OBJECTIVES: Agent based models (ABMs) have been increasingly used for infectious disease simulation, yet validating these models remains challenging because of limited data. Our objective was to validate an ABM of hospital-acquired Clostridioides difficile infection (CDI) using primary hospital data. We then used the model to investigate the impact of a CDI testing strategy that seeks to reduce falsely classified hospital- and community- associated CDIs.

METHODS: We created and validated an ABM of CDI in the UW Health hospital in Madison, Wisconsin. Most parameters were derived from primary hospital data, while few unobservable parameters governing progression of CDI were estimated through calibration. We specifically focused on representing the decreasing trend in CDI between 2013 and 2017. We then used our model to quantify the HA-CDI rate and number of incorrectly classified hospital (HA) and community (CA) acquired CDIs with and without the 2016 implementation of a CDI testing strategy.

RESULTS: We found that our model was able to replicate CDI trends from 2013-2017 (2013: 15.71, 2014: 13.60, 2015: 12.77, 2016: 6.82, 2017: 6.62 modeled vs 2013: 12.14, 2014: 13.70, 2015: 14.01, 2016: 7.58, 2017: 5.78 observed per 10,000 patient days). For example, our model showed a 47% decrease in HA-CDI (12.77 to 6.82) vs the observed 46% decrease (14.01 to 7.58) from 2015 to 2016. Under baseline conditions, use of a testing strategy in 2016 was associated with an additional 4.2%, 3.7%, and 0.9% decrease in HA-CDI per 10,000 patient days, false HA-CDIs, and false CA-CDIs, respectively.

CONCLUSIONS: We used primary hospital data to validate an ABM, which may be a useful tool to estimate the relative impact of infection control interventions. This study also suggests that CDI testing strategies may decrease false HA- and CA-CDI reporting with little negative impact on the overall rate of infection.