Uptake of same-day initiation of HIV treatment in Malawi, South Africa, and Zambia as reported in routinely collected data: the SPRINT retrospective cohort study

Ethics

Ethical clearance was provided by the Boston University Medical Campus Institutional Review Board (IRB) (Ref No. H-40354 (Malawi, May 21, 2020), H-39330 (South Africa, October 15, 2019) and H-40488 (Zambia, July 10, 2020)), the University of the Witwatersrand Human Ethics Research Committee in South Africa (Ref No. M200238 (Malawi, July 10, 2020), M190745 (South Africa, November 6, 2019) and M200599 (Zambia, July 10, 2020)), the National Health Sciences Research Committee in Malawi (Ref No. 20/04/2458, April 24, 2020) and the ERES Converge Institutional Review Board in Zambia (Ref No. 2020-Feb-009, February 27, 2020). The Ministry or Department of Health in all three countries and the National Health Research Authority in Zambia also approved the study.

As this was a retrospective review of medical records and no identifiers were collected, a waiver of informed consent was provided. Data transcription from patient records began on May 27, 2020 in Malawi, September 28, 2020 in Zambia, and November 12, 2020 in South Africa.

IRB approval from the University of the Witwatersrand was sought to allow individuals on the team who were based in South Africa to access Malawi data for analysis purposes. There was no need for University of the Witwatersrand approval in order to collect the data, as that was only under the purview of the IRBs of Boston University and the National Health Sciences Research Committee in Malawi, whose approval was secured prior to the start of data collection. All required approvals were in place before data collection began in all three countries. Team members at the University of the Witwatersrand did not access the data from Malawi prior to their own institutional IRB approval.

Study sites

SPRINT was a retrospective record review conducted in Malawi, South Africa, and Zambia. In each country, in collaboration with national health officials, we first purposively selected provinces and districts that provided diversity of setting (rural/urban) and of nongovernmental supporting partner organizations. We then identified a convenience sample of public sector primary healthcare clinics within these provinces or districts and conducted visits to these sites to assess facility-level and HIV testing and initiation-specific aggregate indicators. The final sites selected provided diversity in size and geographic setting, had relatively strong routine data collection, and were welcoming of the research team. Selected study sites are described in Table 1. All of the study sites offered universal treatment and followed national guidelines for ART treatment eligibility^5–7; eligibility criteria included confirmed HIV-positive status and provider and patient concurrence that the patient was ready to start treatment, based on clinical and personal characteristics. The SPRINT studies are registered with Clinicaltrials.gov (NCT04468399, July 13, 2020, Malawi; NCT04170374, Nov 20, 2019, South Africa; NCT04470011, July 14, 2020, Zambia).

At the time of study enrollment, guidelines for ART initiation in the three study countries encouraged but did not require that SDI be offered to patients whom providers determined to be eligible, without waiting for CD4 count results. In all countries, guidelines encouraged starting HIV treatment as soon as possible^5–7 but also noted the potential need for delay among patients who indicated that they were not yet ready to start. The decision on whether a patient was eligible for same-day initiation was made by the authorized clinician, typically a clinical officer or trained ART nurse in Malawi and Zambia and a public health (senior) nurse in South Africa, each of whom was also authorized to prescribe antiretroviral medications once the decision to initiate was made.

Tuberculosis (TB) symptom screening was recommended in all three countries, with a TB test conducted for symptomatic patients prior to ART initiation. Malawi’s guidelines recommended TB symptom screening but encouraged ART initiation regardless of results: “[Start ART] within 14 days of diagnosis. TBT [TB treatment] + ART can be started on the same day if the patient is stable. Don’t delay TBT or ART.” Only South Africa’s guidelines explicitly required delaying ART initiation for all patients with TB symptoms, stating “Initiate ART after 2 weeks of TB treatment, when the client’s symptoms are improving, and TB treatment is tolerated.” Zambia’s guidelines recommended delaying ART based on suspicion of TB, without further elaboration. Guidelines did not specify actions for symptomatic patients who could not produce sputum samples for TB testing or for whom TB test results were not available, nor how long to delay ART initiation among patients diagnosed with active TB^5–7.

Enrollment and data collection

Patients were eligible for inclusion in the study if they were 1) non-pregnant adults (≥18 years old); 2) tested HIV-positive or provided documentation of HIV infection (prior positive test); and 3) were eligible to start or re-start ART at one of the study sites between 01 January to 31 December 2018 in South Africa or between 01 July 2018 and 30 June 2019 in Malawi and Zambia. We retrospectively identified up to 200 patients per site, starting on the latest date in the inclusion period (30 June 2019 in Malawi and Zambia, 31 December 2018 in South Africa) and working sequentially backward in time until the target sample size was reached (01 July 2018 in Malawi and Zambia, 01 January 2018 in South Africa). Study-eligible patients were selected from the HIV testing registers, testing database, or other testing records kept by the site, which were reviewed on site by study data collectors. Medical record data were downloaded from EMRs where available. Due to concerns about incomplete EMR data, however, all records for participants in Malawi and Zambia were extracted manually from paper files by study data collectors using a structured instrument and either the offline mobile or online web version of REDCap electronic data capture tools hosted at Wits University, depending on internet connectivity at the site. In South Africa, records were extracted manually at all facilities, but for 92 of 534 participants (17%) distributed across all the sites, paper files could not be found. For these records, EMR data were extracted instead.

