We used a pre- and post-intervention study design to evaluate 6-month retention and VS before and after implementation of an intensive case management intervention in six health facilities in Kalangala District, Uganda, supported by the Centers for Disease Control and Prevention (CDC) under the President’s Emergency Plan for AIDS Relief. Three facilities were < 2 h and three were > 2 h from the mainland by boat. Data from the first medical visit following an HIV-positive test onwards were abstracted from medical records, allowing for 6-month follow-up.

For both the pre-intervention and intervention phases, we consecutively enrolled eligible participants, which included ART-naïve adults aged ≥ 18 years who had initiated ART as part of HIV care at one of the six health facilities at least once, following an HIV positive test result. For the intervention phase, only persons who consented to receive the intervention had their data abstracted. The six facilities comprised the primary sampling units. Assuming a 40% relative improvement in retention (from approximately 50% to 90%) in the intervention phase, with an inter-cluster correlation coefficient of 0.05, we calculated a minimum sample size of 340 participants in each phase, with 80% power to detect the assumed difference in retention between the phases.

Ethical approval was granted by the Joint Clinical Research Center in Uganda and the protocol was registered with the Uganda National Council of Science & Technology. This project was reviewed in accordance with CDC human research protection procedures and was determined to be research, but CDC investigators did not interact with human subjects or have access to identifiable data or specimens for research purposes.

During the pre-intervention phase, the standard of HIV care included the performance of a baseline CD4 test, monthly clinical follow-up for the first three months on ART then every 2–3 months for ART refills, screening for opportunistic infections, cotrimoxazole prophylaxis, and viral load (VL) testing conducted at 12 months after ART initiation, thereafter annually if virally suppressed. Health care workers from the health facilities conducted visits monthly at facility-managed outreach sites in the community (e.g., boat landing sites) to provide HIV testing, counselling, and ART initiation and refills. The case manager cadre did not exist, instead facility staff encouraged newly diagnosed individuals to return for follow-up care at the facility or their outreach sites. Health facilities utilized village health teams (VHT) to track persons who did not return for follow-up. Tracking was conducted via home visits and phone calls.

During the intervention phase, the World Health Organization’s Treat All policy (implemented in Kalangala in December 2017) was adopted, making all people with HIV eligible for ART [28]. Virologic testing was implemented at six and twelve months after ART initiation, thereafter annually if virally suppressed. To improve linkage and retention at ART facilities nationwide, linkage facilitators and volunteer counsellors were employed to support outreach activities and tracking of missed appointments.

The implementation of the intervention was initiated in May 2018 with recruitment and training of community and facility case managers. Each of the six health facilities was assigned 2–3 facility case managers (FCM) and 1–2 community case managers (CCM), for a total 15 FCM and 17 CCM. During the intervention phase, newly diagnosed HIV-infected persons initiating ART were provided written consent to be followed by any one of the 3–5 FCM and CCM assigned to the community for the first 3 to 6 months after initiating ART.

FCM were expert clients familiar with the facilities, where they supported medical chart retrieval and triaging. Each FCM was expected to follow 4–6 participants. FCM oriented participants to HIV care, ensured that standard of care was provided, and rendered ongoing supportive counseling to manage barriers to adherence, including stigma and discrimination.

Community case managers, recruited from the VHT serving within 10 km of each facility, signed a confidentiality agreement and received training on HIV adherence counseling and case management, mobilized and educated the community on health issues. CCM followed 10–15 participants and ensured that those diagnosed at outreach sites registered at the facility, assessed barriers to care, tracked missed appointments via phone call, text, or home visits, and delivered ART refills to those unable to attend the clinics. They coordinated tracing with VHTs when participants relocated and liaised with community support groups to mobilize the community in HIV prevention and treatment. CCM also ensured participants received pre-appointment reminders and encouraged participants’ household members to test for HIV and distributed vouchers to be presented at the facilities or at outreach sites to facilitate fast tracking through the clinic. CCM interacted with participants at least two times per month and documented the status of the participant (e.g., alive on ART, LFU, died), adherence, and the presence of any symptoms in their diaries. CCM and FCM for each facility met once per month to review missed appointments and collaborated with peer counsellors and linkage facilitators to ensure treatment continuity.

In both phases an abstraction form was used to collect de-identified data from electronic and paper records documenting their medical visits for HIV care services. Data included demographics, care-entry point (main health facilities or their outreach locations), HIV diagnosis date, ART initiation date, WHO HIV stage, CD4, HIV VL test date and result, and monthly care status (i.e., transferred, died, alive and retained care, and LTFU). During the intervention phase, case managers’ diaries were also used as a data source to assess participants’ care status. Study staff reviewed the diaries for completeness of entries weekly, including details of how case managers contacted participants, whether participants had relocated, and if a future visit was scheduled. Further, facility nurses-in-charge were asked to verify case manager interactions with participants when they presented for ART refill. Data from the abstraction forms were entered into a REDCap database, which was assessed for data quality and completeness through regular queries.

We excluded participants who transferred out of the six facilities from the analysis. The main outcomes assessed were 6-month retention and VS. Based on the President’s Emergency Plan for AIDS Relief (PEPFAR) Monitoring, Evaluation, and Reporting indicators (version 2.2), retention was defined as having a documented provider visit, ART pick-up at the facility, or encounter with a case manager in the community, within 90 days after the scheduled appointment date [29]. VS was defined as having < 1,000 copies/ml. Available VL results within 90 days prior to or after the scheduled follow-up visit were included in the analysis.

Among the remaining participants meeting the inclusion criteria, we used descriptive statistics to compare demographic and clinical characteristics, and 6-month VS (overall and among participants retained) in pre-intervention and intervention phases. We compared frequencies of overall outcomes (i.e., retention, LTFU, and died) by phase.

Logistic regression was conducted to assess factors associated with 6-month retention and to estimate retention probabilities. The covariates of interest were age range (18–24, 25–34, 35–44, and 45 and older), sex, WHO stage (I–IV), entry point (facilities or their outreach sites) and phase (pre-intervention and intervention). Among the 1,011 participants who did not transfer out during the study, there were 1, 51, 21, and 55 missing observations for sex, WHO stage, age range, and entry point, respectively. These missed observations were completed in ten datasets using multiple imputation chained equations [30]. Mixed-effects logistic regression models including an intercept-level random effects for site were used for inference to determine associations with retention in care. Models were fitted for each of the ten imputed datasets, and results were combined using Rubin’s rules [31]. We began with a full model including sex, age group, WHO stage, entry point, phase, and four two-way interaction terms (i.e., sex × phase, entry point × phase, age × phase, and sex × age). We sequentially deleted insignificant interaction terms, followed by insignificant covariates and levels of categorical covariates that did not meet a threshold of P ≤ 0.05 based on t or F tests for binary or categorical covariates, respectively. The marginal (averaging over sites) estimates of retention were obtained from the final multivariable model fitted to the multiply imputed data. All statistical analyses were performed using STATA 16.1, (StataCorp, College Station, TX).