Background: Transmission models for acute infections often assume stable contact patterns or perfect isolation during acute infection. The SickMix study collected longitudinal social contact data throughout the course of infection to better understand temporal variability among people experiencing acute respiratory infection and acute gastroenteritis. We developed mechanistic models to quantify how this variability impacts the estimation of key model parameters and intervention impacts for respiratory syncytial virus (RSV).

Methods: We developed age-stratified deterministic compartmental models with early and late infectious phases to capture changing contact patterns over the course of acute infection. We developed two versions of the model: a behavior naïve model assuming stable contact patterns over time, and a behavior dynamic model incorporating temporal changes in contact patterns among symptomatic cases. We fit these models to nationwide RSV hospitalization data from the 2018-19 and 2019-20 seasons to estimate transmissibility per contact. We simulated multiple interventions reducing disease infectiousness (e.g., masking) and individuals’ daily contacts (e.g., school closure) and compared the estimated impact between the two models.

Results: Estimated hospitalization rates for both seasons aligned well with reported data. The estimated transmissibility per contact in the behavior-dynamic model was 21% and 15% greater than the naïve model for the 2018-19 and 2019-20 seasons, respectively. Estimated impact of interventions varied between the behavior-naïve and behavior-dynamic models (Figure)

Conclusions: A lack of adjustment for reduced contacts during illness can result in underestimates of infectiousness and biases in case reductions from interventions. SickMix’s data allow for more accurate estimation of intervention impacts in infectious disease modeling.