PROJECT SUMMARY/ABSTRACTAntimicrobial resistant (AR) pathogens remain a major cause of healthcare associated infections (HAIs) in theUnited States. Indeed, the prevalence of these existing and emerging drug-resistant agents continues toimpose a heavy burden on U.S. healthcare systems. To better control existing AR pathogen-associated HAIsand prepare for the possible emergence of a novel AR organism, better, more targeted identification andintervention strategies need to be developed. Here, for this Modeling Infectious Diseases in HealthcareResearch Projects to Improve Prevention Research and Healthcare Delivery (MInD Healthcare) networkproject, we propose to develop a hierarchy of new model-inference systems capable of simulating andforecasting HAI outbreaks, quantifying individual patient colonization risk, and identifying optimal interventionapproaches. Specifically, we will use hospitalization records and diagnostic data for multiple AR pathogensfrom four major hospitals in New York City to conduct a series of modeling studies. We will develop twomathematical modeling structures: 1) a metapopulation model capable of simulating AR pathogen transmissiondynamics across multiple healthcare facilities; and 2) an agent-based model capable of simulating individual-level patient infection status, transmission dynamics, and movements within multiple hospitals. These modelswill be used in conjunction with Bayesian inference methods to simulate observed outbreaks of AR pathogens,estimate critical epidemiological characteristics and asymptomatic carriage probabilities among individualpatients, and support development of an AR pathogen forecasting system. As the models are high dimensionand the observations are sparse, new inference methods, capable of data augmentation and efficient modeloptimization, will also be developed. Additionally, we will use the optimized model structures to run freesimulations testing the effectiveness of six interventions: 1) hand hygiene and barrier precautions; 2) isolationof infections; 3) environmental cleaning; 4) active patient screening within hospitals; 5) contact tracing; and 6)screening at admission. These interventions will be tested singly and in bundles and used to inform targetedcontrol approaches. Further, we will develop a framework for identifying intervention bundles that maximallyreduce HAI rates given cost and logistical constraints. Lastly, we propose to collaborate with the CDC and theother research groups in the MInD Healthcare network to develop standardized intervention scenarios andinter-comparisons of simulated intervention outcomes among the different model forms used across thenetwork.
Inference; Forecasting and Optimal Control of Healthcare-Associated Infections
Shaman, Jeffrey L.; Pei, Sen