NEJM AI Study Reveals 84% Faster Resolution of Patient Messages with Clinician-Trained NLP Model
A new study published in the New England Journal of Medicine reveals that clinician-trained AI models can reduce EHR inbox burden by 40%, dramatically improving message response times and reducing administrative workload for health care teams. One area where physicians need support in fighting burnout is their inbox—today, they receive 57% more patient portal messages than they did before the COVID-19 pandemic. (1)
Conducted in partnership with Emory University investigators with support from Switchboard, MD data scientists, the study demonstrated an AI model that uses natural language processing to automatically categorize and route patient messages in real-time, achieving the following:
Health care staff had 2 fewer message interactions per patient inquiry, cutting unnecessary handoffs by 40% and ensuring messages reached the right team instantly.
Conversations were completed 22.5 hours sooner, an 84% reduction in median resolution time.
97.8% model accuracy in message classification across all five categories.
The model tested in this study was an early version of Switchboard, MD’s MDAware solution, which has since evolved into a more advanced AI-driven communication platform and was later adopted by Emory Healthcare. Researchers trained this initial model on thousands of clinician-labeled EHR messages across five categories: urgent, clinician, refill, schedule, and form-related requests.
The routing system was deployed across four outpatient locations at Emory Healthcare. Researchers compared 469 routed patient conversations against a parallel control group of 402 unrouted conversations from the same period to measure the impact of automated message triage.
“As a practicing physician, I know that AI should enhance, not replace, the human connection to medicine. By integrating clinical context into our message routing model, we’re not just improving efficiency—we’re enabling better patient care and reducing burnout without sacrificing the patient experience.”
Blake Anderson, MD
Founder, CEO, CTO, Switchboard, MD
Internal Medicine Physician, Emory Healthcare
This study provides new evidence on how AI can integrate into real-world clinical workflows to address longstanding challenges in patient communication management—creating new opportunities for automation and future clinical use cases.
Read the full analysis, including detailed metrics and network visualizations:
References
1. Zarefsky M. What’s adding to doctor burnout? Check your patient portal inbox. American Medical Association. February 3, 2023. Accessed February 27, 2025. https://www.ama-assn.org/practice-management/digital/what-s-adding-doctor-burnout-check-your-patient-portal-inbox.
