Time for Better Diagnosis: Measuring Outcomes, Stress and Time proposal (MOST)

Brigham And Women'S Hospital, Boston MA

Investigators

Abstract

Time for accurate and timely diagnosis is critical yet increasingly constrained. It is important for ensuring patients have adequate time to share their histories, and for clinicians to collect and reflect on patient data, and communicate and document their diagnostic assessments. Both patients and clinicians report feeling insufficient time and rushed in their diagnostic encounters, with clinicians increasingly reporting time pressures, after-hours work, and lack of control over time needed to care for patients and write their notes. High quality diagnosis for individual clinicians and health systems requires not only adequate time but also the ability to have feedback from patients to continuously learn about diagnostic outcomes and experiences. A team of clinicians and health services researchers who have engaged in more than two decades of work studying diagnostic safety will work with systems engineers to deepen our understanding of relationships between diagnosis quality (processes, outcomes, communication with patients) and time and stress. We will also study a natural experiment, comparing diagnostic encounters at two sites, one that allocates 20 minutes (Hennepin HealthCare) vs. another allocating 30 minutes (Brigham and Women’s) for primary care encounters. The project will leverage new EMR technologies that can: a) screen patients for symptoms to identify diagnostic encounters and collect downstream feedback from patients, b) capture detailed timestamp information recording each keystroke before, during, and after these diagnostic encounters, and c) record and transcribe clinical encounters and using generative AI technology to automatically produce clinicians’ notes. Studying diverse clinics at these two sites, the project has the following three specific aims: 1. Identify 400 patients with diagnostic visits and solicit their feedback on diagnostic processes and outcomes. 2. Collect and analyze data examining relationships between EMR-measured time, clinician stress, and diagnostic outcomes. This includes using large-scale EMR timestamp data from 40,000 visits, deploying a new Assessing the Assessment tool, the SaferDx tool, and validated MiniZ clinician stress measures. 3. Collect transcripts generated from 100 AI-documentation-recorded diagnostic encounters and conduct qualitative analysis of diagnostic communication conversations during the visit, as well as evaluate the features and quality of the AI-generated notes produced by two commercial vendors. Detailed information from each encounter will permit evaluation of key time relationships (e.g. how far behind in schedule clinician is for that encounter, after-hours time), rich evaluation of key elements in the clinical assessment (e.g. diagnostic uncertainty and differential diagnosis, psychosocial issues, don’t miss diagnoses), and correlation with clinician stress (e.g. burnout, perceived time pressures, cognitive load). Triangulating these data sources and analyses, the project will advance our understanding of time required to evaluate common symptoms, along with the quality of these diagnostic assessments and of AI-generated notes.