In silico Randomized Control Trial Framework for Assessing Infection Control and Prevention Interventions in the Hospital

项目总结/摘要 多重耐药微生物（MDRO），特别是碳青霉烯类耐药微生物（CRO）， 是卫生保健相关感染（HAI）的主要原因。尽管研究发现 应对MDRO定植和传播的多种干预措施， 减少HAI，由于不确定最有效的方法， 干预措施的组合。由于患者在医疗机构内通过 医护人员（HCW）和移动的设备的移动， 对传播和感染的影响必须考虑这些网络动态。此外，每个 机构有自己的人员配备比例、设备管理和清洁做法， 告知干预的有效性。解决关于最有效的 在个别医院的干预策略组合，我们将建立一个模型， 该框架将为医院管理人员和感染控制专家提供工具， 量化个体患者定植和感染的风险因素，并比较 干预措施对控制HAI的潜在有效性。该项目的目标是：（1） 开发模型来预测哪些患者的定植和感染风险最高， MDRO使用正常护理过程中收集的临床相关信息 结合人员配置和设备移动的业务数据;（2）利用医院- 水平模型，以量化之间的组合关系，和边际影响， 减少HAI的额外干预措施;（3）为其他医院建立框架， 医疗保健系统检查干预措施的有效性， 机构的数据。模型将以患者为中心，利用丰富的协变量数据 存储在电子健康记录中，以开发基于先进机器的预测模型 学习这些预测模型将成为可翻译的工具， 以帮助临床医生预防HAI。患者连接性数据将 形成医院级模型的基础，允许对 不同干预措施组合的有效性。最后，一个可推广的框架将 在医疗机构网络中构建和测试。这些模型将直接 告知CDC关于MDRO预防的指南，并帮助临床医生降低HAI的风险。

Project Summary/Abstract Multidrug-resistant organisms (MDROs), particularly carbapenem-resistant organisms (CROs), are a major cause of healthcare-associated infections (HAIs). Though studies have found multiple interventions addressing MDRO colonization and transmission to be effective at reducing HAIs, implementation has lagged due to uncertainty regarding the most efficacious combinations of interventions. As patients are connected within healthcare facilities by the movement of healthcare workers (HCWs) and mobile equipment, evaluation of an intervention's impact on transmission and infection must consider these network dynamics. Additionally, each institution has its own staffing ratios, equipment management, and cleaning practices that inform the efficacy of an intervention. To address knowledge gaps regarding the most effective combinations of intervention strategies at an individual hospital, we will build a model framework that will provide hospital administrators and infection control experts with tools for quantifying individual patient risk factors for colonization and infection and for comparing the potential effectiveness of interventions to control HAIs. The aims of the project are: (1) to develop models to predict which patients are at highest risk for colonization and infection with MDROs using clinically relevant information collected during the normal course of care combined with operational data on staffing and equipment movement; (2) to utilize hospital- level models to quantify the combinatorial relationship between, and marginal impact of, additional interventions to reduce HAIs; and (3) to build a framework for other hospitals and healthcare systems to examine the effectiveness of interventions parameterized by their own institution's data. Models will start with the patient at the center, utilizing the rich covariate data stored in electronic health records to develop prediction models based on advanced machine learning. These prediction models will be translatable tools that can be directly incorporated into clinical care to assist clinicians in preventing HAIs. Data on patient-connectedness will form the foundation of hospital-level models that will allow for detailed examinations of the effectiveness of different combinations of interventions. Finally, a generalizable framework will be constructed and tested across a network of healthcare facilities. These models will directly inform CDC guidelines about MDRO prevention and aid clinicians in reducing the risk of HAIs.