Multi-level Determinants of Clinic Attendance and Antiretroviral Treatment Adherence Among Fishermen Living with HIV/AIDS in Communities on Lake Victoria, Uganda

Abstract

This cross-sectional study assessed determinants of HIV clinic appointment attendance and antiretroviral treatment (ART) adherence among 300 male fisherfolk on ART in Wakiso District, Uganda. Multi-level factors associated with missed HIV clinic visits included those at the individual (age, AOR = 0.98, 95% CI 0.97–0.99), interpersonal (being single/separated from partner, AOR: 1.25, 95% CI 1.01–1.54), normative (anticipated HIV stigma, AOR: 1.55, 95% CI 1.05–2.29) and physical/built environment-level (travel time to the HIV clinic, AOR: 1.11, 95% CI 1.02–1.20; structural-barriers to ART adherence, AOR: 1.27, 95% CI 1.04–1.56; accessing care on a landing site vs. an island, AOR: 1.35, 95% CI 1.08–1.67). Factors associated with ART non-adherence included those at the individual (age, β: − 0.01, η² = 0.03; monthly income, β: − 0.01, η² = 0.02) and normative levels (anticipated HIV stigma, β: 0.10, η² = 0.02; enacted HIV stigma, β: 0.11, η² = 0.02). Differentiated models of HIV care that integrate stigma reduction and social support, and reduce the number of clinic visits needed, should be explored in this setting to reduce multi-level barriers to accessing HIV care and ART adherence.

Introduction

Ugandan fisherfolk are significantly more likely to be infected with HIV/AIDS than the general adult population, with HIV prevalence estimated between 22 and 40% [1–4] compared to 6.5% [5]. Thus, fisherfolk are a most-at-risk population (MARP) being prioritized for immediate initiation of antiretroviral treatment (ART) upon HIV diagnosis [6, 7]. Immediate ART initiation has significant benefits for individual health and can reduce HIV incidence on a population-level [6, 8]. The success of these guidelines, however, depends greatly on the ability to engage and retain MARPs across all stages of the HIV care continuum. There is a need to increase engagement among men in sub-Saharan Africa (SSA) in particular, who have worse engagement across all stages of care compared to women [9–11]. Fishermen and other men employed in the fishing industry may face even greater challenges than men generally, given their frequent mobility, the nature of their work, and the structural environment of fishing villages. However, no research to our knowledge has examined HIV care engagement with fisherfolk after linking to care.

Given the unique setting of fishing communities, barriers to HIV care engagement likely occur at multiple levels of the social ecological model (SEM) [12–14]. The SEM is a framework that is commonly used to understand the complex interactions between an individual and factors in the broader environment on health behavior and health outcomes. The SEM recognizes individual factors and factors external to the individual as interacting to influence health behavior/outcomes, including factors at the level of the individual, interpersonal, normative environment, physical/built environment, and policy environment [12–14] (see Fig. 1 for a graphic depiction of the SEM). This model has been used to understand HIV behaviors generally [15], including HIV medication adherence and retention in HIV care in various settings [15–18]. Individual-level factors shown to predict better engagement in care in SSA include motivation and knowledge about ART, and socio-demographics such as older age, higher education, being female, and employment [19–28]. In Ugandan fishing villages, the occurrence of heavy alcohol use [2, 29–31] may also be important, as alcohol consumption has been shown in other settings to impede ART adherence [32, 33].

Fig. 1.

The social ecological model

At the interpersonal-level, HIV status disclosure and treatment supporters are associated with greater ART adherence, as well as greater support to get to HIV appointments [34, 35], highlighting social support as a determinant of care engagement. However, men in fishing communities are often single or separated from their families [36–38]. Thus, the networks of support available to them may be more limited than for men generally, or may have a negative influence on health seeking behavior. For example, a qualitative study in Uganda reported engagement in sexual risk behavior and alcohol consumption among fisherfolk was strongly influenced by the broader social environment [49].

