We conducted a retrospective cohort study of PWH enrolled in HIV care in 12 President’s Emergency Plan for AIDS Relief (PEPFAR)-supported districts in one province of Zambia.

Clinical data from patient records were extracted for PWH aged 15 years and older who were initiated on ART between January 1, 2014 and October 1, 2020, regardless of their clinical outcome at the time of data abstraction. PWH initiated on ART before the launch of the test-and-treat policy (January 1, 2014 to July 31, 2016) formed the pre-test-and-treat cohort, and those initiated in the period following the implementation of the test-and-treat policy (August 1, 2016 to October 1, 2020) formed the post-test-and-treat cohort.

All PWHs that enrolled in care but were not initiated on ART by the time of data abstraction were excluded from the analysis. However, the details of those who were not on ART were provided to the ART health providers for further follow-up and counseling with the aim of commencing ART. PWH with either a missing date of birth or age or a missing ART initiation date were excluded from the study.

Clinical data for PWH were abstracted either from HIV paper-based or electronic medical records known as SmartCare in 45 health facilities located in 12 districts of Southern Province onto the Electronic Data Capture (REDCap) software. Patient outcomes were recorded around the follow-up period of 3, 6, 12, and 24 months ±30 days and the last date of contact with the health care provider. In addition to clinical data (ART regimen, WHO stage, BMI, duration on ART, retention, stopped treatment, death, transferred out, and LTFU), laboratory (CD4 count and HIV viral load), and pharmacy-related data, including medicine refill dates, were collected from laboratory and pharmacy registers, respectively, in facilities where SmartCare was not fully functional. Where possible, the data were triangulated from both SmartCare and paper registers.

A line list of PWH was created from the ART clinic enrollment register per national guidelines. If the site had fewer than 100 patients on ART, data were abstracted from all the patient SmartCare medical records. For sites with more than 100 patient records, a systematic sampling scheme was used. At each site, we determined a sampling interval (X) by dividing the total number of client records at the site by the sample size and taking the whole number only (dropping the decimals); this process was done for the allocated sample sizes in the before and after test-and-treat strategies. The first record for review was selected (N) by randomly selecting a number between 1 and X (sampling interval). Data abstractors selected the Nth record and proceeded with selecting record N + X as the first consecutive record for review, which continued with adding X until the quota for that site had been reached. All the selected records were entered into the REDCap application and screened for inclusion and exclusion criteria. If any patient record from the selected list was excluded from the study, the next eligible record fulfilling the inclusion criteria was selected to replace the excluded record.

Data entries from 10% of the selected patient records were re-abstracted by the team facilitator immediately after the needed records were abstracted by the data extraction team using a systematic random sampling method. Once the files had been re-abstracted, any discrepancies between the data initially abstracted and the data that were re-abstracted were discussed, and the option that was unanimously agreed to be correct was selected. We have used guidelines for strengthening the reporting of observational studies in epidemiology (see Supplementary File 1, STROBE checklist).

The primary outcome was retention in HIV care, defined as regular attendance at appointments or engagement with the ART clinic at 3, 6, 12, 24, and ≥ 24 months after ART initiation. Secondary outcomes were LTFU, transferred out, stopped treatment, and death. LTFU was defined as a PWH whose medical records showed no evidence of any form of contact with a health facility for 30 or more days after the scheduled visit day. Transferred out is when a PWH moves to another ART clinic to continue receiving ART services. Stopped treatment was when a PWH stopped HIV treatment. We defined HIV viral suppression as having a viral load of less than 1,000 copies/mL after being on ART for ≥6 months as per national guidelines (37).

PS software was used to estimate the sample size (38). We hypothesized that the proportion of retention would be slightly higher after a test-and-treat period. An uncorrected chi-squared test was used to test the difference in proportions of retention between the two groups (before and after test-and-treat implementation), and the sample size was estimated accordingly with reference to a study conducted in Malawi (19). The proportion of retention BTT was 76.2%, and we expected a 3.7% increase in ATT (79.9%). We estimated requiring 6,688 records to achieve at least a 90% chance of detecting a difference (one-sided α = 0.05, allowing 5% type I error, conservative design effect of 2, power = 0.9).

