated that enhanced standard of care (adherence education/counseling) plus a treatment partner had greater effects on viral suppression than standard of care plus a treatment partner, although the analysis conflated studies of patient-selected treatment partners with studies of clinic-based volunteers or employees.20 Two RCTs in sub-Saharan Africa, including an 8-country RCT that included Botswana, found nonsignificant effects on viral suppression for treatment partners trained on directly observed medication-taking (vs. untrained treatment partners).17,21 Another RCT of treatment partners trained in adherence promotion found effects on adherence (although viral suppression was not assessed).22 In an RCT in Nigeria, a higher percentage of patients who selected treatment partners vs. in standard-of-care reported improved adherence but not long-term viral suppression.23 Given known strong effects of social support on adherence,24 mixed results may be because of moderating factors related to patient characteristics, treatment partner type, quality of support, and lack of guidelines regarding treatment partner selection and responsibilities.

Nonrandomized studies have found better health and psychosocial outcomes for patients with treatment partners and have helped to outline conditions under which treatment partners may be effective.25–29 A study in South Africa found that patients who selected treatment partners perceived lower stigma over time.30 Other research suggests that patients with both treatment partners and trained peer supporters (eg, who perform directly observed adherence, home visits, and/or appointment accompaniment) have better outcomes (eg, improved self-reported adherence and mental health) than those with a treatment partner only.25,31 Qualitative research in Uganda and South Africa further suggests that effective treatment partners provide material, emotional, and instrumental support, instill a sense of reciprocity (eg, are also HIV-positive), and strengthen social connectedness.32–34

Programs that use treatment partners necessarily leverage patients' social networks, requiring that patients select someone in their social circle with whom they have a close relationship. Research shows that social networks affect health behaviors and outcomes, including adherence, in part via social support provision.24,35–38 Moreover, close relationships have larger or similar effects on risk for mortality as compared to smoking, high body mass index, and low physical activity.39 Thus, we examined treatment partners from a social network perspective, and in the context of social psychological literature on close relationships,40 for characteristics of patients and treatment partners, and patient–treatment partner relationships, associated with patient viral suppression.



Participants were recruited from the Princess Marina Hospital-Infectious Disease Care Clinic (IDCC), the largest HIV clinic in Gaborone, Botswana. The clinic has almost 7000 patients (∼60% women), and 80–100 patients visit daily.

Community Engagement

The team engaged with the community advisory board (CAB) of the research institute that was the main study site. The CAB, which is composed of key stakeholders, including PLWH, considers community issues relevant to project implementation and participant needs. The CAB provided input on the protocol and project design, interpretation of results, and implications for policy and future research.


Patients were eligible if they (1) were 18 years of age or older, (2) on ART and initiated ART at least 6 months before enrollment, and (3) had a treatment partner whom they were willing to refer to the study. Treatment partners were eligible if they (1) were at least 18 years old and (2) selected to be the treatment partner of a patient participant.

Using a case-controlled design, our goal was to purposively select half of the patient sample to be virally suppressed (viral load <400 copies/mL within previous 6 months at enrollment) and half unsuppressed (≥2 documented viral loads ≥400 copies/mL within previous 6 months). Viral load and time from ART initiation were confirmed with clinic records.


Each week, a research nurse checked medical records of patients who had appointments for potential eligibility. When potentially eligible patients arrived at the clinic, the nurse approached them for screening and invited eligible patients to return with their treatment partner.


Patients and treatment partners participated in separate 1-time interviews to assess individual-level and social network characteristics. The interviewer used EgoWeb 2.041 to guide participants through the 2-hour assessment, which was conducted in Setswana. Participants received 40 pula (∼US$5).

Individual-Level Characteristics

The survey assessed sociodemographic characteristics (age, gender, marital status, education level, employment status, and monthly income), number of years since diagnosis, and known correlates of ART adherence (a strong predictor of viral suppression), including depression (Patient Health Questionnaire-9)42, internalized HIV stigma (6-item Internalized AIDS-Related Stigma Scale, 1 = strongly disagree to 6 = strongly agree; α = 0.68)43, perceived HIV discrimination (sum of 5 items assessing frequency of different types of discrimination, dichotomized as “never” vs. “ever”; eg, frequency of being gossiped about44), problem alcohol use,45 and perceived barriers to care (sum of 9 items assessing agreement with different barriers; eg, transportation issues, dichotomized as “somewhat or strongly agree” vs. “somewhat or strongly disagree”). Treatment partners' HIV-serostatus was assessed via self-report.

