Model-driven surveillance and intervention evaluation in highly stochastic healthcare settings

Washington State University, Pullman WA

Investigators

Abstract

Project Summary Healthcare-associated infections (HAIs) are a major source of morbidity and mortality globally. The CDC estimates that in 2011 they caused 721,800 infections in the United States, 75,000 of them fatal. The study of these infections is complicated by the inherently stochastic nature of hospitals. Even the largest hospitals are relatively small populations susceptible to stochastic variation in infection rates, patient risk profiles, and other drivers of transmission dynamics. The problem of stochasticity pervades every aspect of infection control ? from how evidence is gathered to how infection control teams are rated and evaluated. Ironically, the problem of stochasticity is exacerbated by improving levels of infection control ? as case counts drop, it becomes progressively more difficult to discern a clear signal indicating the success or failure of an infection control program from random noise. This challenge is compounded by other aspects of the healthcare environment. These aspects include the patient care mission demanding multiple interventions to control outbreaks, the need to capture nonlinear transmission dynamics, and the incomplete observation of many aspects of HAI infection dynamics. Computational models can alleviate some of these difficulties by simulating a hospital and the patients and staff within it. Within this simulated environment, novel interventions can be designed and tested ? these evaluations taking place thousands of times to capture and account for the effects of stochasticity. This project?s goal is to construct a flexible, high-fidelity model of HAI transmission that makes use of data and expertise from the Duke Center for Antimicrobial Stewardship and Infection Prevention?s network of community hospitals as well as Washington State University?s Veterinary Teaching Hospital. This model will be used to address three pressing clinical questions: 1) is there a level of hand hygiene wherein contact precautions are no longer necessary, 2) at what level of infections being imported from the environment are vertical versus horizontal infection control interventions optimal, and 3) can the interconnection between human and companion animal health be used to improve clinical infection control? The model itself will be an individual-based model that represents the spatial structure of the hospital, the movements of patients, staff, and others within that spatial structure; and disease transmission via either direct contact or contamination of the hospital environment. This model of the hospital will be linked to a general model for the population as a whole, as well as a similarly detailed model of a veterinary hospital. Once developed, this model will be capable of providing sophisticated decision support and advice to clinicians, administrators, and other public health decision-makers. !