Back

Advancing Woman-Centered Global Health Simulations: Improved Accuracy Through Detailed Demographic Modelling in FPsim

Driano, E. K.; Sanz-Leon, P.; Stuart, R.; O'Brien, M.; Zimmermann, M.

2025-09-18 sexual and reproductive health
10.1101/2025.09.17.25336008 medRxiv
Show abstract

BackgroundFPsim is an agent-based model designed to simulate individual family planning (FP) behaviors and outcomes, capturing biological and behavioral heterogeneity among women. The initial release of the model enabled users to estimate the impact of new contraceptive technologies and interventions, but it lacked key contextual factors--particularly those linked to womens socioeconomic status--that influence contraceptive dynamics. In this work, we enhanced FPsim by incorporating additional demographic attributes, including education level, wealth quintile, and urban/rural residence. These attributes influence, and are influenced by, contraceptive behaviors, allowing for a more nuanced simulation of family planning outcomes. MethodsThe key enhancements to the FPsim model are/were: (1) refining the behavioral algorithms that determine a womans probability of contraceptive uptake, method choice, and--importantly--the duration a woman remains on a given method, and (2) integrating additional social and demographic factors to better reflect real-world contraceptive choice. We added demographic attributes of education, urban/rural residence, and wealth quintile. By capturing how social context influences duration and switching behavior, FPsim v2.0 better reflects real-world contraceptive trajectories and heterogeneity across population subgroups. ContributionFPsim v2.0 advances a more comprehensive and person-centered approach to modeling FP behaviors. The new model embeds additional social and demographic diversity into individual agent profiles, offering the ability to simulate more tailored and effective interventions. Further validation in diverse settings is necessary to generalize these findings and maximize the models utility.

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.