Preventing healthcare-associated infections (HAIs) is both a critically important and methodologically challenging problem. Blinded, randomized trials are rare. Instead, the effects of interventions must often be estimated with patients are clustered on multiple levels, where interventions are co-occurring, and where a patient’s risk is a function of current (and past) prevalence of disease.
Computational models can be used, alongside statistics, to address some of these challenges. Of particular utility is the use of these models as a sort of facility-level decision support system, allowing the in silico design of studies, evaluation of interventions, etc. to be conducted to address methodological issues prior to the data collection process. This talk will explore several examples of this in practice, including the translation of statistical findings to a different scale using computational models, the evaluation of several study design problems, and methods for estimating and assessing the role seasonal staffing patterns might have in explaining the dynamics of several HAIs with weak, but consistently present, seasonal patterns.