We conducted a cross-sectional study between May 2017 and June 2017. The Good Health for Women Project (GHWP) is in Mengo Kisenyi, a densely populated slum in the central division of Kampala, the capital city of Uganda. The GHWP clinic was established in 2008 by Medical Research Council/Uganda Virus Research Institute (MRC/UVRI) in collaboration with the Uganda Ministry of Health. The GHWP was set up to study the epidemiology of HIV and other STIs among FSWs, and to provide free enhanced HIV prevention services to this key population. These include treatment of common illnesses, comprehensive HIV care, management of other STIs, reproductive health services, condom promotion and distribution and alcohol risk reduction counselling. These services are delivered at 3-monthly visits and according to the clients’ demand. Although GHWP started offering ART in 2013, ‘test and treat’ began in August 2014; a government of Uganda policy where HIV-positive FSWs start ART immediately irrespective of cluster of differentiation 4 (CD4) count.¹¹

Good Health for Women Project patients who are FSWs are defined as women who engage in sex with men for money and other favours. Female sex workers are recruited to the clinic at night through a peer-led approach from their operating hotspots (e.g. bars, brothels, nightclubs, streets, eating places, lodges and guesthouses). Greater details on the recruitment of FSWs to the GHWP clinic are described elsewhere.⁴ Eligibility criteria were being 18 years old and above, accessing ART from GHWP and having been on ART for at least 6 months. Participants were enrolled consecutively as they came either for their routine clinic visits or for any other reason.

Prior to initiating ART, baseline CD4 counts and pre-ART adherence counselling was performed on confirmed HIV-positive clients. Antiretroviral therapy was dispensed first for 14 days, and then 1-monthly, 2-monthly, and 3-monthly refills were given only when the client was determined to be stable. A fixed-dose combination of tenofovir disoproxil fumarate (TDF), lamivudine (3TC) and efavirenz (EFV) is used as the first-line regimen, unless otherwise indicated. This was a single pill (TRUVADA) used once a day by adults. For follow-up, viral load testing was performed 6 months after ART initiation, repeated yearly if viral suppression had been achieved, and, if not, 4–6 months following Uganda Ministry of Health Guidelines.²⁷ Ongoing adherence counselling was provided for the clients at every clinic visit as appropriate. Specifically, those clients for whom viral suppression was not achieved, at least three sessions of intensive adherence counselling were performed monthly by the nurse counsellor until viral suppression was achieved. Screening for and management of opportunistic infections and other STI was performed at every contact with the clients.

Our trained study staff collected data using pre-tested interviewer-administered tools and data extraction from participant files. Pretesting of the data collection tool was performed in the first week of data collection to determine suitability, face validity, comprehension, and to achieve a common understanding of the tool among the nurse counsellors. Quality control checks were performed for each completed case report form to ensure completeness, accuracy and correctness.

As part of routine clinical care, blood samples were collected to assess HIV viral load and CD4+ T-cell count. Baseline CD4 + T-cell counts were performed on ART initiation and viral load (VL) was performed at least 6 months from ART initiation. Baseline CD4 + T-cell counts were performed on plasma using Multi-test Trucount tube CD4, CD8 and CD3-Lyse no wash (Becton D Biosciences, United States). Viral load testing was performed on plasma using an Abbott Real-time HIV-1 PCR assay (Abbott Park, Illinois, US) from the Uganda Central Public Health Laboratories.

We entered data into epi-data, cleaned and exported to STATA 15.0 (StataCorp, College Station, TX, United States) for analysis. We categorised and summarised participants’ characteristics by counts and percentages. We analysed continuous variables using appropriate measures of central tendency, means or medians, and their relevant measures of dispersion, standard deviations and interquartile ranges, respectively. We determined the proportion of non-adherent FSWs who scored ≥ 95% at assessment divided by the total number included in the analysis and expressed as a percentage. We used logistic regression to identify factors associated with adherence at bivariate analysis. We selected variables with statistical significance p < 0.1 on log likelihood ratio tests (LRT) for multivariable logistic regression. We maintained ‘alcohol use’ variable in the final model because literature shows that it is a strong predictor of ART adherence. We presented results adjusted odds ratios (AOR) with p-values and 95% confidence intervals (CI). The p-values < 0.05 were considered significant.

