Infectious Disease
Incorporating Behavior Change into Estimates of Transmission Parameters and Intervention Impact in Acute Infections Kayoko Shioda* Dehao Chen Chen Chen Chen Chen Chen Emory University
Although behavioral changes in infectious individuals drive transmission, empirical contact data for acute infections remain scarce. Most transmission models for acute infections assume that contact rates remain constant or decrease to perfect isolation over the course of infection, introducing biases in key model-derived quantities—transmissibility per contact [β] and intervention impact. We aimed to characterize these biases. We conducted a longitudinal survey of social contacts in a cohort with acute respiratory or enteric infection from 2024-2025. We followed up with cases to measure their contact rates on the index date (when feeling most sick), 1 week, and 2 weeks after. We developed two versions of the mechanistic model for rotavirus: a behavior-dynamic model, incorporating our empirical data on behavioral feedback, and a behavior-naive model, assuming constant social contact patterns. We fitted both models to rotavirus gastroenteritis hospitalization rates in the US prior to vaccine introduction. We estimated β from both models. We simulated a non-pharmaceutical intervention of daycare closure by reducing contact among < 5-year-old cases. We found that contact rates among cases increased steadily from the index day through the 2-week follow-up. The dynamic model yielded β that was 1.3 times greater than that of the naïve model. Compared with the dynamic model, the naïve model overestimated the intervention impact (% reduction in cases with daycare closure vs. no closure) in the <5 age group by 3.1% when all infections (symptomatic and asymptomatic) were included, and by 14.3% when only symptomatically infected individuals stayed home. Overestimating contact rates leads to underestimating transmissibility per contact. In turn, the projected intervention impact for the target population—based on the underestimated β—is biasedly inflated.