At the normative-level, fishing villages have been characterized as having high levels of HIV stigma [36, 39], which has been associated with poor linkage to care and uptake of HIV testing among fisherfolk [40, 41], and poor adherence and retention more broadly in SSA [42, 43]. For example, experiencing prejudice and discrimination first hand from others (i.e., enacted stigma) emerged as a driver of treatment interruption among ART patients in Malawi [44]. The fear of, or anticipated, HIV stigma and internalized HIV stigma are also associated with poor retention and treatment adherence [45, 46], and may be especially influential for men in SSA [47, 48].

The physical and built environment of fishing villages poses a number of challenges to clinic access. Communities on islands are only accessible by boat [37, 49] and there may be few clinics with ART availability near island outreach testing [40, 41]. The fishing occupation itself includes frequent mobility, irregular work schedules, and an inconsistent cash-based income that varies based on seasonal fish yields [36, 49–51], all of which are likely to affect men’s care engagement. Finally, factors within the broader health system (e.g., drug stock out, quality of care, human resource availability and training), and societal-level policies affecting ART eligibility [18, 22, 52–58] have also been shown salient in their impact on access to care in SSA more broadly.

The overall goal of this study is to identify determinants of HIV clinic attendance and ART adherence among male fisherfolk on ART in Wakiso District, Uganda. Using a social ecological perspective, we examine individual, interpersonal, normative, and physical/built environment determinants to understand which factors and levels have the greatest effect, which can inform the development of targeted interventions to improve care engagement. We also explore the role of individual and interpersonal factors in moderating the associations between normative and physical/built environment factors and engagement outcomes.

Methods

This cross-sectional study included male fisherfolk receiving HIV care from one of seven ART clinics in Wakiso District, Uganda. We partnered with Mildmay Uganda, a nonprofit organization that provides free HIV services in several districts in Uganda, to purposively sample participants at HIV clinics between October 2016 and March 2017. Research sites included Mildmay “outreach clinics” (Zzinga, Kachanga, Kasenyi), governmental facilities (Bussi Health Center III [HCIII], Entebbe Hospital, Kigungu HCIII), and a private facility (Rapha Medical Center). The catchment areas of the clinics have a wide range—the smallest estimated to be 500 at Kachanga HCIII and the largest being 300,000 at Entebbe Hospital. Outreach clinics are Mildmay established and operated sites that provide HIV treatment once weekly or biweekly in communities with poor access to services. Three of our sites were on land (Entebbe Hospital, Kasenyi outreach, Kigungu HCIII), and four were on one of three islands on Lake Victoria (Bussi, Kachanga, Zzinga Islands). Each island was accessible only by boat and required approximately 30–60 minutes of travel time from the landing site.

Men were eligible to participate if they were: at least 18 years of age, fisherfolk by occupation, living with HIV, on ART, and enrolled in HIV care for at least 6 months (to assess ART adherence and retention). On ART clinic days, a male research assistant (RA) reviewed patient records for men presenting to care to pre-screen for eligibility. The RA approached potentially eligible men, confirmed eligibility using a verbal checklist, and obtained written informed consent for men agreeing to participate. To reach men not regularly attending the clinic, we invited men to meet with the RA to learn more about the study during routine reminder calls for clinic appointments, either at their next appointment or an agreed upon location. Of the 322 eligible men invited to participate, 22 declined for the following reasons: did not have time (n = 14), did not want to disclose sensitive information (n = 4), intoxicated (n = 1), hearing impairment (n = 1), wanted more compensation (n = 1), wanted to be interviewed with wife (n = 1). Participants completed an interviewer-administered questionnaire with the RA in a private space in or around the clinic using computer assisted personal interviewing (CAPI) programming in Questionnaire Development Software version 3.0. Men received 10,000 Ugandan Shillings (~ 3 USD) for their participation. Institutional Review Boards at San Diego State University and Makerere School of Public Health approved the study, as did the Uganda National Council for Science and Technology.