Social Network Characteristics

We used personal network methods,46–48 which have been shown to be reliable for capturing real-world variability of most network structure and composition characteristics49 and valid for capturing basic network member characteristics.50 In the Alter (ie, social network member) Elicitation Section, participants listed initials of 20 people with whom they were in contact in the past year (eg, phone, in-person), starting with those most important to them. To assess network structure, participants indicated how often they thought each pair of alters interacted (0 = never, 1 = almost never, 2 = sometimes, 3 = always/almost always). In the Network Composition Section, for each alter, participants provided sociodemographic information (eg, age), relationship to participant (eg, spouse), ratings of closeness and trust (0 = not at all to 2 = very), and whether the patient had disclosed their serostatus to the alter.

Because social support is a strong predictor of adherence,24 we asked participants to report the extent to which each alter was likely to provide emotional, informational, affectionate, and instrumental support using 4 items from the Medical Outcomes Study social support scale,51 with responses 0 = not likely, 1 = somewhat likely, and 2 = very likely (α = 0.66). HIV discrimination was assessed with the sum of 6 yes/no items, asking whether each alter expressed stigmatizing attitudes or showed discrimination (eg, asked them to use separate plates/utensils).52

Network Variables

We derived network density, a structural measure, as percentage of existing ties relative to total number of possible ties.53 Network composition indicators included network member characteristics (eg, proportion of women).

Statistical Analysis

After examining descriptive statistics, we conducted bivariate logistic regressions predicting viral suppression with patient and treatment partner individual-level variables and social network–level variables. Because of the large number of potential predictors (and inadequate statistical power to examine all simultaneously), we examined variables in 3 separate conceptually distinct sets, conducting multivariate logistic regressions predicting viral suppression with (1) patient individual-level variables, (2) treatment partner individual-level variables, and (3) social network–level variables at least marginally significant in the bivariate models (at P < 0.10). We then conducted a fourth regression that included all significant variables (at P < 0.10) not highly correlated with each other (at r > ±0.50) from the first 3 regressions. Our analysis assumed each dyad was independent and each patient was associated with a unique social network.54 Data of patients and treatment partners were linked using unique dyadic identification numbers.

Because research suggests that men may react more negatively than women to spouses' attempts to control their illness management,40 we additionally tested whether gender moderated the effects of having a spouse/partner as a treatment partner; the interaction term was not significant (P > 0.10) and thus will not be discussed further.

Our analysis sample was designed such that half of those recruited would be virally suppressed, but in Botswana, 90% of PLWH on ART are suppressed.2 Thus, as a sensitivity analysis, we repeated bivariate regressions using poststratification weights to account for the sample's underrepresentation of suppressed patients. Suppressed patients were assigned a weight equal to (population mean)/(sample mean), and unsuppressed patients were assigned a weight of (1 − population mean)/(1 − sample mean).55 Furthermore, because our main outcome was viral suppression at one point in time, we repeated the bivariate regressions using long-term viral suppression (past 4 measurements) as another sensitivity analysis.

Qualitative Component and Analysis

After the quantitative assessment, the interviewer used EgoWeb's network visualization capabilities to create a network diagram and recorded responses to open-ended questions about the diagram. Figure 1 displays 2 network diagrams from the sample, one of a virally unsuppressed patient (Fig. 1A) and the other of a virally suppressed patient (Fig. 1B). To ensure that the diagrams were representative of each subgroup, we selected participants whose characteristics were closest to the subgroup mean (within 1 SD) on key individual sociodemographic and network-level variables (listed in Table 1). Each node represents a network member and each line means that the 2 network members know each other; more lines represent greater interconnections between members. Treatment partners are triangles. Females are represented in black and males in white. Larger circles represent greater emotional closeness, showing that the suppressed patient reported greater average closeness to network members than did the unsuppressed patient.

Examples of participant social network diagrams. A, Diagram of a patient who is not virally suppressed. B, Diagram of a patient who is virally suppressed. Note that patients themselves are not depicted in the diagrams.
Characteristics of the Patient and Treatment Partner Samples
Characteristics of the Patient and Treatment Partner Samples

Patients were asked to focus on the treatment partner in their network diagram and describe how the treatment partner supported them. Treatment partners were asked to focus on the patient in their network diagram and describe how they supported the patient; they also were asked their opinion about the kinds of training treatment partners need. The interviewer used probes and open-ended questions to avoid biasing answers and allow for new topic exploration.56–58 Because responses were brief, the interviewer was able to transcribe responses verbatim into EgoWeb during the interview.

