890 Modeling Transmission in a Multi-ward Hospital

Sunday, March 21, 2010
Grand Hall (Hyatt Regency Atlanta)
Molly Leecaster, PhD , University of Utah School of Medicine, Salt Lake City, UT
Karim Khader , University of Utah, Salt Lake City, UT
Tom Greene, PhD , University of Utah School of Medicine, Salt Lake City, UT
Michael Rubin, MD, PhD , IDEAS Center, Salt Lake City, UT
Matthew Samore, MD , IDEAS Center, Salt Lake City, UT
Background: Mechanistic modeling of the underlying process of infectious disease transmission has the potential to provide enhanced understanding of the spread of bacteria within hospitals and improve estimation of the effect of interventions. However, current models for hospital transmission are limited either by the number of patients in the model or the complexity of the model itself. Objective: The objectives of this work were to implement models for large hospitals and to incorporate between ward transmissions. Methods: Models of transmission assume that colonized patients interact with uncolonized patients and become colonized at an unknown transmission rate. We develop a transmission model with distinct within- and between-ward transmission rates which also incorporates the importation rate and the sensitivity of laboratory tests to confirm colonization. Observed hospital data were assumed to include stay (admission and discharge) and transfer information as well as test dates and results, but not true colonization status, which was treated as unknown (augmented data) and estimated along with the model parameters. The multi-ward hospital transmission model was specified by likelihood function that described the importation of colonized patients, transmission events, and laboratory testing. Using a Bayesian approach, the model parameters were assumed to have diffuse prior distributions to ensure that the observed data played the primary role in determining the results. Parameters, including augmented data, were estimated using Markov chain Monte Carlo methods. The model was assessed using simulated data that allowed knowledge of the true augmented data and parameter values. Various values were specified in the simulations and compared to estimates from the model. Results: The model was run on 100 simulated scenarios with hospital sizes up to 2000 patients. The model estimates closely reproduced the values specified in the simulations. The model of transmission improved on conventional methods to estimate prevalence and incidence by providing these as well as estimates of two transmission rates and an importation rate. Conclusions: Improved transmission models provide information for decision-makers beyond estimates of prevalence and incidence. We have shown that this extended model provides accurate estimates of transmission parameters and separates the within and between ward transmission events for estimating prevalence and assessing interventions. The model can be used for large hospital scenarios and uses generally available hospital information. Future work includes incorporating organism strain information and differentiation between clinical and surveillance culture results.