Beyond viral suppression: Quality of life among stable ART clients in a differentiated service delivery intervention in Tanzania

The T&T project is hosted by the Catholic Diocese of Shinyanga which covers both Shinyanga and Simiyu regions in north-western Tanzania. Besides Shinyanga urban, Kahama urban, and Bariadi districts, the regions are largely rural. Project sites are four HIV care and treatment centers (CTC) referred to as hubs, two hubs each in the Shinyanga (Ngokolo and Bugisi) and Simiyu (Songambele and Mwamapala) regions. Standard of HIV care in Tanzania is clinic-based and includes one clinic visit every one to three months for consultation, health screening, routine labs and ART refill. DSD in ART clubs was rolled out in the T&T project from July 2018, details of which have been described elsewhere [38]. Briefly, ART clubs are CHW managed groups of 15–30 clients living within the same locality who meet every 3 months for routine health screening and ART distribution. Club members have a clinical consultation every year. Data were collected from May to August 2019. Participants were recruited at the two hubs in the Shinyanga region and their related ART clubs. Eligibility criteria included being adults ≥ 18 years and stable on ART according to the Tanzanian guidelines: on ART 1^st line regimen ≥ 6-months, viral load < 50 copies/ml, and no current chronic illness [39]. At the hubs, participants were randomly sampled from a list of all eligible clients who had a clinic appointments within the data collection period. Eligible participants were approached as they attended clinic appointments. All clubs that had a meeting during the data collection period and were at least 6 months or older, were visited. At the clubs, all members were approached as stability was an eligibility criterion for DSD participation. Those clients who gave written consent, completed the interviewer-administered questionnaire.

We used an HIV-specific HRQoL tool that has been validated for the low literacy Swahili population, the Kiswahili translation of the Functional Assessment of HIV Infection (FAHI) [40]. Outcomes of interest were the total and domain-specific FAHI scores. The FAHI is a 47-item tool with five domains namely physical wellbeing (PWB) – 10 items, emotional wellbeing (EWB) – 10 items, functional & global wellbeing (FGWB) – 13 items, social wellbeing (SWB)-8 items, and cognitive functioning (CF) – 3 items [41]. Scores ranged for each item between 0 and 4. We derived (a) domain scores by summing respective item scores – ranges for PWB and EWB were 0 to 40, FGWB 0 to 52, SWB 0 to 32, and CF 0 to12; (b) total FAHI scores by summing all five domain scores – ranging between 44 and 176—note that three items in the PWB domain were not scored as recommended by the FAHI scoring document [42]; (c) FAHI proportional score by calculating each individual score as a proportion of the maximum possible total or domain scores; and (d) a dichotomous (satisfactory/less than satisfactory FAHI HRQoL) variable for total and domains. We considered a score in the highest quintile i.e. ≥ 80% of FAHI total or domain scores as satisfactory to capture all participants who report at least ≥ 4 on the 5-point FAHI tool. This represents all those who report at least above “Somewhat” (i.e. 3 – the midpoint) for all items in all domains of the FAHI instrument.

Secondary outcomes were factors associated with satisfactory HRQoL and domain scores. Three categories of additional data were collected to assess these factors: socio-demographic (location, sex, age, educational level, marital status, employment status, and income level), HIV care (duration on ART, CD4 count at ART start and recency of viral load result) and service access (location, time spent during clinic visit/club meeting[wait time], time spent traveling to clinic/club[travel time] and frequency of service delivery). Data entry, collation, and cleaning were done using EpiData [43].

Our sample size calculation was based on EQ-index scores and extrapolated to proportional FAHI scores. We assumed a difference in proportional scores of 0.10 (0.80 to 0.90) between the clinic and DSD participants, a standard deviation of 0.40 as determined by Louwagie et al. in South Africa, and a 10% refusal rate requiring thus a minimum of 542 participants overall with 271 participants per service delivery group to have 80% power to reject the null hypothesis of no difference [1].

Categorical variables were presented as percentages and continuous variables as means (± standard deviation) or medians (± interquartile range) as appropriate. Comparisons between clinic and DSD participants were done using Mann Whitney or Kruskal Wallis tests. Association between socio-demographic, HIV care, and service access factors and satisfactory FAHI HRQoL were examined using logistic regression. Sex, age, marital status, and variables showing significant bivariate association at the p-value of < 0.1 were included in the multivariable model. A 3-step hierarchical multiple linear regression was used to quantify the contribution of the three-factor categories to the variance of FAHI scores observed. Socio-demographic variables were entered in the model in the first step, followed by HIV care variables and lastly, service access variables.

We examined variables for multicollinearity using tolerance values and variance inflation factor (VIF) statistics. We generated a Receiver Operative Characteristic (ROC) i.e. area under the curve (AUC) to test the discriminative ability of the model (with all covariates included) to categorize observations as satisfactory/less than satisfactory HRQoL. We assessed the 33 and 28 missing observations dropped from the clinic and DSD in step 3 hierarchical linear models, respectively, to observe any significant differences in mean FAHItotal. All analyses were performed using STATA software version 16.0.

