Predictors of the observed high prevalence of loss to follow-up in ART-experienced adult PLHIV: a retrospective longitudinal cohort study in the Tanga Region, Tanzania

This study was a retrospective longitudinal cohort analysis of adult PLHIV attending care and treatment clinics within eight districts supported by Amref Health Africa Tanzania in the Tanga region from October 2018 to December 2020. The HIV prevalence in Tanga is 5% (3.7 males: 6.2 females), a bit higher than the country's HIV prevalence of 4.9% (3.4% males and 6.3% females [17].

The participants were adult PLHIV aged 15 years and above on ART and attended the clinic at least once after ART initiation. Data on PLHIV on ART from October 2018 to December 2020 was extracted from the Care and Treatment Clinics form number 2 (CTC2) database, whereby all PLHIV below 15 years of age were excluded.

The following variables were considered as possible determinants of loss to follow-up; sex (male, female); age (15–24; 25–34; 35–44; 45–54; 55 +); BMI (< 18.5, 18.5–< 25, 25–< 30, 30 +); WHO HIV disease stage (I, II, III, IV); district of residence (Handeni, Kilindi, Korogwe, Lushoto, Mkinga, Muheza, Pangani and Tanga); pregnant state (yes, no); marital status (Cohabiting, Divorced, Married, single and Widow); ART duration (≤ 1, > 1 years); adherence (good, poor); Viral load suppression (suppressed, non-suppressed); ART regimen (first line, second line); DTG based drug (yes, no).

This study aimed to identify factors associated with LTFU using two-year data retrieved from the CTC2 database. In this study, LTFU was defined as missing clinic visits for 90 days or more consecutively from the last clinic attendance for drug refills and not confirmed dead or transferred out [18]. Confirmed deceased PLHIV and those transferred to other health facilities not supported by Amref were censored. Confirmed Transfer out were those PLHIV who have initiated ART and later received official transfer letter to continue ART services to other non-Amref supported health facilities. Death of PLHIV was ascertained when it was documented in the ART care and treatment follow-up sheet by health care professionals. In this study, good ART adherence is considered when ≥ 95% of the medication is taken as prescribed, meaning three or fewer doses are missed. In contrast, poor adherence is considered when ≤ 85% of the medication is not taken as prescribed, i.e., nine or more doses are missed [18].

We did a descriptive statistical analysis to provide basic information about the variables of interest in our study to highlight potential relationships between the variables. Numbers and frequencies were used to describe categorical variables, while means/medians and interquartile ranges were used to describe continuous variables. In addition, we used bivariate analysis to ascertain the degree of association between independent and dependent variables in various periods.

We used Cox proportional hazard regression models to describe the hazard ratio of participants lost to follow-up with an associated 95% confidence interval (CI). We used Kaplan–Meier curves to estimate the cumulative incidence of loss to follow-up. The death date and last clinic visit marked the end time of follow-up. Potential risk factors that were statistically significant at a p-value of 0.2 or less in univariate analysis were included as potential confounders in multivariate models on a two-sided test with a significant p-value less than or equal to 0.05. We used the missing indicator method to handle missing data. Statistical analysis was performed using the SAS^® statistical software package, Release 9.3 (Cary, NC, USA).

In the 2-year program review period, the median age of the adult PLHIV was 42.6 (interquartile range, IQR: 35–52). The median CD4 count, cell/mm³ 478 (IQR: 286–714). As shown in Tables 1 and 2, a total of 57,173 adults PLHIV ever on ART, females were more (69.21%), the majority (28.94%) lived in the Tanga district, and most (76.81%) had received ART for more than one year and the majority (97.89%) on first-line regimen. The majority (97.03%) had good drug adherence, and 75.08% of the participants were virally suppressed. Results showed that most (99.67%) never had a TB history, and 77.43% of the participants were on a DTG-based regimen. More than half (52.37%) of the participants were married, and only 2.19% of women were pregnant. The majority (73.5) % (n = 42,062) were retained in the study, while 26.43% (n = 15,111) were loss to follow-up.

As depicted in Table 1, a slightly greater proportion of males (26.79%) than females (26.27%) were LTFU. Also, the age group of 15–24 years (33.64%) accounted for the highest proportion of LTFU, followed by the age group of 25–34 years (32.16%), respectively, compared to other age groups. Furthermore, pregnant women (37.14%) accounted for a significant proportion of LTFU compared to non-pregnant. A high proportion of patients from Kilindi (34.68%), Handeni (30.02), and, Muheza (29.96%) districts accounted for LTFU compared to the rest of the districts. In addition, single participants (29.67%) accounted for the highest proportion of LTFU compared to their counterparts.

Table 2 describes clinical factors and their association with LTFU. A slightly greater proportion (26.94%) of participants on the second-line regimen (drug type) were LTFU compared to those on first-line treatment (25.75%). Results showed that PLHIV on ART for more than 1 year were more likely (21.58%) to be LTFU than those on treatment for less than one year, but the difference was statistically significant. Of the 2,924 (19.35%) participants who did the viral load test, 14.90% who were virally suppressed were LTFU, compared to 31.78% who were virally non-suppressed. A greater proportion of participants (69.18%), and 69.15%, with BMI (kg/m²) < 18 and 25–< 30, respectively, were LTFU than those with a BMI between 18.5 – < 25 and 30 +. In addition, more participants in WHO stage I (30.72%) were LTFU than WHO stage II, III, and IV.

We used a univariate logistic regression model to ascertain the independent variables contributing to LTFU. We included the variables in univariate analysis with a p < 0.05 in the multivariate logistic regression model (Table 3). Age between 15 and 24 (HR: 1.85, 95% CI 1.66–2.07) had about a two times higher risk of LTFU than other age groups, p < 0.0001, whereas males (HR: 2.00, 95% CI 1.51–2.62) have about twice the risk of being LTFU compared to females, p < 0.001. Divorced participants (HR: 1.35, 95% CI 1.24–1.48) were more likely to be LTFU than their counterparts. PLHIV on adult second-line drug type (HR: 1.13, 95% CI 1.09–1.18) were more likely to be LTFU than those in adult first-line drug type, p < 0.0001. Participants with poor drug adherence (HR: 1.50, 95% CI 1.23–1.75) had a 1.5 times higher risk of LTFU than those with good drug adherence, p < 0.0001. PLHIV with unsuppressed viral load (HR: 2.15, 95% CI 2.02–2.29) has about twice the risk of being LTFU compared to their suppressed counterparts. PLHIV who were not on DTG-related drugs (HR: 7.51, 95% CI 5.88–10.79) have about seven times increased risk of being to LTFU than those on DTG-related drugs, p < 0.0001. Additionally, PLHIV with WHO stage III and IV (HR: 2.51, 95% CI 2.32–2.72) had a higher risk of LTFU than their counterparts, p < 0.0001.

Cohabiting (HR: 0.42, 95% CI 0.35–0.52), pregnant women (HR: 0.73, 95%CI 0.63–0.85), and ART duration of less than 1 year (HR: 0.72, 95% CI 0.65–0.81) were the factors that reduced the likelihood of LTFU. While variables on tuberculosis history, WHO stage II, BMI, being widowed, and age above 55 + were statistically significant in the univariate analysis but lost their statistical significance level in the multivariate analysis, i.e., they were not the factors that were independently associated with LTFU.