Fact · Dictation for work and study
12,999clinician-annotated sentences
The error metrics in the npj Digital Medicine framework study were derived from 18 experimental configurations and 12,999 clinician-annotated sentences.
Checked against its source on October 16, 2026
In the source's own words
“Our clinical error metrics were derived from 18 experimental configurations involving LLMs for clinical note generation, consisting of 12,999 clinician-annotated sentences.”
A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation, Asgari E et al., npj Digital Medicine, 2025, Abstract.
Open the original (DOI 10.1038/s41746-025-01670-7)- What it measures
- The sentence count is the size of the annotated evidence base behind the hallucination and omission rates.
- What it does not show
- The sentences come from the authors' configurations. A larger, independent set of real visits would be needed to show how often a given commercial scribe makes these errors.