A key to optimal management of COVID-19 is development of evidence-based recommendations and associated strategies to ensure implementation of treatment recommendations. This is particularly important when evidence is emerging as rapidly as is the case for COVID-19. GRADE (Grading of Recommendations Assessment, Development and Evaluation) has emerged as the leading system for rating the quality of evidence and strength of recommendations. Endorsed by more than 110 professional organizations, GRADE has codified key normative factors that clinical practice guidelines (CPGs) panels ought to take into consideration. During activities conducted over the last 4 years on our parent R01 grant (5R01HS024917), we have discovered that in addition to normative GRADE factors, important non-GRADE factors affect the group judgment of CPG panels. We now propose to leverage these findings from the parent R01 grant to help generate optimal management strategies for COVID-19. To provide the most rational framework for managing COVID-19 patients, we propose to develop GRADE-based CPGs that we will implement in the electronic medical record (EMR) at the point-of-care within the Rush health system in Chicago. An increasingly popular strategy for improving patient care is to standardize care by translating CPGs into clinical pathways (CPs), which typically use flow charts or clinical algorithms to provide detailed steps about a course of management for a particular clinical problem or an entire spectrum of care. However, despite the promise of CPs and their increasing use, no theoretical framework has been developed to guide their development. This means it is not possible to rigorously analyze the efficiency of CPGs/CPs, nor their influence on patient health outcomes. We hypothesize that solid theoretical grounds for developing CPGs/CPs can be provided by converting them into fast-and-frugal decision trees (FFTs). FFTs are constructed as a series of sequentially- ordered, clinical information or ?cues? whose relation is defined by a series of if?then statements. Every cue in an FFT can correctly or incorrectly classify a signal (e.g., patient has COVID-19) vs. noise (e.g., patient does not have COVID-19) and this classification pattern can be measured (e.g., is the signal a true positive or negative). This property of FFTs allows them to be integrated within a broader theoretical framework of signal detection and related theories which, in turn, allows the accuracy of the clinical strategies they represent to be evaluated. In this application, we propose to develop GRADE CPGs for COVID-19 (Aim 1), translate the CPGs into CPs, and, convert the CPs into FFTs. Subsequently, we will implement FFTs in the Rush EMR (Aim 2), and conduct an interrupted time series to evaluate the effect of GRADE-based FFTs on management of patients with COVID-19. The proposed application is directly informed by the parent R01 grant and has potential for immediate and sustained impact to improve clinical management of patients with COVID-19.
The proposed project aims to improve clinical management and treatment outcomes in patients with COVID-19 by developing an evidence-based, decision support system. We will initially implement this approach in the electronic medical record at the point-of-care within the Rush health system in Chicago, but it could be rapidly deployed to other diseases and health systems across the nation if effective.