Measures

The individual-level factors measured and used in the present analysis include: age (continuous), monthly income (continuous, 1 unit equates to 50,000 shillings, ~ 14 USD), education (none vs. primary or secondary level), occupation (fishermen vs. other), months on ART (continuous, 1 unit equates to 6 months), and ART dosing (once daily vs. twice daily dosing). We used the World Health Organization’s Alcohol Use Disorders Identification Test (AUDIT) to assess alcohol risk [59], shown reliable among people living with HIV (PLHIV) in rural Uganda (α = 0.71) [60], using the total AUDIT score as a continuous variable for analysis (α in the present sample = 0.63). Internalized HIV stigma, or the endorsement of negative beliefs and feelings associated with HIV about themselves among PLHIV, was measured using Earnshaw’s HIV stigma framework measures (α = 0.79) (response options: 0–4, “never” to “very often”) [61]. We used Chesney’s AIDS Clinical Trials Group (ACTG) adherence questionnaire [62] to measure individual-level barriers to ART adherence. The scale includes reasons for missed antiretroviral (ARV) doses (response options 0–3, “never to often”). For analysis, we calculated the mean score of all items measuring individual-level barriers (e.g., forgot, side effects, lost pills) (α = 0.69).

Among interpersonal-level factors measured, relationship status was dichotomized into living together vs. single or separated (including men who reported being single, divorced, widowed, and married but separated). Social support was measured through the Social Support Scale [63], which contains 6 items measuring the participants’ access to emotional support (e.g., “You get chances to talk to someone you trust about your personal and family problems” and “You have people who care what happens to you”) and 4 items measuring access to instrumental support (e.g., “You get help when you need transportation” and “You get help when you are sick”). This scale (α = 0.67) was adapted from the Duke-UNC Functional Social Support Questionnaire [64], and asks participants how much of these types of social support they receive (response options 0–3, “never” to “as much as I would like”). The scale has been validated among PLHIV in rural Rwanda (α = 0.91) [65]. A single item assessed HIV disclosure, capturing the total number of people the participant disclosed their status to.

At the normative-level, we measured two dimensions of HIV stigma using Earnshaw’s measures [61], including enacted HIV stigma (perceived degree of prejudice and discrimination experienced by PLHIV) and anticipated HIV stigma (expectation among PLHIV that they will experience prejudice and discrimination in the future) (α = enacted: 0.65, anticipated: 0.65); response options for both: 0–4, “never” to “very often”.

Factors in the physical and built environment included mobility, measured with an item asking the participant if they have traveled/slept away in the prior year for work (yes/no), originating from a scale created for fisherfolk in the communities surrounding Lake Victoria in Kenya [51]. We accounted for the differences in the research site (i.e., HIV clinics the men were recruited in) by examining the differences for men accessing a clinic on an island vs. on land (referred to as a “landing site”). Average travel time to the HIV clinic was measured with one item asking how long in minutes it takes to get to the clinic (continuous, 1 unit equates 30 min). We used Chesney’s [62] AIDS Clinical Trials Group (ACTG) adherence questionnaire to measure structural-barriers to ART adherence by calculating the mean of items capturing travel and work-related barriers (response options 0–3, “never to often”), as described above (α = 0.70).

Our first outcome, missed HIV clinic visits, was measured using participants’ HIV clinic records. The RA extracted the number of clinic appointments scheduled in the prior year from participants’ HIV clinic cards, and recorded which of these visits were kept as scheduled vs. missed. We used participant self-report if data were missing or conflicted with participant’s clinic records. We operationalized missed visits as the proportion of HIV clinic visits missed in the prior year.

We use Chesney’s [62] AIDS Clinical Trials Group (ACTG) questionnaire to measure our second outcome, self-reported adherence. The scale includes items on participants’ regimen and number of pills taken per day, which we confirmed using participants’ HIV clinic cards. The ACTG asks for the number of doses missed over a 4-day recall period (yesterday, 1 day, 2 days, 3 days ago), and an additional four-items on adherence more broadly (i.e., last missed dose, frequency of missed doses on weekends, difficulty following schedule, and difficulty following special instructions). The ACTG has demonstrated construct validity in Uganda and similar settings [66]. We calculated a non-adherence index score which incorporates more items than the 4-day recall period (see data analysis section); the index score has been used in large clinical trials [67, 68] and shown to correlate with plasma HIV RNA more than the 4-day recall period alone [68].

