We collected data for the period of April 2018–September 2021 at 10 health facilities with POC EID testing. Health facilities in Uganda are classified into seven levels. In ascending order, these are: clinic (community-based preventive and promotive health services), Health Centre Two (HC II), Health Centre Three (HC III), Health Centre Four (HC IV), general hospital, regional referral hospital (RRH), and national referral hospital (NRH) [14]. We collected data at three RRH (Fort Portal RRH, Mubende RRH, and Kawempe RRH), four general hospitals (Kiboga Hospital, Lyantonde Hospital, Mityana Hospital, and Kyenjojo Hospital), and 3 HC IVs (Kyegegwa HC IV, Mpigi HC IV, and Sembabule HC IV). With the exception of Fort Portal RRH which had Cepheid GeneXpert, the rest of the nine health facilities used Abbott m-Pima q HIV-1/2 Detect. Kiboga hospital also used Cepheid GeneXpert. The study health facilities were selected because they had POC testing introduced, reported the highest numbers of HEI tested for HIV in 2020 in their regions, and had a minimum EID testing volume of 12 infants per month according to the District Health Information System version 2 (DHIS2), a national electronic health database. The 10 sites were selected from the initial pool of 33 pioneer sites being an opportunity for the study to leverage the experience gained from early implementation efforts.

Socioeconomically, agriculture is the main source of income in 53% of the households in Uganda [15].

We conducted a retrospective evaluation of data for HEI at the 10 health facilities before and after the implementation of POC EID testing. At each facility, we abstracted data from EID registers for 12 months following the rollout of POC testing (during POC) at the facility. Since POC testing was introduced at health facilities at different times, the POC period ranged from April 2019 to September 2021. For comparison, we also abstracted data for 12 months before POC rollout (pre-POC period) when centralized testing at a reference laboratory was the standard of care. The pre-POC period ranged from April 2018 to September 2020. According to national guidelines, two HIV DNA PCR tests are conducted for HEI. The first test should be done at 4–6 weeks. HEI with a negative first HIV DNA PCR test should be retested using the same test 6 weeks after cessation of breastfeeding. Those with a negative second HIV DNA PCR should receive a final rapid HIV antibody test at 18 months [16]. The study utilized results for the first DNA PCR test.

Under POC testing, the health facilities had different set-ups based on size and type of machine. For example, sites with Cepheid GeneXpert machines had them placed in the laboratory which is a separate room or site, while sites with Abbott m-Pima q HIV-1/2 Detect machines had them at mother-baby care points (MBCP) and some in the laboratories. Generally, either the sample was drawn at the clinic and sent to the laboratory or the client was sent to the laboratory. The sample was processed at the POC (either MBCP or laboratory) where it was run through the machine; this could take up to an hour. Results were printed, returned to the clinic, and given to clients within the same day to the greatest extent possible.

During Pre-POC period, DBS samples were sent to the Central Public Health Laboratory (CPHL), the central reference laboratory via the hub system. Samples were collected at health facilities, delivered by a laboratory hub rider from the health facilities to the laboratory hub. A CPHL driver then picked the samples twice a week and delivered them to CPHL where they were sorted, coded and then tested. It took 3–7 days to test samples. Results were uploaded on the EID results dashboard, the hub downloaded these results, and the hub rider delivered them to the health facilities. At health facilities, results were recorded and given to the caregivers at follow up visits which could be between 2 weeks and 3 months.

We used Kobo Collect application to program a questionnaire on tablets. The questionnaire included infant’s identification number, sex, date of birth, date of registration at the facility, date of collection of first PCR test, dates results received, dates results given to caregiver, ART enrollment status and date, and final EID outcome (discharged negative, referred for ART, lost, died negative, died positive). We used the date of registration of the infants to determine whether they were registered for HIV testing pre-POC or during POC. HIV DNA PCR result turnaround times were defined as the number of days from sample collection to return of results to the clinic, or results return to the caregiver. HIV diagnosis date was defined as the time when HIV test results were received at the clinic. For turnaround time from sample collection to results receipt at the clinic, we used dates that results were received at the clinic or date of last clinic visit (date of last follow-up) for censored observations. Censored observations were those for which survival times were unknown because they had no date of results returned. For turnaround time from sample collection to results receipt by caregiver, we used dates of caregiver results receipt.

A sub-group analysis was conducted among infants who tested positive for HIV to assess turnaround times and the effect of POC testing on time to ART initiation. The primary outcome in this study was time to ART initiation. Time to ART initiation was defined as the number of days between dates of sample collection and initiation on ART; same-day ART initiation was defined as starting ART on the same day of sample collection. Positivity rate was defined as the proportion of infants that tested HIV positive out of the number tested who had valid results. The secondary study outcome was time to first HIV DNA PCR sample collection from birth. When calculating proportion of infants with HIV test results, only HEI who had date of result returned to the clinic were included in the analysis.

Trained research assistants working in the EID clinics at health facilities completed the questionnaire. Data were sent from the tablet computers to the Kobo Collect server each day. We analyzed the data in Stata version 14. Duplicate entries were removed using the exposed infant identification number and health facility name. HEI missing dates for a step in the care cascade were excluded in the analysis for that particular step.

We calculated summary statistics for all variables. Categorical variables were presented as frequencies and proportions and continuous variables were described using medians and interquartile ranges. We compared time to sample collection, results receipt at the clinic and by the caregiver, and ART initiation between pre-POC and POC periods using the Wilcoxon rank-sum test and Kaplan Meier curves. The log-rank test was used to test for differences in time to ART initiation between the pre-POC period and the POC period as displayed in Kaplan–Meier curves.