Objective: To investigate how ‘real-world’ constraints on the allocative and technical efficiency of HIV prevention programmes affect resource allocation and the number of infections averted.

Design: Epidemiological modelling and economic analyses in Benin, South Africa and Tanzania.

Methods: We simulated different HIV prevention programmes, and first determined the most efficient allocation of resources, in which the HIV prevention budget is shared between specific interventions, risk-groups and provinces to maximise the number of infections averted. We then identified the efficient allocation of resources and achievable impact given constraints to allocative efficiency: earmarking (provinces with budgets fund PrEP for low-risk women first), meeting treatment targets (provinces with budgets fund UTT first) and minimizing changes in the geographical distribution of funds. We modelled technical inefficiencies as a reduction in the coverage of PrEP or UTT, which were factored into the resource allocation process or took effect following the allocation. Each scenario was investigated over a range of budgets, such that the impact reaches its maximum.

Results: The ‘earmarking’, ‘meeting targets’ and ‘minimizing change’ constraints reduce the potential impact of HIV prevention programmes, but at the higher budgets these constraints have little to no effect (approximately 35 billion US$ in Tanzania). Over-estimating technical efficiencies results in a loss of impact compared to what would be possible if technical efficiencies were known accurately.

Conclusions: Failing to account for constraints on allocative and technical efficiency can result in the overestimation of the health gains possible, and for technical inefficiencies the allocation of an inefficient strategy.