Data Analysis Approach

In SPSS version 24, we used generalized linear modeling with a logit link (events/trial format) for our missed HIV clinic visits, and linear regression modeling non-adherence index score. To create the non-adherence index score, we followed Reynolds et al.’s methodology [68], which includes principle component analysis to approximate a latent variable. The index score was standardized on a scale from 0 to 100, with greater scores equating to worse adherence. For analysis, we used the square root method to transform the variable so that it met the assumptions for normality for linear regression. In preliminary bivariate models, we assessed the unadjusted associations between independent variables with missed clinic visits and non-adherence, respectively. Factors associated with the outcomes (p < 0.10) were included in multivariate models for our two outcomes. We also tested hypothesized interactions between independent variables—hypothesizing that older age, greater income, longer time on ART, greater social support, and HIV status disclosure, would mitigate the effects of HIV stigma and physical/built environment factors on our outcomes.

For missed clinic visits, Odds Ratios (OR) and Adjusted Odds Ratios (AOR) from bivariate and multivariate logistic regression analysis with 95% CI are presented. Betas (β) and standard errors (SE) are presented for bivariate and multivariate linear models assessing non-adherence index score. In the linear models, we include the eta squared (η²) to indicate the variance explained by each individual item, and we report R² to indicate the variance explained in adherence by the full model.

Results

Participant Characteristics

On average, men were 36.86 years old (SD = 8.58), and about half were living with a partner (49.00%), while the other half were single, divorced, widowed, or separated. Of the married men, approximately 30% had two or more wives. The average monthly income of men was 226,517 Ugandan Shillings (~ 62.12 USD, SD = 29.73). Men had been on ART for an average of nearly 2 years (22.90 months, SD = 13.50). The average AUDIT score was 4.45 (SD = 5.13), indicating low-risk drinking overall. The average time men traveled to the HIV clinic was 41.4 minutes (SD = 39.56), and 64.7% of men reported mobility for work. In the prior year, men had missed 39.00% of their scheduled clinic appointments. Men had taken 87.79% (SD = 21.46%) of their ARV doses over the 4-day recall period, making 31.00% “sub-optimally adherent” (< 95% of ARVs taken as prescribed). The average score on the non-adherence index score was 41.70 (SD = 79), with greater scores indicating worse adherence (range 0–100). See Table 1 detailed participant characteristics.

Table 1.

Participant characteristics, N = 300, Uganda 2016–2017

	% (n)/Mean	SD	Range
Individual-level factors
Age	36.86	8.58	20–70
Education
No schooling	202 (67.30%)
Primary level	84 (28.00%)
Secondary level	14 (4.70%)
Occupation (other vs. fishermen)
Fishermen	246 (82.00%)
Fish seller, cleaner, dryer	35 (11.67%)
Boat operator, repairer, loader	19 (6.33%)
Monthly income (in USD)	62.16	29.73	4.14–160.35
Months on ART	22.90	13.50	2–55
Number of pills prescribed per day
Once daily dosing	260 (86.67%)
Twice daily dosing	40 (13.33%)
AUDIT score	4.45	5.13	0.00–20.00
Internalized stigma	0.61	0.53	0.00–2.17
Individual-level barriers to ART adherence	0.35	0.24	0.00–1.42
Interpersonal-level factors
Marital status
Never married	30 (10.00%)
Divorced	78 (26.00%)
Widowed	16 (5.30%)
Married and separated most of the time	29 (9.70%)
Married and living together most of the time	147 (49.00%)
Social support	1.43	0.44	0.30–3.00
Total number of people disclosed HIV status to	2.04	1.59	0.00–8.00
Interpersonal-level factors
Anticipated stigma	0.86	0.53	0.00–2.83
Enacted stigma	0.62	0.48	0.00–3.00
Physical and built environment-level factors
Mobility (traveled/slept out of the community in prior 12 months)
Yes	194 (64.67%)
No	106 (35.33%)
Average travel time to HIV clinic (min)	41.40	39.56	2–180
Structural-level barriers to ART adherence	1.11	0.57	0.00–2.67
Research site
Island	175 (58.30%)
Landing site	125 (41.70%)
HIV care engagement outcomes
Non-adherence index score (greater score indicating worse ART adherence)	41.79	25.95	0–100%
Proportion of missed clinic appointments (prior year)	39.00%	28.79%	0–100%
ART adherence (% of pills taken over 4 day recall period)	87.79%	21.46%	0–100%