Records were reviewed from the point of HIV treatment eligibility (HIV-positive diagnosis) to the point of treatment initiation or for a 6-month period, whichever was shorter. By starting with HIV diagnosis, rather than with treatment initiation, the study captured information about patients who failed to initiate treatment within 6 months (dropped out of care before being dispensed medications) and allowed us to observe services received and days elapsed before initiation. We note that the HIV test used to identify study participants may not have been any individual participant’s first positive test anywhere, but only their first positive test at the study site during the study period. Because repeat testing is common⁸, we assume that many participants had previously tested positive at another location but failed to initiate treatment at that time or place.

Sample size, outcomes, and statistical analysis

Study inclusion numbers were determined by resource availability. Our maximum sample size was 200 patients per site, or up to a total of 2,400 patients each in Malawi and Zambia and 1,600 in South Africa. The primary outcome of interest for the study was SDI, defined as being dispensed an initial supply of ARV medications on the day of testing positive for HIV or the day of first HIV-related clinic interaction, if a positive test had been conducted previously at another location, without subsequent linkage to care.

Simple descriptive statistics were used to describe the study sample. In order to assess site-specific characteristics, we stratified results by country, site, and presence of TB symptoms and ranked facilities by volume (number) of ART patients. Predictors of SDI were assessed using modified Poisson regression⁹, clustered by study site to allow intragroup correlation in the standard errors. For our first model (Model 1) including data from all three countries, we adjusted for biological sex, age (categorized as 18–29, 30–39, 40–49 and ≥50 years of age), WHO stage III/IV (vs. I/II), urban/peri-urban (vs. rural), volume on ART at each site, and country, all assessed at ART initiation, as potential risk factors for SDI. Considering that there were no CD4 cell count or TB symptom data from Malawi, we then ran the same model (Model 2) restricted to Zambia and South Africa and included a variable indicating whether the patient received a CD4 count (yes/no) and the presence of TB symptoms (yes/no) at initiation.

Time to ART initiation

Time to ART initiation is presented in Table 3. Same-day initiation was the norm in Malawi (734/825, 89%) and Zambia (1,813/1,984, 91%), where a vast majority of patients did start ART on the day of diagnosis or first HIV-related clinic interaction. It was much less common in South Africa (304/534, 57%), where a fifth of patients (116, 22%) did not have a record of initiating ART at all within the 6-month study observation period. Among patients who did not initiate on the same day, varying proportions proceeded to initiate within 1, 2, 4, and 24 weeks, as shown in Table 3. Those who did initiate within 6 months, but not on the same day, made a median of 1 additional clinic visit (2 in total) between diagnosis and initiation (inclusive.).

Across all three countries, as shown in Table 3 and Table 4, we identified a total of 181 participants who did not initiate treatment within 6 months of HIV diagnosis or their first recorded HIV-related clinic visit (non-initiators). These patients comprised 8% (n=64) of the cohort in Malawi, 21% (n=114) in South Africa, and just 0.2% (n=3) in Zambia. We note that while these proportions are consistent with expectations in Malawi and South Africa, it is likely that non-initiators in Zambia who should have been included in our sample were not. Although we made multiple attempts to confirm the completeness of our cohort in Zambia, we believe it is likely that more than 0.2% of patients in Zambia who were eligible to initiate ART did not start within 6 months. Among those who did not initiate within 6 months, in Malawi most were either lost to follow-up or remained alive and in care, but not on ART, while in South Africa most were either lost to follow-up or were simply missing follow up records (Table 4), a category which likely includes both unreported (silent) transfer to other healthcare facilities and loss to care.

When we stratified results by country, clinic, and TB symptom status and ranked sites according to ART patient volume (Supplemental Table 1), our results suggested that patients presenting with TB symptoms in South Africa were less likely to initiate ART on the same day than were those with TB symptoms. In rural clinics in South Africa, the likelihood of SDI tended to decline as patient volume increased. We did not see these associations in Zambia or Malawi.

Predictors of same-day initiation

Table 5 presents the results of the modified Poisson regression. In Model 1, which included data from all three countries, we found that patients with a WHO stage III/IV (vs. stage I/II) had an 8% decrease in the likelihood of SDI (adjusted risk ratio (aRR): 0.91; 95% confidence interval (CI): 0.85-0.98). Patients in South Africa had a 19% decrease in the likelihood of SDI (aRR: 0.81; 95% CI: 0.73-0.89) compared to Zambia, while the uptake of SDI in Malawi was comparable to that of Zambia. We did not control for CD4 counts or TB symptoms in this model, as these were not available in Malawi.

In Model 2, restricted to South Africa and Zambia, we found similar results to Model 1. Patients with a WHO stage III/IV (vs. stage I/II) had an 8% decrease in the likelihood of SDI (aRR 0.92; 95% CI: 0.84-0.99) and those in South Africa had an 18% decrease in the likelihood of SDI compared to Zambia (aRR: 0.82; 95% CI: 0.76-0.87). We also found that those with TB symptoms had a 15% decrease in the likelihood of SDI (aRR 0.85; 95% CI: 0.73-0.97). Our results also suggest that patients who did have a CD4 count done (aRR: 1.10; 95% CI: 1.02-1.19) and larger vs. smaller clinic size (5,000-10,000 vs. <2,000 patients – aRR 1.09; 95% CI: 1.01-1.18 and >10,000 vs. <2,000 patients – aRR 1.07; 85% CI: 0.99-1.15) were more likely to initiate ART on the same day. Sex, age and location (urban/peri-urban or rural) were not predictors of SDI in either model.