Cost-effectiveness of differentiated HIV service delivery in North Wollo, Ethiopia: a retrospective study

Background Differentiated service delivery (DSD) models have been introduced to optimise HIV care by tailoring services to patient needs, but evidence on their cost-effectiveness in Ethiopia is limited. Objective To evaluate the cost-effectiveness of DSD models for HIV care compared with standard of care (SOC) in North Wollo, Ethiopia. Study design Retrospective economic evaluation using bottom-up microcosting from a healthcare provider perspective. Setting One hospital and three health centres in North Wollo Zone, Ethiopia, from January 2024 to December 2024. Participants 786 adults (≥18 years) on antiretroviral therapy, including 393 receiving DSD (across nine models: 6-month multimonth dispensing, 3-month multimonth dispensing, appointment spacing, community antiretroviral therapy (ART) groups (CAG), FastTrack, maternal and child health, peer-led, adolescent groups and key population models) and 393 receiving SOC. Main outcome measures Total annual cost per patient (USD), retention in care at 12 months (%) and incremental cost-effectiveness ratios (ICERs) expressed as cost per additional patient retained. Statistical analysis Costs were annualised at a 3% discount rate. ICERs were calculated comparing each DSD model to SOC. Deterministic one-way sensitivity analysis (±20% for medication cost and retention rate) assessed robustness. Results SOC had the highest annual cost per patient (US$295) and lowest retention (76.6%). All DSD models were dominant (lower cost, higher retention) with negative ICERs. CAG was the most cost-effective (US$199 per patient; 93.0% retention; ICER –5.85), followed by FastTrack (US$212; 92.9%; –5.09) and adolescent groups (US$238; 90.5%; –4.10). Sensitivity analyses confirmed robustness; no scenario reversed dominance. Conclusion DSD models, particularly CAG, FastTrack and adolescent groups, provide more cost-effective HIV care than SOC in Ethiopia by achieving higher retention at lower costs. Scaling up these models could optimise resource allocation and reduce health system burden.