Agent-Based Model Using NATSAL-3 Calibration Shows Concurrency and Bisexuality Drive STI Network Structure

The model implements partnership formation and dissolution probabilities stratified by age, sex, and sexual orientation, including bisexual individuals, with monogamous and concurrent ties dissolving via duration-dependent hazards. Latin Hypercube Sampling generated parameter sets, and the lowest-MSE combination was selected against NATSAL-3 targets for mean partnerships, concurrency prevalence, and duration distributions. This approach directly tests whether individual-level heterogeneities can generate the observed population-level contact network without assuming fixed aggregate statistics.

Static models commonly used in STI epidemiology treat concurrency as a fixed attribute and therefore miss how age-specific partnership turnover and bisexual bridging reshape network connectivity over time. The agent-based framework reveals that even modest increases in concurrent partnerships among younger age groups produce disproportionate rises in component size and epidemic potential, consistent with earlier empirical findings from the UK and US that concurrency explains residual variance in gonorrhea and chlamydia incidence after controlling for partner number.

Compared with prior NATSAL-based analyses that reported concurrency at roughly 12-15% in men, the calibrated model identifies bisexual men as critical bridges whose removal reduces giant-component size by 18-22% in sensitivity runs. This suggests targeted interventions could exploit orientation-specific mixing patterns rather than broad concurrency reduction.

Next steps include coupling the network to an STI transmission layer and validating against 2025-2027 surveillance data from England; divergence above 10% in predicted incidence would indicate missing behavioral feedback mechanisms.