Early virologic diagnosis is critical to initiating antiretroviral therapy (ART) promptly in infants born to people with HIV. Conventional diagnosis relies largely on centralized laboratory-based testing, which can delay result return and defer treatment initiation. Point-of-care (POC) testing devices provide same-day results and can therefore enable more timely ART starts, but device availability and placement are constrained in many low-resource settings. This analysis evaluated whether relocating POC machines based on programmatic data could improve clinical outcomes and economic value for infants tested for HIV in Zimbabwe.
The investigators combined the validated CEPAC-Pediatric HIV microsimulation model with a location-optimization model. The linked modeling framework incorporated detailed subnational programmatic data from 122 clinics across the seven districts of Matabeleland South, Zimbabwe. The models simulated infant diagnostic and treatment pathways and estimated individual- and population-level outcomes under alternative POC placement strategies.
Two testing strategies were simulated: conventional laboratory-based testing and POC machine–based testing. Key projected outcomes included the proportion of infants receiving test results within 30 days (30-day result-return), the proportion of children with HIV (CWH) initiating ART within 30 days, life expectancy (LE), HIV-related healthcare costs (discounted lifetime costs per infant), and net health benefit (NHB). The modeling considered both maximizing LE and maximizing NHB as objectives for device placement.
Using the 17 POC devices in their current locations as the baseline, the model projected that 43.8% of tested infants would receive HIV test results within 30 days. Among infants with HIV, 41.6% were projected to initiate ART within 30 days. Undiscounted life expectancy for the cohort was projected at 67.85 years overall and 25.91 years among children with HIV. Average discounted lifetime HIV-related healthcare costs were projected at $202 per infant.
When the model optimized the locations of the 17 devices to maximize life expectancy, it retained 6 devices in their current locations and moved 11 devices to new sites. Under this optimized configuration, projected 30-day result-return increased to 52.6% and 30-day ART initiation among CWH increased to 50.0%. Projected life expectancy rose slightly to 67.88 years overall and to 27.04 years among children with HIV. Average discounted lifetime costs increased to $211 per infant under the LE-maximizing placement.
Maximizing net health benefit (NHB) produced the same optimal device locations as maximizing life expectancy when the willingness-to-pay threshold was at least $1,164. The authors note that this threshold corresponds to approximately 66% of Zimbabwe’s GDP per capita. The equivalence of LE and NHB objectives at that threshold indicates the optimized placement yields both clinical and economic advantages under that valuation.
The analysis examined a policy constraint of ensuring at least one POC machine per district while maximizing life expectancy. Achieving one machine per district would require adding one additional device to the existing 17 (i.e., 18 machines total). Under that scenario, projected 30-day result-return increased to 53.3% and 30-day ART initiation to 50.6%. Life expectancy was projected at 67.88 years overall and 27.11 years among CWH. Average discounted lifetime costs under this scenario were $212 per infant.
Across scenarios, optimized placement increased the share of infants receiving timely results and the share of CWH starting ART within 30 days, with small absolute increases in projected cohort life expectancy. Average discounted lifetime costs per infant rose modestly in optimized configurations ($202 baseline to $211–$212 after optimization or district coverage expansion). Additional optimally located machines beyond the modeled set were predicted to yield further life expectancy gains, although specific numbers for additional machines were not detailed beyond the scenarios reported.
The authors interpret these findings to indicate that data-driven, location-optimization of POC infant HIV testing devices can improve early HIV detection and increase both life expectancy and net health benefit for infants undergoing HIV testing in Zimbabwe. Optimizing placement of a limited number of POC machines can meaningfully increase 30-day result-return and timely ART initiation, and modestly improve long-term outcomes while marginally increasing lifetime healthcare costs per infant. Ensuring at least one device per district is feasible with one additional machine and yields further improvements in timely diagnosis and treatment.
The abstract reports the model structure, inputs at a subnational level, and key outcome projections for specific optimization scenarios. The abstract does not report detailed model parameter values, uncertainty ranges or confidence intervals around the projected percentages and life expectancies, nor does it present granular results by individual districts or clinic-level changes. Specific assumptions used in the CEPAC-Pediatric model runs, sensitivity analyses, and implementation considerations at the health-system level are not described in the abstract.
Linking a validated pediatric HIV microsimulation with a facility-level location-optimization model for Matabeleland South, Zimbabwe, suggests that relocating existing POC machines and selectively adding devices can increase timely result-return and ART initiation for infants, improve cohort life expectancy, and enhance net health benefit at modest additional cost per infant. The abstract supports a policy argument for data-driven placement of limited POC resources to maximize clinical and economic impact.