Determine the prevalence of examination fabrication in examination-free consultations

Determine how many consultations without a physical examination produce no fabricated examination findings in notes generated by the three audited commercial ambient scribes, thereby enabling a rate rather than only a count of such failures.

Background

The study identifies 42 verified findings in which consultation history was documented as an objective examination, including cases from telephone consultations where no examination occurred. However, the corpus does not reliably label which other consultations were remote or examination-free. Consequently, the authors report the ten consultations containing this failure as a count and explicitly leave the denominator—and therefore the prevalence of the failure—unresolved.

References

The corpus carries no reliable label for which of its other consultations were remote, so how many examination-free consultations produced no such finding is unknown, and we state this as a count, not a rate.

— One note in three: a verified census of three deployed AI scribes, and the instrument that counted it  (2608.31017 - Fox et al., 31 Aug 2026) in Section 3.4, “The next three: dropped diagnoses, invented dates, invented examinations”