ART antiretroviral therapy

Social Ecological Model Predictors of Missed HIV Clinic Visits

Younger age, less time on ART, once-daily dosing, greater AUDIT score, not living with a partner, less social support, less HIV status disclosure, anticipated stigma, mobility, greater travel time to clinic, accessing care on a landing site, and structural barriers to ART adherence were associated with missed clinic visits (see Table 2 for detailed bivariate statistics).

Table 2.

Bivariate and multivariate models testing independent variables at each level of the social ecological model and missed HIV clinic visits, Uganda 2016–2017, N = 300

	Bivariate models			Multivariate model
	OR (95% CI)	χ2	p	AOR (95% CI)	χ2	p
Individual-level factors
Age	0.97 (0.96–0.99)	18.36	< 0.001	0.98 (0.97–0.99)	6.00	0.01
Monthly income (50,000 UGX/unit)	0.98 (0.94–1.03)	0.57	0.45
Education (no education vs. any education)
Any schooling	1.14 (0.92–1.40)	1.46	0.23
No schooling (reference)
Occupation (other vs. fishermen)
Fishermen	1.06 (0.82–1.37)	0.18	0.67
Other fisherfolk (reference)
Months on ART (6 month/unit)	0.92 (0.88–0.97)	11.71	0.001	1.07 (0.97–1.17)	1.72	0.19
Number of pills prescribed
Twice daily	0.72 (0.53–0.96)	4.95	0.03
Once daily (reference)
AUDIT score	1.02 (1.01–1.04)	5.88	0.02
Internalized stigma	1.14 (0.95–1.37)	1.85	0.17
Individual-level barriers to ART adherence (Scale:0-3)	1.36 (0.89–2.08)	2.01	0.16
Interpersonal-level factors
Marital status
Not married or separated	1.30 (1.06–1.58)	6.61	0.01	1.25 (1.01-1.54)	4.30	0.03
Married and living together (reference)
Interpersonal-level factors
Social support (Scale 0-3)	0.74 (0.58–0.93)	6.60	0.01
Total number of people HIV disclosed status to	0.93 (0.88–0.99)	5.07	0.02
Normative-level factors
Anticipated stigma (Scale: 0-4)	1.23 (1.02–1.49)	4.71	0.03	1.55 (1.05–2.29)	4.83	0.03
Enacted stigma (Scale: 0-4)	1.16 (0.95–1.42)	2.00	0.16
Physical and built environment-level factors
Mobility (traveled/slept out of the community in prior 12 months)
Yes	1.27 (1.03-1.57)	5.05	0.03
No (reference)
Average travel time to HIV clinic (30 min/unit)	1.12 (1.04–1.21)	8.49	0.004	1.11 (1.02-1.20)	5.93	0.02
Research site
Landing site	1.74 (1.42–2.13)	29.03	< 0.001	1.35 (1.08-1.67)	7.12	0.008
Island (ref)
Structural barriers to ART adherence (Scale: 0-3)	1.42 (1.19–1.71)	14.39	< 0.001	1.27 (1.04–1.56)	5.36	0.02
Interactions
Months on ART × anticipated HIV stigma	–	–	–	0.89 (0.81–0.98)	5.72	0.02
Mobility (traveled/slept out of the community in prior 12 months)

