| Abstract: |
| Human behavior plays a central role in shaping infectious disease transmission. As risk changes, people modify their mobility, adopt protective measures, and respond to health policies and rising case numbers. Despite this, most epidemic models still represent behavior only indirectly, either through fixed transmission parameters or by inferring behavioral effects from epidemiological trends alone.
In this study, we develop a modeling framework that explicitly incorporates measured behavioral change into epidemic dynamics and enables comparison across alternative behavioral compartmental structures. Using repeated survey data from the COSMO project in Germany, we quantify variation in risk perception and protective behavior, and link these responses to policy interventions using indicators from the Oxford COVID-19 Government Response Tracker. On this basis, we construct three compartmental formulations in which behavioral change is represented through transitions between behavioral states and through behavior-dependent modulation of contact patterns and transmission intensity.
Population mixing is modeled using age-structured contact matrices from POLYMOD, rescaled with time-varying mobility data from Google. All model variants are calibrated to weekly COVID-19 mortality data across 16 German states using Approximate Bayesian Computation with Sequential Monte Carlo. Our results show that explicitly incorporating behavioral dynamics improves model fit and captures regional and temporal heterogeneity effectively. |
|