Introduction We quantified and characterised the temporal evolution of the contribution of several types of sexual-partnership HIV statuses to total HIV incidence in representative countries at different HIV epidemic scales in sub-Saharan Africa.

Methods A pair-based mathematical model was constructed to accommodate for movement of individuals between different forms of sexual partnerships and HIV infection statuses. Stable couples (SCs) were divided into sero-concordant negative (SCNCs), sero-discordant (SDCs), and sero-concordant positive couples. Throughout epidemic phases, historical and future trends (1980–2030) of sources of HIV incidence by partnership status were projected based on nationally-representative epidemiologic and demographic data.

Results As the epidemics emerged and peaked, HIV incidence resulting among SDCs increased from 6%–32% to 19%–44% of total HIV incidence, mostly due to transmissions within the SDCs. As the epidemics started their natural and further decline with antiretroviral therapy (ART) scale-up, this proportion declined in intermediate and high HIV-prevalence countries, but increased in low HIV-prevalence countries. As the epidemics emerged and peaked, HIV incidence resulting among SCNCs from extramarital sex declined from 54%–80% to 35%–73% of total HIV incidence. As the epidemics started their decline, this proportion increased in intermediate and high HIV-prevalence countries, but declined in low HIV-prevalence countries. HIV incidence resulting among individuals not in SCs was 9%–29% of total HIV incidence and was stable throughout the epidemics, but larger with higher HIV prevalence.

Conclusion The contribution of different forms of sexual-partnership HIV statuses to total HIV incidence was dependent on HIV epidemic phase and scale. Throughout the epidemics, more than two thirds of HIV incidence occurred among individuals engaged in SCs. The majority of incidence among SCs was due to extramarital sex. To achieve reductions in HIV incidence, prevention approaches should target both SCs and individuals not in SCs.