Two team members developed a codebook by reviewing open-ended responses and writing up summary notes.59 They then marked areas of text pertaining to codes capturing types of support provided (adherence support, medical appointment support, or nonmedical support), from the patient and treatment partner perspective, and suggestions for treatment partner training (treatment/adherence education, basic HIV knowledge, or counseling skills). Coder consistency across responses was good (Cohen Kappa mean across codes = 0.87; range = 0.69–1.0).60 Responses were used in this mixed-methods analysis to complement our quantitative results, by helping to explain why some treatment partners might have been more effective than others.61



Using convenience sampling, 465 potentially eligible patients (adults who had been on ART >6 months) were screened, and of those, 141 patients were enrolled. Although data were not systematically collected on the number of eligible and ineligible patients screened, patients screened as ineligible primarily because they were unable to visit the clinic with their treatment partner for the interview (eg, they were not in contact with their treatment partner or their treatment partner was unavailable). Treatment partners of all 141 enrolled patients were offered participation; 10 did not show up for the scheduled interview, despite multiple attempts at contact. Thus, the present analysis includes only the 131 patients whose treatment partners participated.

Table 1 shows participants' sociodemographic and social network characteristics. Patients and treatment partners had similar characteristics: they were almost 40 years old on average, about a quarter had low education, and about 10% had low income. Patients were significantly less likely to be female (P < 0.05) and married/cohabitating (P < 0.001) than treatment partners. Patients were diagnosed with HIV more than 10 years prior on average. About half of patients selected a family member as their treatment partner, 37% selected a spouse/partner, and 19% considered their treatment partner to be a friend (because the categories were not mutually exclusive, the percentages add up to >100%). Half of treatment partners were HIV-positive. By design, about half of the patients (51.1%) were virally suppressed.

Bivariate and Multivariate Correlates of Viral Suppression

Individual-Level Patient Predictors

In bivariate tests, patients were more likely to be suppressed if they were older, male, married/cohabitating with a partner, and employed; endorsed fewer barriers to care; and did not report discrimination (Table 2). In the multivariate model, being older and male remained significantly related to suppression.

Bivariate and Multivariate Logistic Regressions Predicting Viral Suppression With Individual-Level Variables
Bivariate and Multivariate Logistic Regressions Predicting Viral Suppression With Individual-Level Variables

Individual-Level Treatment Partner Predictors

In bivariate analyses, patients were more likely to be suppressed if their treatment partner was a spouse/intimate partner and not another family member, and HIV-positive (Table 2). In the multivariate model, only one variable remained significant: patients whose treatment partners were spouses/intimate partners were more likely to be suppressed.

Social Network–Level Predictors

Patients were more likely to be suppressed if their alters were older on average, if they had a higher percentage of alters who were male, and if they felt closer emotionally to and had more trust in alters on average (Table 2). In the multivariate model, older age was marginally significant, and percentage of male alters and average alter closeness remained significant.

Final Multivariate Model: Individual-Level and Social Network–Level Predictors of Viral Suppression

The multivariate model combining all significant and marginally significant (P < 0.10) predictors from the individual- and network-level models indicated that patients were more likely to be virally suppressed if they had a treatment partner who was a spouse/intimate partner and if they felt emotionally close to their alters on average (Table 3).

Final Multivariate Logistic Regression Predicting Viral Suppression With Patient Characteristics, Patient Network Composition and Structure, and Treatment Partner Characteristics

Sensitivity Analyses

Results of the sensitivity analyses (weighted regressions and regressions predicting long-term viral suppression) were similar in the effect sizes and significance levels of the unweighted bivariate tests.

Qualitative Analysis

Table 4 shows example quotes for coding categories related to treatment partner support and training. Both patients and treatment partners discussed ways in which treatment partners supported patients, through adherence and medical appointment support and nonmedical support. Adherence support most often included reminders to take medications, picking up refills, bringing medications to patients, and appointment accompaniment.

