In compartmental models of sexually transmitted infections, populations are stratified into pre-specified states ("compartments"), which are assumed to be homogeneous and memoryless. Thus, individual sexual partnerships cannot be tracked. However, these models still need to account for continued sexual contact within partnerships where transmission has already occurred.

The prevailing approach to this problem defines a cumulative probability of transmission per partnership (or partnership-year), which approaches but never exceeds 100%. This probability is then multiplied by an average partnership change rate to give the overall force of infection. In this way, partnerships are effectively modelled as instantaneous.

In this talk, we review the limitations of this "instantaneous partnerships" approach, and introduce a new approach: the Effective Partnerships Adjustment (EPA), which overcomes some of these limitations. EPA introduces a new population stratification to explicitly track individuals who already transmitted or acquired infection within an ongoing partnership. These individuals can then be removed from the force of infection until they form a new partnership. In this way, EPA avoids the need to estimate and apply an instantaneous cumulative probability of transmission per partnership.

Using an existing model of heterosexual HIV transmission in Eswatini, we illustrate how the choice of approach can substantially influence model-estimated resource allocation priorities. Specifically, the instantaneous partnerships approach tends to overestimate the importance of prevention in longer partnerships and underestimate the importance of prevention in shorter partnerships, as compared to the EPA approach.