Ethical approval for the study was obtained from the National Institute for Medical Research (NIMR; approval number NIMR/HQ/R.8c/Vol. I/674).

Of 667 PLHIV approached to participate, 641 consented to participate (response rate of 96.1%), and 629 were included in the final analysis (12 excluded due to missing data). While the overall majority of participants were female (63%), there were significantly more men in clinic-based care compared to DSD, and DSD participants were also significantly older (see Table 1). The mean numbers of years-on-ART and mean CD4 count at ART start were significantly longer (4.9 vs 4.1 years, p < 0.001) and higher (398.1 vs 341.4 cells/mm³, p < 0.001) for DSD participants. They also spent shorter time on travel (84.7 vs 34.3 min, p < 0.001) and during club meetings (140.2 vs 83.8 min, p < 0.001). There were more DSD participants in the urban area (60.6% vs 39.4%). Table 1 provides details on the characteristics of study participants according to the service delivery model.

Clinic and DSD participants show comparable mean HRQoL scores across domains with only slight differences in the physical and emotional wellbeing domains (36.4 vs 35.5, max-40 p < 0.01) and (32.1 vs 32.8, max-40 p < 0.05) (Fig. 1a). No differences were observed in satisfactory HRQoL percentages across domains except for FGWB where more clinic participants revealed satisfactory HRQoL as compared to DSD. Satisfactory HRQoL overall was highest in the CF domain (89.2 vs 93.6) and lowest in the EWB (68.8 vs 68.5) and SWB (74.1 vs71.7) domains (Fig. 1b).

Satisfactory overall HRQoL was associated with being male ((odds ratio 2.59, 95% confidence interval 1.36–4.92) among clinic participants and with living in an urban setting (4.72, 1.70–13.1) in DSD care (see—Table 2). Less than satisfactory HRQoL was observed with increasing age among clinic participants, and with increasing income, and increased meeting duration among DSD participants. HIV care factors were generally not associated with satisfactory overall HRQoL.

Compared to being single, satisfactory HRQoL was associated with being married in the PWB domain for both clinic and DSD. Being married or separated, divorced, or widowed was positively associated with satisfactory HRQoL for both clinic and DSD participants in the SWB domain and only among DSD participant in the FGWB domain. Living in an urban area was significantly associated with satisfactory HRQoL for both clinic and DSD participants in the EWB domains and only among DSD participants in the SWB and PWB domains. Across domains, declining age was generally not associated with satisfactory HRQoL. Significantly less satisfactory HRQoL was only seen among clinic participants aged over 65 years in the PWB and FGWB domains. Surprisingly, less satisfactory HRQoL was linked with increased income levels in the EWB domain among clinic and DSD participants. Generally, HIV care factors were not associated with satisfactory HRQoL. Among service access factors, DSD participants alone reported less than satisfactory HRQoL for spending longer time during service access in the PWB, EWB, and SWB domains (Table 2 and Additional file 1 [for additional results of FGWB & CF domains]).

Table 3 shows the contribution of sociodemographic, HIV care, and service access factors to the variance observed in HRQoL scores. The analyses revealed that among clinic participants, the variance in overall HRQoL score FAHI total explainable by sociodemographic variables in the first step was 10.2%. The addition of HIV care variables in the second step increased the variance explained to 14.5%. Finally, service access variables in the third step brought the total to 14.9%. For DSD participants, the variance explained was 22.9%, 28.9%, and 43.5% in the first, second, and third steps, respectively. Across all domains, the variance in HRQoL explainable by the 3-step hierarchical model for clinic participants was modest (see Table 3). The highest was reported in the EWB domain i.e. 8.5%, 11.4%, and 18.5%, and lowest in the CF domain i.e. 5.9%, 9.1%, and 9.8%, respectively. A much higher proportion of variance was explained in overall FAHI 43.5%, PWB 30.2%, EWB 53.4%, and SWB 35.1% among DSD participants. Additional file 2 shows the details of the hierarchical linear regression with coefficients of all covariate in each step.

Our study is among few HRQoL studies conducted recently in SSA in the era of DSD. It provides useful insights into factors influencing HRQoL in an African population. Our participants were drawn from different geographical settings that mimic the reality of our population and generated valuable information about the impact of DSD in these settings. Though observational with known biases, the analytical design of our study allowed for comparisons that produced a rich resource useful for informing implementation and policy.

Clinic participants were selected for stability as defined by the Tanzanian guideline at the time of data collection while DSD participants were assumed to be stable. This might have biased our results in favor of clinic participants, however, viral load-related variables were similar in both groups and not independently associated with HRQoL in our study.

The project sites were mission clinics which may limit the generalizability of our findings. However, we might expect that larger differences in HRQoL scores would be found when comparing DSD and clinics outside the mission hospital setting, as better funding and service which characterize our setting likely obscured the effect of DSD.