Coding Categories, Coder Consistency, and Example Quotes for Open-Ended Responses

Of patients who reported receiving nonmedical support from their treatment partner, 69% were suppressed and only 31% were unsuppressed. Treatment partners of unsuppressed patients tended to focus more narrowly on medication reminders, appointment reminders, and accompaniment, whereas treatment partners of suppressed patients tended to provide more in-depth adherence support (eg, dropping off medications) and broadly supporting patients to stay healthy in nonmedical ways not directly tied to medications. For example, treatment partners of suppressed patients were more likely to prepare food, bathe patients, and proactively attend to patients' needs so that patients were healthy overall. They reminded patients to eat healthy, to exercise, and not to use alcohol or drugs; they also provided general emotional support and acceptance, to motivate adherence. Few treatment partners of unsuppressed patients mentioned supporting patients for nonmedical needs, and those who did tended to help with food-related needs directly related to medication-taking (eg, reminders to eat in conjunction with medication-taking) rather than taking a broader focus on overall health (see “Nonmedical Support” quotes in Table 4).

Treatment partners of both suppressed and unsuppressed patients requested ongoing training on basic HIV knowledge and medication side effects, as well as counseling skills education to help them to be patient, motivate patients, and provide emotional support. However, a higher percentage of treatment partners of unsuppressed patients (50%) than treatment partners of suppressed patients (37%) requested counseling skills training.


This study indicates that characteristics of patients' social networks and of the patient–treatment partner relationship are significantly associated with viral suppression, controlling for patient and treatment partner individual-level characteristics. Patients who selected spouses or other intimate partners and who trusted and felt close to their social network members as a whole were more likely to show viral suppression. Our results are consistent with the large body of work showing robust effects of social support on health and mortality.39 By taking a social network approach, our results extend previous research on treatment partners that have taken an individual-level patient perspective.11,32–34

Our qualitative results elucidated some reasons why treatment partners vary in effectiveness. We found that treatment partners of patients who are virally suppressed support patients beyond their adherence needs, providing more than adherence and appointment reminders, to consider patients' overall health and well-being. These results are consistent with previous research on diabetes adherence, showing that adherence increases if spouses' daily adherence reminders are supportive (eg, show appreciation and understanding) but decreases if spouses' daily adherence reminders are used as an attempt to control patients' behavior (eg, pressuring patients to adhere).62 Similarly, a study in Uganda found that HIV treatment partners who interacted with patients daily (after receiving SMS notifications about patients' nonadherence) were perceived to be unsupportive and overbearing.63 Thus, in this study, treatment partners who took a holistic approach to supporting patients could have been perceived to be more caring, which in turn motivated adherence. Alternately, treatment partners with a narrow focus on adherence may have been perceived negatively, as attempting to persuade patients without fully recognizing their life context and challenges to adherence.

Previous research on close relationships sheds light on why spouses may be more effective treatment partners. Close relationships may improve illness management to the extent that the illness is appraised by both members of the dyad to be shared or communal, in which both members of the couple take responsibility for illness management.40 In this study, spouses may have been more likely than other types of treatment partners to view patients' HIV as a shared stressor and to feel interdependent on patients, as the patients' survival may have been tied to their families' (and potentially children's) survival.

Our research not only has relevance for HIV but also for other conditions requiring strict adherence that may benefit from treatment partners. Treatment partners with whom patients are in close and trusted relationships, and who interact with the patient frequently, are likely to be effective for improving health outcomes. Treatment partners who are in close relationships with patients have frequent opportunities to observe patients and provide comfort, security, and tangible support in anticipation of patients' needs rather than waiting for patients to request support directly or simply pressuring patients to adhere without attending to other needs. Although our results indicated that spouses were particularly helpful, not all patients have spouses. However, treatment partners, regardless of whether they are spouses, can be trained on nonconfrontational counseling skills (such as motivational interviewing64) and made aware that support that is perceived to be badgering is ineffective and could lead to nonadherence.

Our research is limited by the cross-sectional design and correlational analysis. Although we surmise that, consistent with previous research,39 higher quality social relationships led to viral suppression in this study, it is possible that patients were more able to develop close relationships because they were feeling healthier. Moreover, our study was conducted in one large urban clinic in one high prevalence country, limiting generalizability. Cultural factors, such as gender and relationship norms, may have influenced the results, possibly explaining why another study (in Kenya) found significant effects for patient gender and nonsignificant effects for spouses as treatment partners.65 In addition, we could not independently validate that the treatment partner participants were patients' originally selected treatment partners. We also did not audio-record participants' responses to open-ended survey questions and thus could not check accuracy. Furthermore, our results could have been biased by only recruiting patients who could bring their treatment partners to the interview. Patients who were not willing or able to bring their treatment partner may have had less a supportive treatment partner and been getting support from social network members other than their selected treatment partner; interviewing this group would have enabled additional insights about treatment partner support and viral suppression in the context of the overall social network. In addition, our research may have less relevance for newly diagnosed patients in the era of test-and-treat, as they may not have time to disclose their serostatus before ART initiation and thus may have limited options for selecting treatment partners.