We selected 45 health facilities [21 rural health centers (RHC), 16 urban health centers (UHC), and eight hospitals] across the 12 districts from 274 ART sites. In the initial step, the 274 ART sites were stratified into rural and urban sites, and then, a probability based on the number of PWH on ART at each health facility was used to select the required number of sites from each stratum.

The data were entered in REDCap and thereafter exported to STATA version 15 and R for standard data analysis. Data were described using frequency and percentages for categorical variables and medians (interquartile ranges) for continuous variables. To test for normality, the Shapiro–Wilk test was used. The Wilcoxon rank-sum test was used to ascertain the statistical difference between the two medians. A relationship between two categorical variables was determined using a chi-squared test. Multivariable logistic regression was used to examine the factors associated with retention using generalized estimating equations (GEE) via the logit link function to account for correlation structure within observations. The variables [cohort (ATT), age, sex, facility location, marital status, duration to ART initiation after HIV diagnosis, WHO stage, ART regimen, and BMI] in the adjusted analysis were selected based on previous literature (19, 26, 31, 32) and knowledge from HIV treatment and management experts. To ascertain statistical significance, a p value of <0.05 was used.

Ethical approval was obtained from the Macha Research Trust Institution Review Board (IRB) and the Zambia National Health Research Authority (ZNHRA). This project was also reviewed in accordance with the Centers for Disease Control and Prevention’s (CDC) human research protection guidelines. In this study, we analyzed de-identified data from healthcare facilities.

Owing to the anticipated potential presence of duplicates and missing information, we abstracted 6,910 records. After removing duplicates and records with missing information on the outcome variable, only 6,640 records were included in the analysis (Figure 1). We identified 2,991 (45.1%) records for PWH initiated on ART before test-and-treat (BTT) and 3,649 (54.9%) after test-and-treat (ATT) policies (Figure 1). The median age was 38 years [interquartile range (IQR), 31.0, 46.0], and participants in the BTT policy were significantly older, 40 vs. 37 years, p < 0.001. Women comprised 60.2% (n = 3,995) of PWH in the study sample. The majority of the participants, BTT (70.5%, n = 2,110) and ATT (55.9%, n = 2040), were from urban areas. There were more married individuals than those who were never married, divorced, or widowed in both the BTT (67.8%, n = 1813) and ATT (62.9%, n = 2022) cohorts. The baseline body mass index (BMI) was significantly higher among participants in the ATT than the BTT (21.6 vs. 21.3 kg/m², p < 0.001). BTT participants took 14 days (IQR: 0, 32) to be initiated on ART from the time of HIV diagnosis compared with the ATT cohort that took 0 days (IQR: 0, 1), p = 0.001. A high proportion of the participants were taking non-nucleotide reverse transcriptase inhibitors (NNRTIs) at baseline (99.7% BTT and 77.6% ATT) and currently (71.4% BTT and 56.8% ATT). At baseline, the majority of patients had WHO stage I HIV, with a higher proportion of patients in the ATT cohort with Stage I (83.9%) vs. those in the BTT cohort (77.2%; Table 1).

The overall prevalence of CD4 ≥ 500 cells/μL was 22.9, 37.9, and 47.8% at baseline, 6 months, and 12 months, respectively. Similarly, an increase in the proportion of patients with viral suppression (VL <1,000 copies/μL) was observed at different time points: 6 months (80.5%), 12 months (92.4%), and 24 months (94.6%).

At baseline, a significantly higher proportion of participants in the ATT cohort had CD4 ≥ 500 cells/μL than BTT (26.6 vs. 19.5%, p < 0.001). However, at 12 months, a higher proportion of participants in BTT than ATT had CD4 ≥ 500 cells/μL, 49.6 vs. 45.1%, p = 0.018. A higher proportion of viral suppression was observed among the participants in the BTT cohort at 6 months (86.1 vs. 77.4%, p < 0.001) and 12 months (93.6 vs. 91.5%, p = 0.030). At 24 months, the viral load categories were comparable (Table 2).

At the time of data abstraction, 5,543 (83.4%) were actively on ART, 13 (0.2%) had stopped ART, 412 (6.2%) were transferred out, 621 (9.4%) were LTFU, and 51 (0.8%) were dead (Table 3). There was a statistically higher proportion of retention on ART for the ATT cohort as compared to the BTT cohort (90.5 vs. 74.8%, p < 0.001), while attrition was generally high in the BTT cohort as compared to the ATT cohort: death (1.4 vs. 0.2%, p < 0.001), stopped treatment (0.3 vs. 0.1%, p < 0.001), transferred out (9.9 vs. 3.2%, p < 0.001), and LTFU (13.5 vs. 5.9%, p < 0.001).

