AI Scribes Reduce Documentation Time in ERs but Show No Measurable Effect on Patient Throughput

AI scribes using ambient listening and large language models have been rolled out in select emergency departments to address clinician burnout from electronic health record tasks. In one 18-month pilot across 12 sites, physicians saved an average of 1.8 hours per shift on notes, according to internal metrics shared with STAT. However, these time savings did not translate into faster discharges or reduced hallway boarding, which remained driven by inpatient bed availability and nursing ratios.

Observational data from two academic centers published in JAMA Network Open in 2024 similarly showed scribe-related documentation improvements without altering key operational endpoints such as door-to-provider time or left-without-being-seen rates. Resource constraints and workflow fragmentation appear to limit downstream efficiency gains. Broader system issues, including boarding driven by hospital capacity rather than documentation speed, dominate ER bottlenecks.

Next steps require randomized trials that pair scribe implementation with concurrent operational interventions such as inpatient flow optimization. Without those, documentation relief may improve clinician retention but leave core capacity problems untouched.