In sum, this study indicates that social network factors may moderate the effects of treatment partners on health, including the quality and type of the patient–treatment partner relationship, and the level of trust and closeness that patients have with treatment partners and other network members. Moreover, treatment partners need a basic level of training that includes not only fundamental education about treatment (ie, information about treatment, adherence and side effects) but also counseling skills education. Given that many countries globally have treatment guidelines recommending patient selection of treatment partners, strengthening treatment partner selection and training could have wide-reaching impact.


The authors would like to thank the Princess Marina Hospital IDCC staff for a good working relationship during the conduct of the study and the Botswana-Harvard AIDS Institute Partnership Community Advisory Board for their key input on the study design and results.


1. UNAIDS. Country factsheets: Botswana 2016. 2016; Available at:
2. Joint United Nations Programme on HIV/AIDS (UNAIDS). Ending AIDS: Progress Towards the 90-90-90 Targets. Global AIDS Update. Geneva, Switzerland: UNAIDS; 2017.
3. Wouters E, Van Damme W, Van Rensburg D, et al. Impact of community-based support services on antiretroviral treatment programme delivery and outcomes in resource-limited countries: a synthetic review. BMC Health Serv Res. 2012;12:194.
4. National Drug and Theraputics Policy Advisory Committee, The AIDS and TB Directorate, Ministry of Health and Child Welfare, , et al. Guidelines for Antiretroviral Therapy in Zimbabwe. Harare, Zimbabwe: Ministry of Health and Childcare; 2013.
5. Malawi Ministry of Health. Clinical Management of HIV in Children and Adults. 2011.
6. Republic of Uganda. National Antiretroviral Treatment and Care Guidelines for Adults, Adolescents, and Children. Kampala, Uganda: Ministry of Health; 2009.
7. United Republic of Tanzania Ministry of Health and Social Welfare, Tanzania Mainland, National AIDS Control Programme. In: National Guidelines for the Management of HIV and AIDS. Dar es Salaam, Tanzania: Government of Tanzania; 2012.
8. National Department of Health, South Africa. Clinical Guidelines for the Management of HIV and AIDS in Adults and Adolescents. Republic of South Africa, 2010. Available at:
9. Ministry of Medical Services, Repubic of Kenya. Guidelines for Antiretroviral Therapy in Kenya. Nairobi, Kenya: National AIDS/STI Control Programme; 2011.
10. Ministry of Health Government of Lesotho. National Guidelines on the Use of Antiretroviral Therapy for HIV Prevention and Treatment. Maseru, Lesotho; 2016. Available at:
11. Foster S, Nakamanya S, Kyomuhangi R, et al. The experience of “medicine companions” to support adherence to antiretroviral therapy: quantitative and qualitative data from a trial population in Uganda. AIDS Care. 2010;22:35–43.
12. Federal Ministry of Health Abuja - Nigeria. National guidelines for HIV and AIDS treatment and care. Abuja, Nijeria: Federal Ministry of Health; 2010.
13. Government of the Republic of Zambia Ministry of Health. Adult and Adolescent Antiretroviral Therapy Protocols. Lusaka, Zambia: Zambia Ministry of Health; 2010.
14. Republic of Namibia, Ministry of Health and Social Services, Directorate of Special Programs. National Guidelines for Antiretroviral Therapy. Windhoek, Namibia: Namibia Ministry of Health and Social Services; 2014.
15. Ministry of Health Botswana. 2016 handbook of the Botswana Integrated HIV Clinical Care Guidelines. Gaborone, Botswana: Ministry of Health Botswana. 2016.
16. Simoni JM, Pantalone DW, Plummer MD, et al. A randomized controlled trial of a peer support intervention targeting antiretroviral medication adherence and depressive symptomatology in HIV-positive men and women. Health Psychol. 2007;26:488–495.
17. Gross R, Zheng L, La Rosa A, et al. Partner-focused adherence intervention for second-line antiretroviral therapy: a multinational randomized trial (actg a5234). Lancet HIV. 2015;2:e12–e19.