Retention was higher in the ATT cohort as compared to the BTT cohort (Figure 2) at 3 months (92.8 vs. 75.8%, p < 0.001), 6 months (91.6 vs. 76.0%, p < 0.001,), 12 months (92.1 vs. 77.0%, p < 0.001), 24 months (92.1 vs. 78.3%, p < 0.001), and > 24 months (92.8 vs. 87.3%, p < 0.001), that is, all time points.

In the cohort BTT, the factors significantly associated with retention in HIV care were older age, facility location, marital status, higher baseline BMI and CD4 count, and being in baseline WHO stage 1. While ATT participants, older age, lower viral load at 24 months, and being initiated on ART on the same day of HIV diagnosis were the factors significantly associated with retention in HIV care (Supplementary Table 1).

In the univariate analysis, there was a statistically significant association between retention and the following determinants: Participants in the ATT cohort were 3.72 times more likely to be retained in HIV care compared to individuals in BTT (OR: 3.72, 95%CI 3.44, 4.02). A year increase in age was significantly associated with 1% increased odds of being retained in HIV care [odds ratio (OR) 1.01; 95%CI 1.01, 1.02]. Participants from rural areas had a significantly higher probability (54%) of being retained in HIV care compared to those from urban areas (OR 1.54; 95%CI 1.42, 1.67). Individuals who had never been married had 22% reduced odds of being retained in HIV care compared to those who were married (OR 0.78; 95%CI 0.70, 0.85). An increase in the number of days from HIV diagnosis to ART initiation was significantly associated with a 1% reduced chance of being retained in HIV care. Participants who were in WHO stage 3 at baseline had 35% reduced odds of being retained in HIV care compared to the ones in stage 1 (OR 0.65; 95%CI 0.58, 0.73). Clients on a DTG and protease inhibitor (PI)-based regimen were 2.27 and 2.29 times more likely to be retained in HIV care relative to those on NNRTIs, respectively. A unit increase in baseline BMI was significantly associated with 3% increased odds of being retained in HIV care (OR 1.03; 95%CI 1.02, 1.04).

In the adjusted analysis, individuals from the ATT cohort were 3.67 times more likely to be retained in care. A unit increase in age was associated with a 1% increase in the odds of being retained in HIV care. Participants from rural areas had 32% higher odds of being retained in care than those from urban areas. Never married and divorced participants had 24 and 16% reduced odds of being retained in HIV care, respectively, compared to married ones. Baseline participants in WHO stage 2 had 18% increased odds of being retained in HIV care (OR 1.18; 95%CI 1.05, 1.32). However, baseline participants with WHO stage 3 had 23% reduced odds of being retained in HIV care. A unit increase in BMI increased the chance of being retained in care by 2% (Table 5).

In the BTT participant group, factors significantly associated with retention upon adjusted analysis included receiving care at a rural health facility (OR 1.54; 95%CI 1.21, 1.94) and being never married (OR 0.70; 95%CI 0.52, 0.95). In contrast, for the ATT cohort, factors linked to ART retention were increasing age (OR 1.02; 95%CI 1.01, 1.04) and being in WHO stage 2 (OR 2.32; 95%CI 1.20, 4.52). See Supplementary Table 2.

Missing data led to delays in data abstraction. Some characteristics, such as viral load and CD4 count, were missing. We did not collect data on alcohol consumption, HIV disclosure status, year of enrollment for each participant, distance to the health facility, or adherence counseling, which could have helped explain retention rates among our study participants. Additionally, we could not get the patients’ perspectives regarding UTT policy. However, the study was able to quantify retention in HIV care and clinical outcomes before and after the universal test-and-treat policy.

We used a large sample size in our study compared to similar studies. Hence, this study was highly powered, and the results may be applicable to PWH living in Zambia. The methodology and statistical analysis were rigorous. The multicenter model and sampling methods used strengthened the study and minimized bias, providing a comprehensive overview of the whole program in the province.