Do Doctors Still Use Medical Transcriptionists?

Two things happened at once. Speech recognition became good enough to produce a usable first draft, and electronic health records made structured template entry the default for routine visits. Together those removed most of the straightforward typing volume: a clinician clicking through a templated well-visit does not need a transcriptionist, and never really did.

What did not go away is the documentation that resists templating. A complex consultation, a procedure note, a difficult history - these are narrative, and narrative is where recognition errors concentrate and where templates do not help. That work still gets dictated, and it still gets reviewed.

The pattern is fairly consistent. Specialties with dense or unusual vocabulary - oncology, radiology, pathology, orthopaedics - see higher recognition error rates and so get more value from review. Clinicians who dictate fast or with an accent see the same. Practices doing high narrative volume outsource because the alternative is clinician time, which is the most expensive hour in the building.

The economics are usually clearer than they look. A transcription service costing a few dollars per note competes against a clinician spending fifteen minutes correcting a draft, and the clinician loses that comparison at almost any realistic rate.

AI drafting introduced a documentation risk that did not exist when a person typed from a recording. A model produces fluent text, so its errors read correctly - and the tools generate timestamps that can evidence the absence of the review a signature asserts. A note signed nine seconds after the encounter closed is not evidence of review; it is evidence of its absence, and CMS audit posture now treats it that way.

That shifts what a practice should want from a service. Not just an accurate transcript, but a record of who checked it and when - because the transcript is the easy part and the attestation is what an audit or a claim actually asks for.

Separate the documents. Routine templated visits probably need nothing. Narrative documentation that enters the chart needs a reviewed transcript and a record that review happened. Internal material - meetings, notes to self, research - needs neither, and paying for review on it is waste.

ScribeForms is built for exactly that split: review is a per-job choice at a stated rate, so the same account handles both without a separate contract, and the document records which mode produced it so the distinction survives into the record.

There are four realistic options and they price very differently. An in-house transcriptionist is a salary plus overhead, and only makes sense at sustained volume. An outsourced service is per minute, scales with use, and carries no fixed cost. Unreviewed speech recognition is cents per minute and shifts the correction work onto the clinician. A scribe is the most expensive per encounter and the only one that also removes the documentation burden during the visit itself.

The comparison most practices get wrong is the third against the second. Speech recognition looks nearly free, but the correction time lands on the clinician, and clinician time is the most expensive hour in the building. A service costing a few dollars a note beats fifteen minutes of physician attention at essentially any rate a practice might assign to it.

The work is now editing rather than typing, which is a higher-skill task at lower volume. A transcriptionist reviewing an automated draft has to know which errors a model makes - the phonetically similar drug name, the normalised dose, the inferred laterality - because those are the ones that survive a casual read. Catching them needs the terminology knowledge the old job required plus a different kind of attention.

It also means the quality of review is harder for a practice to assess from outside. A clean-looking transcript tells you nothing about whether anyone checked the figures in it. That is why the record of review has become as important as the review itself: a signature over a digest of exactly what the reviewer saw is checkable in a way that "we have a QA process" is not.