Bolded text is used to indicate statistically significant associations at the p < 0.05 level

^*Used to indicate statistically significant associations at the p < 0.10 level

OR odds ratio, AOR adjusted odds ratio, χ² Wald Chi square, AUDIT the alcohol use disorders identification test, ART antiretroviral treatment

In the final multivariate model, at the individual-level, younger men were more likely to miss clinic appointments than older men (AOR = 0.98, 95% CI 0.97–0.99, p = 0.01). At the interpersonal-level, men not living with a partner were 1.25 times more likely to report missed clinic visits in the prior year compared to men living with a partner (AOR = 1.25, 95% CI 1.01–1.54, p = 0.03). Anticipated HIV stigma remained positively associated with missed clinic appointments at the normative-level (AOR: 1.55, 95% CI 1.05–2.29, p = 0.03). At the physical/built environment level, for every 30-minute increase in time spent traveling to the clinic, men were 1.11 times more likely to report missed clinic visits (AOR: 1.11, 95% CI 1.02–1.20, p = 0.02). The odds of missed HIV clinic appointments were 1.35 times greater for men accessing HIV care at a landing site compared to on an island (AOR: 1.35, 95% CI 1.08–1.67, p = 0.008). Greater scores on the structural barriers to ART adherence scale were associated with increased odds of missed visits (AOR = 1.27, 95% CI 1.04–1.56, p = 0.02). Finally, there was a statistically significant interaction between months on ART and anticipated HIV stigma, with anticipated stigma having a greater effect on missed clinic visits for men more recently initiated on ART (AOR = 0.88, 95% CI 0.81–0.98, p = 0.02). See Table 2 for multivariate model statistics, and Fig. 2 for a graphic depiction the observed interaction.

Fig. 2.

Interaction between anticipated HIV stigma and months on antiretroviral treatment (ART) on the probability of missed HIV clinic visits in the prior year, controlling for covariates. Men newly initiated on ART are more likely to missed their HIV clinic visits due to anticipated HIV stigma compared to men on ART for a longer time period. The figure displays this difference for men on ART for 3 months compared to men on ART for 3 years

Social Ecological Predictors of Non-Adherence Index Score

Younger age, no education, less time on ART, once daily dosing, greater AUDIT score, individual-level barriers to ART adherence, less social support, having disclosed one’s status to less people, anticipated and enacted HIV stigma, mobility, structural barriers to ART adherence, and greater proportion of missed clinic visits in the prior year were associated with greater non-adherence in bivariate models (p < 0.10) (See Table 3 for detailed statistics).

Table 3.

Bivariate and multivariate models testing independent variables at levels of the social ecological model and ART non-adherence index score, Uganda 2016–2017, N = 300

	Bivariate models				Multivariate model
	β	Std. Error	p	η2	β	Std. Error	p	η2
Individual-level factors
Age	− 0.01	0.00	< 0.001	0.05	− 0.01	0.00	0.002	0.03
Monthly income (50,000 UGX/unit)	− 0.03	0.01	0.004	0.03	− 0.01	0.01	0.04	0.02
Education (no education vs. any education)
Any schooling	0.03	0.05	0.62	0.001
No schooling (reference)
Occupation (other vs. fishermen)
Fishermen	0.07	0.06	0.26	0.004
Other fisherfolk (reference)
Months on ART (6 month/unit)	− 0.02	0.01	0.08*	0.01
Number of pills prescribed
Twice daily	− 0.13	0.07	0.05*	0.01
Once daily (reference)
AUDIT score	0.01	0.00	0.03	0.02
Internalized stigma	0.04	0.04	0.32	0.003
Individual-level barriers to ART adherence	0.23	0.88	0.009	0.02	0.14	0.09	0.09*	0.01
Interpersonal-level factors
Marital status
Not married or separated	0.03	0.05	0.47	0.002
Married and living together (reference)
Social support	− 0.23	0.05	< 0.001	0.06
Total number of people HIV disclosed status to Community-level factors	− 0.38	0.02	0.009	0.02	− 0.02	0.01	0.08*	0.01
Anticipated stigma	0.17	0.04	< 0.001	0.05	0.10	0.04	0.03	0.02
Enacted stigma	0.19	0.05	< 0.001	0.05	0.11	0.05	0.03	0.02
Physical and built environment-level factors
Mobility (traveled/slept out of the community in prior 12 months)
Yes	0.15	0.05	0.002	0.03
No (reference)
Average travel time to HIV clinic (30 min/unit)	0.01	0.02	0.49	0.002
Research site
Landing site	0.07	0.05	0.12	0.01
Island (ref)
Structural barriers to ART adherence	0.14	0.04	< 0.001	0.05
Proportion of missed clinic appointments (prior year)	0.27	0.08	0.001	0.04	0.06	0.07	0.40	0.003

