SHM Converge Scientific Abstract ePoster Gallery-Development and Piloting of a Real-Time Ehr-Based Predictive Algorithm for Diagnostic Error

Pdf Summary

This study developed and piloted a real-time electronic health record (EHR)-based predictive algorithm to estimate a patient’s risk of diagnostic error (DE) during hospitalization. The goal was to improve clinician awareness of diagnostic uncertainty and encourage use of a “Diagnostic Time-Out” (DTO) when risk was elevated.<br /><br />The algorithm used baseline and dynamic EHR risk factors identified from literature and expert opinion, such as altered mental status, language barriers, inter-hospital transfer, multiple consultants, early readmission, recent ED visits, team changes, and increased oxygen needs. Patients were assigned a score that mapped to low (green), moderate (yellow), or high (red) DE risk flags in an EHR dashboard.<br /><br />To test whether clinicians found the risk flags meaningful, the team sent weekly yes/no polls to care team members for randomly selected general medicine patients. Clinicians were asked whether they would take a DTO for the patient. Of 99 patients polled (393 total polls), at least one response was received for 43 patients (43%). “Yes” responses increased with predicted risk: 7% for low-risk cases, 25% for moderate-risk cases, and 54% for high-risk cases. High-risk diagnoses such as altered mental status were significantly associated with “Yes” responses. When moderate or high-risk flags were used to predict a “Yes” response, the algorithm showed high sensitivity (92%) but low specificity (42%).<br /><br />Overall, the findings suggest clinicians were more likely to agree to a DTO as predicted DE risk increased, indicating the algorithm may help identify cases where diagnostic reflection is needed. Next steps include comparing predicted risk with actual chart-reviewed DE outcomes and refining the algorithm using larger datasets and adjudicated outcomes.

Asset Subtitle

Anuj K. Dalal

Meta Tag

Author List Alyssa Lam, Anuj K. Dalal, Daniel Motta-Calderon, Jeffrey L. Schnipper, Kevin Carr, Pamela Garabedian, Savanna Plombon

Category Innovations

Concept Prediction Algorithm

Concept Diagnostic Error

Concept Patient Risk

Concept Risk Factor

Concept Perioperative Mortality Risk Prediction Tool

Distinguished Non-Finalist

Presenter Organization Brigham and Women's Hospital

Presenter Organization Harvard Medical School

Presenting Author Anuj K. Dalal

Track Technology in Hospital Medicine

Keywords

diagnostic error

predictive algorithm

electronic health record

hospitalization

diagnostic time-out

clinical decision support

risk stratification

diagnostic uncertainty

altered mental status

real-time dashboard

Prediction Algorithm

Diagnostic Error

Patient Risk

Risk Factor

Perioperative Mortality Risk Prediction Tool