18. Mugusi F, Mugusi S, Bakari M, et al. Enhancing adherence to antiretroviral therapy at the HIV clinic in resource constrained countries; the Tanzanian experience. Trop Med Int Health. 2009;14:1226–1232.
19. Nakigozi G, Makumbi FE, Bwanika JB, et al. Impact of patient-selected care buddies on adherence to HIV care, disease progression and conduct of daily life among pre-antiretroviral HIV-infected patients in Rakai, Uganda: a randomized controlled trial. J Acquir Immune Defic Syndr. 2015;70:75.
20. Mills EJ, Lester R, Thorlund K, et al. Interventions to promote adherence to antiretroviral therapy in Africa: a network meta-analysis. Lancet HIV. 2014;1:e104–e111.
21. Nachega JB, Chaisson RE, Goliath R, et al. Randomized controlled trial of trained patient-nominated treatment supporters providing partial directly observed antiretroviral therapy. AIDS. 2010;24:1273–1280.
22. Kunutsor S, Walley J, Katabira E, et al. Improving clinic attendance and adherence to antiretroviral therapy through a treatment supporter intervention in Uganda: a randomized controlled trial. AIDS Behav. 2011;15:1795–1802.
23. Taiwo BO, Idoko JA, Welty LJ, et al. Assessing the viorologic and adherence benefits of patient-selected HIV treatment partners in a resource-limited setting. J Acquir Immune Defic Syndr. 2010;54:85–92.
24. DiMatteo MR. Social support and patient adherence to medical treatment: a meta-analysis. Health Psychol. 2004;23:207–218.
25. Thomson DR, Rich ML, Kaigamba F, et al. Community-based accompaniment and psychosocial health outcomes in HIV-infected adults in Rwanda: a prospective study. AIDS Behav. 2014;18:368–380.
26. Stubbs BA, Micek MA, Pfeiffer JT, et al. Treatment partners and adherence to HAART in Central Mozambique. AIDS Care. 2009;21:1412–1419.
27. Wouters E, Van Damme W, Van Loon F, et al. Public-sector ART in the Free State Province, South Africa: community support as an important determinant of outcome. Soc Sci Med. 2009;69:1177–1185.
28. Birbeck GL, Chomba E, Kvalsund M, et al. Antiretroviral adherence in rural Zambia: the first year of treatment availability. Am J Trop Med Hyg. 2009;80:669–674.
29. Igumbor JO, Scheepers E, Ebrahim R, et al. An evaluation of the impact of a community-based adherence support programme on ART outcomes in selected government HIV treatment sites in South Africa. AIDS Care. 2011;23:231–236.
30. Masquillier C, Wouters E, Mortelmans D, et al. The impact of community support initiatives on the stigma experienced by people living with HIV/AIDS in South Africa. AIDS Behav. 2015;19:214–226.
31. Pearson CR, Micek MA, Simoni JM, et al. Randomized control trial of peer-delivered, modified directly observed therapy for HAART in Mozambique. J Acquir Immune Defic Syndr. 2007;46:238.
32. O'Laughlin KN, Wyatt MA, Kaaya S, et al. How treatment partners help: social analysis of an African adherence support intervention. AIDS Behav. 2012;16:1308–1315.
33. Nachega JB, Knowlton AR, Deluca A, et al. Treatment supporter to improve adherence to antiretroviral therapy in HIV-infected South African adults: a qualitative study. J Acquir Immune Defic Syndr. 2006;43:S127–S133.
34. Duwell MM, Knowlton AR, Nachega JB, et al. Patient-nominated, community-based HIV treatment supporters: patient perspectives, feasibility, challenges, and factors for success in HIV-infected South African adults. AIDS Patient Care STDS. 2013;27:96–102.
35. Berkman L, Glass T. Social integration, social networks, social support and health. In: Berkman L, Kawachi I, eds. Social Epidemiology. New York, NY: Oxford University Press; 2000.
36. Pescosolido B, Levy J. The role of social networks in health, illness, disease and healing: the accepting present, the forgotten past, and the dangerous potential for a complacent future. In: Levy J, Pescosolido B, eds. Advances in Medical Sociology: Social Networks and Health. Vol. 8. Greenwich, CT: JAI; 2002:3–25.
37. Heaney C, Israel B. Social networks and social support. In: Glanz K, Lewis FM, Rimer BK, eds. Health Behavior and Health Education: Theory, Research, and Practice, 2nd ed. San Francisco, CA: Jossey-Bass; 1999:179–205.