Bolded text is used to indicate statistically significant associations at the p < 0.05 level

B Beta, Std. error standard error, AUDIT the alcohol use disorders identification test; ART antiretroviral treatment

^*Used to indicate statistically significant associations at the p < 0.10 level

In the multivariate model (Table 3), at the individual-level, younger men (β = − 0.01, SE = 0.00, p = 0.002, η² = 0.03) and men with lower income (β = − 0.02, SE = 0.01, p = 0.04, η² = 0.02) had worse adherence. Greater scores on the individual-level barriers to adherence scale (β = 0.14, SE = 0.09, η² = 0.01) were also associated with non-adherence, but only marginally (p = 0.09). The only interpersonal-level factor that remained in the model was HIV disclosure (β = − 0.02, SE = 0.01, η² = 0.01), though the association was only marginally significant (p = 0.08). Finally, at the normative-level, greater scores on the anticipated HIV stigma scale (β = 0.10, SE = 0.04, p = 0.03, η² = 0.02) and enacted HIV stigma scale (β = 0.11, SE = 0.05, p = 0.03, η² = 0.02) were associated with greater non-adherence. No barriers in the physical and built environment remained in the final model, and no interactions were identified between factors. Taken together, the total model explained 16% of variation in non-adherence (R² = 0.16).

Discussion

This study is the first to our knowledge to assess HIV clinic attendance and ART adherence among fishermen on ART [69]. We identified factors at multiple levels of the SEM [12–14] associated with HIV care engagement for men (see Fig. 3 for a graphic depiction of our study’s primary findings mapped onto the SEM). Individual, interpersonal, normative, and physical/built environment-level factors were associated with missed clinic visits. For ART adherence, individual and normative-level factors had the greatest effect. These findings can inform the development of differentiated care models [70] tailored to the needs of fishermen. Our study justifies the need for such efforts by highlighting less than optimal engagement among men; a third of men reported sub-optimal ART adherence (31%), and men missed a significant portion (39%) of their HIV clinic visits over the prior year.

Fig. 3.

The study’s primary findings by level of the Social Ecological Model. Note “Missed Clinic Visits” and “ART Non-Adherence” are the study’s primary outcomes; the factors listed below each are the independent variables found to be associated (p < 0.05) with the outcomes in multivariate analysis

Though HIV stigma was reportedly low in our sample, it was still influential for both clinic attendance and ART adherence for men. Anticipated HIV stigma had the greatest effect compared to any other independent variable on missed visits, and both enacted and anticipated HIV stigma predicted ART non-adherence. We also found anticipated HIV stigma had a greater effect on missed clinic appointments among men more recently initiated on ART (interaction displayed in Fig. 2). This finding supports other research that shows after the initiation of ART, HIV stigma may decrease over time [71]. Thus, strategies for HIV stigma reduction should occur soon after diagnosis and treatment initiation to prevent men from dropping out of care. These findings are in line with that of prior studies that report HIV stigma reduced men’s uptake of HIV testing and timely linkage to HIV care in fishing villages [29, 30], and contributes to a large literature in SSA supporting HIV stigma’s negative effect on engagement across the care continuum [31–35], which may be particularly true for men in SSA [36–40].

