AI That Frees Clinicians: How a Detroit Trauma Service Cut Charting Time by 60%

Sinai-Grace Hospital is a Level I trauma center in Detroit, part of the Detroit Medical Center. It handles over 4,000 critical cases a year. The trauma team runs 24/7 with attending surgeons, residents, nurses, and support staff rotating through high-pressure shifts where minutes matter.

Sinai-Grace operates with its own protocols, documentation standards, and chain of command. Its reputation was built on clinical outcomes, not technology adoption — and the staff had earned a healthy skepticism of tools that promised to help but added friction.

Trauma clinicians were spending more time charting than treating. After a critical case, attendings and residents sat down to document what happened — injuries, interventions, vitals, decision rationale, medications — often reconstructing events from memory hours after the fact.

The EHR required structured fields, dropdown selections, and narrative notes for every encounter. Clinicians copy-pasted from prior records, left fields incomplete, or dictated notes that a transcriptionist cleaned up the next day. The result: documentation that was slow, inconsistent, and frequently inaccurate.

Leadership knew the documentation problem was real, but past attempts at “AI charting” had failed. Tools that auto-generated notes hallucinated details. Templates that tried to standardize input slowed clinicians down. Nobody trusted the output, so nobody used it.

A documentation layer that listens during trauma activations and drafts structured chart notes in real time — reviewed and approved by the clinician, never auto-submitted.

• Clinical workflow mapping — observed and documented the actual sequence of a trauma activation from door to disposition, identifying where charting happens, where it stalls, and where information gets lost
• AI-assisted drafting — built a model pipeline that captures ambient audio during activations and produces structured EHR-ready notes within minutes of case close
• Clinician-in-the-loop — every generated note requires review and approval before it touches the record; flagged fields highlight uncertainty so clinicians correct rather than accept blindly
• EHR integration — connected directly to Epic so approved notes populate the right fields without re-entry or copy-paste
• Zero protocol disruption — the system fits around existing triage workflows; nothing about how clinicians treat patients changed