Ethical clearance to conduct this study was obtained from Uganda Virus Research Institute Research Ethics Committee (No. MRCU/07/815) and Uganda National Council for Science and Technology (No. HS 364). Our study presented more than minimal risks including potential unintended disclosure of HIV and FSW status to the community during home visits. Noteworthy, we performed a secondary analysis of clinical information without any invasive procedures that diminished some risks. Furthermore, we obtained written informed consent from all participants, and all data were de-identified prior to analysis.

The FSWs had a mean age of study participants of 32 years (standard deviation [s.d.]: ±6.5), with 56.2% falling within the 25–35 years age bracket. A total of 56.6% had attained at least primary-level education. While only 37.6% were separated or divorced, most FSWs (88.5%) reported disclosing their HIV status to a significant other (Table 1). Over half (56.2%) reported consuming alcohol over the last 3 months. The majority (90.3%) reported owning a mobile phone.

Clinically, most FSWs (87.6%) had initiated ART early at WHO stage 1 and had a mean duration on ART of 22 months (s.d.: ±8). Half FSWs had a baseline CD4+ T-cell count above 532 cells/mm³ (interquartile range [IQR]: 390–692). The majority (96.9%) of participants were on first-line regimen (TDF, 3TC and EFV). Viral suppression (< 1000 copies/mL) was observed in 89.4% of participants.

Overall, 59.2% of participants were adherent to ART. Being away from home was the most important reason for missing pills, 40.8% (n = 120) (Figure 2). In the bivariate analysis, adherence to ART was more likely among participants who had social support compared to those who did not (odds ratio [OR]: 1.85; 95% CI: 1.02, 3.34; p = 0.041), those who owned a phone at the time of the study (OR: 3.53; 95% CI: 1.38, 9.05; p = 0.006) and those who reported possession of a medicine reminder tool (OR: 2.35; 95% CI: 1.23, 4.40; p = 0.009). A 10-year increase in age of the client was associated with increased odds of adherence (OR: 1.59; 95% CI: 1.03, 2.43; p = 0.031).

In multivariable analysis, participants who owned a phone at the time of study were more likely to adhere to ART than those who did not (AOR: 2.90; 95% CI: 1.07, 7.88) (Table 2). There was an increased odds of adherence to ART for each 10-year increase in age (AOR 1.60; 95% CI: 1.00, 2.60). Compared to married women, widowed participants were less likely to adhere to ART (AOR: 0.22; 95% CI: 0.05, 0.87). The finding on alcohol use may have suffered social desirability bias from, especially, among drinkers of alcohol.

Firstly, our study discovered that adherence to ART was suboptimal among FSWs in the GHWP, with only 3 out of 5 attaining ≥ 95% level. Secondly, adherence to ART was independently associated with ownership of a mobile phone, older age by 10-year bands and marital status. Specifically, widows were less adherent to ART. Alcohol use was not associated with adherence in this key population.