In the broader literature, social support has been shown protective against HIV stigma [72–74], and interventions that enhance patients’ support mechanisms report promising effects both on stigma and care engagement [75, 76]. A recent study in Kenyan fishing villages reported improvements in care retention in an evaluation of “microclinic interventions,” which included 10 sessions of social support and education, and bolstered social capital and reduced HIV stigma by using patient-defined support networks [75]. While social support did not remain significant in our final models, we did find men living with a partner had better clinic attendance compared to those not living with a partner, perhaps because their partners served as treatment supporters. With half of our sample being unmarried or separated, the integration of a male-to-male peer support program may be beneficial in this setting, and could have the additional benefit of HIV stigma reduction. Community-based ART delivery models, such as community adherence groups (CAGs) and adherence clubs, might also be considered in this context; these service delivery models have been shown to reduce HIV stigma (89, 90) by leveraging or expanding patient’s existing support networks (91–93), while reducing structural barriers to care by bringing services closer to the community.

The greatest number of factors associated with clinic attendance fell within the physical/built environment. However, clinic attendance was not associated with ART adherence in the final model and factors in the physical/built environment did not associate with ART adherence. Though research is needed to understand what other negative effects men’s poor clinic attendance may have, this finding suggests men may be able to maintain optimal treatment adherence without attending all scheduled clinic appointments. Research to evaluate and optimize strategies that reduce the number of visits men need to maintain optimal adherence would not only benefit men, but would also reduce burden on the overall health system. Though our data cannot speak to this, the use of multi-month refills for patients may be explaining this finding; future research should assess whether multi-month prescriptions are being optimally utilized with fishermen, and whether they can be expanded on to further reduce clinic burden and the time and monetary costs associated with transportation for stable patients, as has been demonstrated in other studies [77–79].

While our study provides support for the effect of multi-level factors on men’s engagement in HIV care, this study provides only a narrow view of the potential factors that affect men’s health seeking behavior. The observed effect sizes were small, and some variables shown important for HIV care engagement in this context in qualitative research, such as mobility [49], did not remain in our final models. However, we found men accessing care on a landing site were more mobile than men accessing care on an island, which likely explains why they had worse clinic attendance than those accessing care on an island. Future research is needed to replicate and expand on our findings, and to include policy-environment factors, which were not assessed in this study.

Our clinic attendance measure was collected through clinic records and corroborated by self-report, which may be subject to data entry errors or missing information. Though Chesney’s ACTG questionnaire [62] has been shown predictive of viral load [68], our measure of adherence would have been strengthened through objective measures of pill count or viral load samples. We attempted to collect viral load data from patient records, but found the data too incomplete for use. Future research should explore these constructs against biological outcomes, and use longitudinal designs to infer temporality between factors measured. While most of our measures were validated or previously used in East or Southern African settings and piloted for comprehension, they were not formally validated for use with this specific population. The generalizability of our study is limited by use of non-random sampling directly from the clinic, which may have resulted in the exclusion of men with the greatest barriers to engagement. Future research should focus on engaging men lost-to-follow up, as well as younger men and men of lower income who, consistent with the literature [23–28], had more trouble with engagement. Nevertheless, this study provided insight into multi-level barriers to care for fisherfolk on ART, who struggled with engagement despite being sampled from the clinic.

Conclusion

In line with global HIV guidelines [80, 81], Uganda is moving towards the scale-up of universal antiretroviral treatment (ART) for all [7] making efforts to engage and retain PLHIV along the care continuum a growing priority. Understanding factors influencing engagement in care among most-at-risk populations is especially needed, as achieving population viral load suppression in high prevalence settings would optimize the effectiveness of treatment as prevention [82, 83]. This study begins to fill a gap in the literature on determinants of HIV care engagement among Ugandan fishermen, and highlights the need for multi-level interventions to improve men’s care engagement. We recommend future research explore the integration of social support, HIV stigma reduction, and strategies to reduce the number of visits needed for HIV care (e.g., multi-month refills, CAGs) into differentiated models of care for fishermen on ART.