Structured chronic disease review

The problem

Reviews drift. The same clinician does a thorough review in a quiet clinic and a partial one in a busy clinic, and the difference is invisible until an audit finds it.

The steps

1. Patient record, for the history and current management.
2. Sources, pointed at your protocol for that condition. This is what decides what the review covers and what the targets are, rather than Rhazes deciding.
3. Calculate the measures the review depends on. Put every number here rather than in a reasoning step, so the arithmetic is identical on every patient and the working is kept for checking.
4. Branch on whether the targets in your protocol are being met, so a patient at target and a patient off target get different handling instead of the same paragraph.
5. Ask Rhazes for the review content on each path, drawing on your protocol.
6. Document, using the Chronic Disease Review template.
7. Create tasks for monitoring now due and anything the review flagged.

Why the calculation step matters here

A model asked to work out a score will usually get it right and occasionally will not. For a number that decides whether a patient is at target, usually is not good enough. A Calculate step removes that risk rather than reducing it, and keeps the working so the result can be checked later.

Why the branch matters

Without it the workflow writes one review for everyone and leaves you to notice which patients need something different. With it, the workflow surfaces that difference itself.

Set the else path deliberately. It runs for the patient who fits none of your cases, and that is usually the one worth the closest look.

What to check

The calculated numbers against the source data, at least while you are learning to trust the workflow. Every clinical recommendation, before acting on it. And whether the protocol the Sources step points at is still the current version.