Multiscale Modeling of Patient Movement to Determine Effects of Surveillance on Healthcare-Associated Infections

Abstract

Background: The transmission of pathogenic organisms in healthcare settings is a major cause of healthcare-associated infections (HAIs). In recent years, infections with carbapenem-resistant Enterobacteriaceae (CRE) have become a significant public health threat, in part because many patients are arriving at the hospital already colonized, and colonization is a major risk factor for infection. Reducing transmission requires understanding how patient movement drives the spread of CRE; however, analysis of this issue has mostly been modeled at a hospital-level without much consideration for the population dynamics that occur outside of the hospital setting and how patients move between healthcare settings. Patients move between hospitals, other healthcare settings, such as long-term care facilities (LTCFs), and the community, all of which pose different colonization risks. Thus, studying each environment in isolation fails to realistically address the consequences of large-scale policy interventions. One such intervention is a statewide electronic registry to track patients who are known to be colonized or have had a CRE infection. Understanding the potential for reducing CRE morbidity and mortality requires consideration of small- and large-scale effects on patients’ movement and transmission. Methods: We developed a multiscale, metapopulation model for hospitals, communities, and LTCFs in the state of Maryland. In our computational simulation, we included a regional- as well as a local-scale model that were informed by the patient-mix data from the Maryland Health Service Cost Review Commission. We examined the impact of implementing a registry compared to less coordinated scenarios. Results: The most effective policy was the implementation of an electronic registry which resulted in 9.6% median reduction in CRE HAIs in Maryland for simulated outcomes (Fig. 1). Other interventions included colonization screening at various or all hospitals and using a predictive algorithm to determine at-risk patients that need to be screened. These interventions only resulted in ~1%–3% reductions in HAIs. We also observed that coupling other interventions with an electronic registry does not aid in reducing more HAIs.

Funding: None

Disclosures: None

Boxplot of simulated outcomes of all scenarios compared to the baseline scenario. The scenarios include (1) complete screening with no electronic registry, (2) selective screening with no electronic registry, (3) predictive screening with no electronic registry, (4) baseline with an electronic registry, (5) complete screening with an electronic registry, (6) selective screening with an electronic registry, and (7) predictive screening with an electronic registry. Scenarios 5-7 include the same interventions as 1–3 coupled with an electronic registry.