Dictation as an authored channel

A dictation is a voicemail with the speaker flipped. A voicemail is an anonymous caller whose message you route to a mailbox; a dictation is a known clinician whose spoken note you file to a patient chart. Two things change, and everything else is shared.

The same pipeline, a third source

Classification, the routing rules, the review queue, auto-mode, and the audit trail do not care whether text came from a scanned page, a recorded call, or a clinician speaking at a workstation. So the inbound logic spans all three rather than forking. A fax is OCR then classify; a voicemail is transcribe then classify; a dictation is transcribe then classify. From the classify step on, the path is the same.

flowchart LR
  FAX[Fax] --> OCR[OCR] --> CL[Classify]
  VM[Voicemail] --> STT1[Transcribe] --> CL
  DICT[Dictation] --> STT2[Transcribe] --> CL
  CL --> RULES[Rules] --> RQ[Review queue]
  RQ --> AUTO[Auto-mode eligible?]
  RQ --> FILE[Human-confirmed filing]

What flips

Two things differ from a voicemail. The author is the signed-in clinician rather than an unknown caller, so a dictation is private to its author until it is filed, and there is no caller ID to block. And the routing target is a cosign or notify recipient, the person who should also see the note, rather than a delivery mailbox. Filing the note to the chart is a separate step from that routing.

One trust score across three channels

Auto-mode scores a rule’s eligibility over a union of its fax, voicemail, and dictation decisions, so a shared rule keeps one trust score across every channel instead of fragmenting into three. A rule that is clearly trustworthy from its combined history can earn its way into routing even if no single channel would have graduated it alone. That union is the whole point of keeping the channels in separate tables but the logic shared.

On-prem, like everything else

Transcription runs on your own hardware. Audio never leaves the building, the same rule that governs page images. Reading or streaming a stored recording, or playing the transcript back, is a PHI access and is audited.

The line that does not move

Auto-mode can route a dictation’s cosign target without a person. It can never file a note to a chart without a person confirming the patient, because filing needs an independent identity cross-check that a routing decision does not provide. A dictation files to MEDITECH the same human-confirmed way a fax does, with one difference that fits what it is: the transcript is rendered to a one-page note and filed as a medical document (an MDM message), not as a result.