Network Models, Patient Transfers, and Infection Control

Residents of long-term care facilities are at high risk for acquiring multidrug-resistant organisms [1–5]. These individuals are often admitted to acute-care hospitals, promoting the propagation of multidrugresistant (and other) infections to other patients. Thus, hospital transfers represent an important potential point of intervention, where informed decisions can be made to control the spread of multidrugresistant organisms, if only one could sufficiently understand the complex interactions that enable the spread of infections. Ray et al’s article “Spread of Carbapenem-Resistant Enterobacteriaceae Among Illinois Healthcare Facilities: The Role of Patient Sharing” in the current issue of Clinical Infectious Diseases uses a model constructed from healthcare data to help elucidate the relationship between multidrug-resistant organisms and the sharing of patients between long-term care facilities and acute-care hospitals [6]. Scientists and engineers use models to understand the behavior of complex systems in many fields. Obviously, any resulting model-based predictions are only as good as their underlying models are faithful to reality; in general, finer-grained models support more refined predictions. In epidemiology, relatively coarse epidemiological models based on randommixing (ie, the assumption that every member of the population is at similar risk) have long been used to predict how infections spread across large populations [7]. In reality, interactions between healthcare workers are not uniformly random [8], individual healthcare workers come into contact with diverse sets of patients [9–11], and some hospitals are much more likely to transfer patients to or from other hospitals [12]. These inconsistencies with the underlying model assumptions may compromise the effectiveness of interventions based on that model. For example, vaccinating a particular subset of healthcare workers (ie, thosewho aremore densely connected with patients and other healthcare workers) may be much more effective than vaccinating healthcare workers selected at random [8]. In addition, increasing hand-hygiene adherence may be much more important among healthcare workers with more direct contact with patients [9, 13], and networks via peer effects may even influence healthcare worker behavior [14]. Thus, random mixing models do a poor job of predicting both how infections spread within and across hospital facilities and how effective a particular intervention meant to prevent, halt, or slow the spread of an infection might be. As a consequence, interventions designed to interrupt the spread of multidrug-resistant organisms may be much more effective if patterns of connectivity between healthcare and long-term care facilities are explicitly considered in the design of the intervention. Ray et al clearly show that a higher volume of transfers between long-term care facilities and hospitals is associated with higher rates of extensively drug-resistant organisms (XRDOs), providing much needed empirical evidence for the importance of considering long-term care facilities when designing infection-control practices in closely associated hospitals. The authors highlight the importance for the infectioncontrol community to gain not only a greater understanding of the flow of patients between facilities but also the need for a framework to understand how potential interventions may be affected by patient transfers. Thus, while Ray et al focus on patients “shared” between long-term care facilities and hospitals, the implication that hospital transfers amplify the spread of healthcare-associated infections extends well beyond the facilities considered. Of course, the association between hospital XRDO rates and transfer volumes described by Ray et al may not actually be caused by the transfers. The associations observed could be due to unconsidered factors such as (unmeasured) severity of illness or the patient’s transfer history prior to the present transfer. Hospitals with higher XRDO rates may receive a greater volume of transfers from long-term care facilities or may admit more critically ill and/or frail patients, both of whom are more likely to acquire XRDOs. In addition, as Ray et al were unable to adjust for the direction of patient transfers, additional work should examine the direction of transfers, patient transfer history, severity of illness, and similar elements. Even so, the causation that Ray et al propose is surely epidemiologically plausible. Simmering et al showed that indegree (defined as the number of hospitals from which transfer patients are accepted into a specific hospital) was significantly associated with Clostridium difficile rates, even Received 27 June 2016; accepted 28 June 2016; published online 2 August 2016. Correspondence: P. M. Polgreen, Carver College of Medicine, Department of Internal Medicine, University of Iowa, Iowa City, Iowa 52242 ([email protected]). Clinical Infectious Diseases 2016;63(7):894–5 © The Author 2016. Published by Oxford University Press for the Infectious Diseases Society of America. All rights reserved. For permissions, e-mail [email protected]. DOI: 10.1093/cid/ciw465