38. Bogart LM, Wagner GJ, Green HD, et al. Social network characteristics moderate the association between stigmatizing attributions about HIV and non-adherence among Black Americans living with HIV: a longitudinal assessment. Ann Behav Med. 2015;49:865–872.
39. Holt-Lunstad J, Robles TF, Sbarra DA. Advancing social connection as a public health priority in the United States. Am Psychol. 2017;72:517.
40. Martire LM, Helgeson VS. Close relationships and the management of chronic illness: associations and interventions. Am Psychol. 2017;72:601.
41. EgoWeb 2.0 Open source software for social network analysis [computer software]. Available at: Accessed November 5, 2017.
42. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med. 2001;16:606–613.
43. Kalichman SC, Simbayi LC, Cloete A, et al. Measuring AIDS stigmas in people living with HIV/AIDS: the internalized AIDS-related stigma scale. AIDS Care. 2009;21:87–93.
44. Neuman M, Obermeyer CM. Experiences of stigma, discrimination, care and support among people living with HIV: a four country study. AIDS Behav. 2013;17:1796–1808.
45. Cherpitel CJ. Screening for alcohol problems in the U.S. General population: comparison of the CAGE, RAPS4, and RAPS4-QF by gender, ethnicity, and service utilization. Alcohol Clin Exp Res. 2002;26:1686–1691.
46. Campbell K, Lee B. Name generators in surveys of personal networks. Soc Networks. 1991;13:203–221.
47. McCarty C. Structure in personal networks. J Soc Struct. 2002;3:20.
48. McCarty C, Bernard H, Killworth P, et al. Eliciting representative samples of personal networks. Soc Networks. 1997;19:303–323.
49. McCarty C, Kilworth PD. Impact of methods for reducing respondent burden on personal network structural measures. Soc Networks. 2007;29:300–315.
50. Green HD, Hoover MA, Wagner GJ, et al. Measuring agreement between egos and alters understanding informant accuracy in personal network studies. Field Meth. 2014;26:126–140.
51. Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci Med. 1991;32:705–714.
52. Genberg BL, Kawichai S, Chingono A, et al. Assessing HIV/AIDS stigma and discrimination in developing countries. AIDS Behav. 2008;12:772–780.
53. Wasserman S, Faust K. Social Network Analysis: Methods and Applications. New York, NY: Cambridge University Press; 1994.
54. Green HD Jr, Tucker JS, Wenzel SL, et al. Association of childhood abuse with homeless women's social networks. Child Abuse Negl. 2012;36:21–31.
55. Little RJA. Post-stratification: a modeler's perspective. J Am Stat Assoc. 1993;88:1001–1012.
56. Bernard HR. Research Methods in Anthropology: Qualitative and Quantitative Approaches. 2nd ed. Thousand Oaks, CA: Sage Publications; 1994.
57. Becker HS. Problems of inference and proof in participant observation. Am Sociol Rev. 1958;23:652–660.
58. Spradley JP. The Ethnographic Interview. New York, NY: Holt Rinehart and Winston; 1979.
59. Ryan GW, Bernard R. Data management and analysis methods. In: Denzin NK, Lincoln YS, eds. Handbook of Qualitative Research. 4th ed. Thousand Oaks, CA: Sage Publications; 2011:796–802.
60. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas. 1960;20:37–46.
61. Palinkas LA, Aarons GA, Horwitz S, et al. Mixed method designs in implementation research. Adm Policy Ment Health. 2011;38:44–53.
62. Stephens MAP, Franks MM, Rook KS, et al. Spouses' attempts to regulate day-to-day dietary adherence among patients with type 2 diabetes. Health Psychol. 2013;32:1029.
63. Atukunda EC, Musiimenta A, Musinguzi N, et al. Understanding patterns of social support and their relationship to an ART adherence intervention among adults in rural Southwestern Uganda. AIDS Behav. 2017;21:428–440.
64. Miller WR, Rollnick S. Motivational Interviewing: Helping People Change. 3rd ed. New York, NY: Guilford Press; 2012.
65. Kibaara C, Blat C, Lewis-Kulzer J, et al. Treatment buddies improve clinic attendance among women but not men on antiretroviral therapy in the Nyanza region of Kenya. AIDS Res Treat. 2016;2016:9124541.

adherence; antiretroviral therapy; Botswana; social support; viral load

Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.