Although our study had some strengths, these findings are better understood with regard to its limitations. We had a relatively large sample size of FSWs and conducted this study in a typical urban slum setting for FSW providing for generalisable findings. However, selection bias could have occurred as we enrolled only those FSWs who came to the clinic for their visits. As a result, we may have overestimated adherence to ART among the participants. Firstly, our measurement of adherence to ART was based on patients’ self-reports of missed pills, which are subject to recall and social desirability biases, potentially resulting in overestimated adherence. Good enough, self-reported adherence correlates well with viral load changes, is cheaper and more practical in low-income settings.^28,29,30 A recent article that compared adherence to ART by appointment keeping, 7-day recall versus 30-day recall and found that only 30-day self-report measure was associated with a lower risk of virological failure (VF) (adjusted hazard ratio (aHR) = 0.14, 95% CI: 0.05–0.37).³¹ Another older study compared 3-day caregiver recall, 30-day visual analogue scale (VAS), Medication Event Monitoring System (MEMS), unannounced pill counts (UPC) and liquid formulation weights were collected monthly over a 1-year period. This study reported that 3-day caregiver recall; VAS and MEMS were found to be positively associated with increasing Nevirapine (NVP) concentration in hair, although associations were not statistically significant.³² However, recall approaches to measuring adherence are not without limitations. For example, a study on patient health literacy and adherence to ART demonstrated that overall, 57% of patients reported a 30-day recall of 100% adherence.³³ Finally, we opted for the most recent viral load results instead of real-time viral load tests that could have overestimated adherence. Given our cross-sectional design, we recognise the limitations of correlating self-reported results with patient important outcomes.

Adherence among FSWs was relatively low compared to findings from a systematic review of studies largely from higher-income countries at 76%.²⁶ Evidence from other studies reported FSW adherence to ART at 67%³⁴ and 81%.³⁵ In addition, our study reported lower adherence to ART compared to the general populations in SSA.^18,36 Low proportions of ART adherence among FSWs have also been reported in India, in a study that estimated adherence at 48% using a combined measure of pill counts and self-reports.²⁵

The major reason for missing pills by participants in our study was being away from home. Most FSWs do not usually operate from their areas of residence, where they are known, but rather go to other hot spots within the city where they are less likely to be known. It is possible that sometimes they either forget carrying their pills with them or spend more days away from home than they had anticipated.

We found a strong association between owning a mobile phone and adherence. This is consistent with the fact that most of the study participants who reported owning a mobile phone used the phone alarm or clocks as their medicine reminder tool. This is also consistent with findings from a qualitative study conducted in Lesotho that showed that mobile phone alarms and clocks are major facilitators of adherence among people living with HIV.³⁷ Our findings build a case for the use of mobile phone technologies to enhance adherence to ART, as evidenced by the various studies.^{38,39,40,41,42,43,44}

Our study also showed that old age was associated with ART adherence, a finding consistent with what has been found among the general population^18,36,45,46 and other key populations.^47,48 Although not examined in this study, it is possible that older FSWs have lived with HIV for a longer time and therefore have more ART experience than their younger counterparts. This finding suggests the need for targeted interventions to improve adherence among the young FSWs. Studies have shown that younger FSWs have unequal footing when negotiating in their communities. They are usually street-based, poorly educated, occupying low-skilled and poorly paying jobs, and are always in conflict with their older counterparts because of competition for clients.^49,50

In addition, our study found lower odds of adherence to ART among FSWs who reported being widowed. This finding is consistent with results from a study among the general population of people living with HIV in Dakar, Senegal.⁵¹ Being widowed and a single mother who takes care of the children without a hand of the husband (many times originally the sole breadwinner), has been shown to be associated with depression.⁵² This, coupled with living in settings where FSWs are discriminated against and criminalised, could possibly affect adherence to ART.

Contrary to our expectation and evidence from other studies,^16,53,54 data from our study did not show an association between alcohol use and adherence to ART. However, our finding is similar to that from a study among FSWs in India.²⁵ Over half of the FSWs in our study reported alcohol use, similar to findings among FSWs in another study in Malawi⁵⁵ and one in Uganda.⁶ Studies have documented the use of alcohol and other substances by FSWs as a way of coping with demands of their work.^56,57 Our measurement of alcohol use was based on a Yes/No response to alcohol use in the last 3 months and not on the standard tool like audit score or the CAGE. This could partly explain this contradictory finding. Nonetheless, alcohol risk reduction counselling is necessary intervention for FSWs. Studies have indicated that alcohol consumption is not only associated with high-risk sexual behaviour^6,56 but may also influence the survival of HIV-infected patients⁵⁸ without altering the effectiveness of ART.⁵⁹