Telehealth

Care delivered over video still produces a chart, a message and a bill. Staff work from home offices and shared devices, so the gateway is often the only common control point. This starter treats it that way. Every request must carry a verified identity, and every answer meets a scored floor before a clinician sees it.

Virtual visits produce a lot of drafting work. Visit summaries, patient messages and follow up instructions arrive faster than a clinician can write them. Each of those artifacts can reach a patient directly. That is a different exposure from an internal note, and it deserves a gate on the way out.

The request path runs in a fixed order for a reason. Injection screening comes first, because a message pasted from a patient portal may carry an instruction of its own. Identity enforcement follows and rejects an unverified caller. A health data stage brings record level matching to the shared redaction path, and masking hides member, patient and policy numbers.

Reimbursement rules ask for accurate documentation, not for a fluent paragraph. Output scoring answers part of that. A reply must score 0.88 overall, with relevance at 0.78 and accuracy at 0.80, and a failure is refused rather than trimmed. Anything under 120 characters or a single sentence fails the same check.

Three things stay in your hands. Nothing here approves a message for release to a patient, because the review point ships empty and escalates nothing until you configure it. Metadata only history keeps visit text out of the captured record, which fits a service that must not retain transcripts. Source checking flags weak grounding and still returns the answer. Coverage and billing decisions remain a policy matter outside the gateway.