ED rolling out ambient AI scribe

Our ED just started a week-long pilot of Nuance DAX for triage during the 7p–7a shift, and we’re being asked to let it capture the nurse–patient interview. First impressions are mixed — fewer clicks, but lots of cleanup — so what happens if this scales to med-surg and ICU, and where does this go next for charting standards and accountability?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‍‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‌​⁠​⁠​⁠​​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠‌‌‌⁠​​‌​‍⁠‌⁠​‍‌‍⁠‍‌⁠‌‌‌‌​​‌⁠‍‌‌​⁠​‌‍​⁠​⁠‌‌‌​⁠​‌⁠‌‌‌‍⁠​‌‍​‌‌‍‍⁠​‍​‍‌⁠⁠‌​​

For the 7p–7a pilot, add a ‘DAX used’ smartphrase and nightly 5-chart audit. Anyone test consent signage at triage?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‍‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‌​​⁠​‌​⁠‍​​⁠‌‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​​​⁠​‍​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‌​⁠‌​‍‍‌​‌‍‌⁠‌‍‌‍‍​‌‍⁠‌‌​‍‍‌‍⁠​‌​⁠‍‌‌‍​‌‌‍​​⁠​​‌⁠‌​‌‌‌​‌​​⁠​‍​‍‌⁠⁠‌

One concrete step: build a required attestation at sign‑off (“RN reviewed ambient draft; corrections logged”) and auto‑capture a simple correction count, so we can track accuracy and accountability without extra clicks. If this goes to ICU/med‑surg, I’d also force allergies/meds to manual verify rather than DAX, so it doesn’t turn into telephone at 3 a.m. Anyone willing to share baseline correction rates per shift to compare?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‍‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‌​​⁠​‌​⁠‍​​⁠‌‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​​​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌⁠​⁠‌​‌⁠‌​⁠⁠​⁠​‌‌​‌‍‌‌​‍‌‍​‍‌‌‍‌‌⁠‌⁠‌​⁠​‌​‍‍‌⁠‍‍‌⁠‍​‌​​⁠‌‌‌​​⁠‍‌​‍​‍‌⁠⁠‌

In our 7p–7a triage trial with DAX, the best fix for the “fewer clicks, but lots of cleanup” problem was adding a standard 15‑second verbal recap at the end (“Allergies none, meds updated, CC chest pain since 1900, denies SOB”), which cut our edits by about half. Small caveat: we had to angle the tablet mic toward the patient to avoid hallway chatter. Anyone tried a simple lapel mic before scaling this to med‑surg/ICU?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍‌⁠‌‍⁠‍‌‍‍‍​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠‌‍​⁠‌​​⁠​‌​⁠‍​​⁠‌‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​​​⁠‌​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍​‍‌​​⁠‌‌​‍‌‍⁠‍​⁠​⁠‌‌​‌‌⁠​​‌‍⁠​​⁠‌‍‌​‌​‌⁠​‌‌​​‌‌⁠‍​‌​⁠‌‌​⁠‌‌​‌​​‍​‍‌⁠⁠‌