Practice operations
From Annual Wellness Visit Findings to Completed Follow-Up
Caesar Health · September 11, 2026
After an Annual Wellness Visit, the practice needs a way to connect each finding to a clinician-reviewed decision, an assigned owner and a documented result. A referral sent, an appointment scheduled and care completed are different milestones. Tracking them separately makes unfinished work visible.
The Medicare Annual Wellness Visit includes a Health Risk Assessment and a personalized prevention plan. It is different from a routine physical examination. CMS describes the visit and its requirements in its Medicare Wellness Visits guidance.
This article describes a workflow practices can evaluate. It develops the operational idea behind Caesar Health’s comment to CMS on AI and Annual Wellness Visits. It is a proposed approach to coordination, not evidence that a particular system has improved clinical outcomes.
Give each finding a next step
A finding in a questionnaire or visit note is the starting point. The responsible clinician determines whether it calls for further assessment, an order, a referral, monitoring or no additional action. Staff need to see that decision without having to infer it from a long narrative.
We recommend maintaining a record with these fields:
| Field | What the team should be able to see |
|---|---|
| Finding and source | What was reported, when it was recorded and where the original information lives. |
| Clinical review | Who reviewed the finding and the decision they made. |
| Agreed action | The next step approved by the clinician and discussed with the patient. |
| Responsible owner | The person or team accountable for moving the work forward. |
| Timing | The clinician-directed timeframe and next review date. |
| Current status | Whether the action is recommended, approved, ordered, scheduled or completed. |
| Result or disposition | Evidence of completion, patient declination, a clinical decision to close, or the reason work remains open. |
The practice should define urgency and escalation through its clinical protocols. A single default deadline cannot express the difference between an urgent concern and a routine follow-up.
Separate scheduling from completion
Consider an illustrative workflow: a patient raises a concern during the AWV, and the clinician recommends a referral after reviewing it. A coordinator sends the referral and confirms that the receiving practice has it. The patient schedules an appointment. The team then checks for evidence that the visit occurred and routes the resulting information for clinical review.
Each step answers a different question. Sending the referral proves that the request went out. Scheduling proves that an appointment was arranged. Neither proves that care was delivered or that the clinician reviewed the result.
If the patient declines, record that outcome separately. If the clinician decides that the referral is no longer appropriate, record the decision and reason. If the team cannot confirm completion, retain that uncertainty instead of converting it into a completed-care count. This example is illustrative and does not describe a real patient.
Use AI for coordination with visible clinical control
In the workflow we recommend testing, AI organizes reported information, prepares work queues, records outreach attempts and flags unresolved items. The care team validates the information and remains responsible for clinical decisions.
An operator evaluating a system should ask it to demonstrate an ordinary unresolved case: a referral without a returned report, an unanswered outreach attempt or a changed patient preference. Can staff identify the owner, see the source information, correct the record and route the issue to the appropriate clinician?
An appointment reminder is useful only if the team can also see what happens when the patient cannot attend or needs a different way to respond. Telephone and staff-assisted options should remain available alongside digital channels.
Our clinic operating system overview describes the broader coordination problem. The same operational question applies here: can the team see what needs to happen next and who owns it?
Measure delivered care separately from closed tasks
We recommend defining a cohort and a follow-up window before comparing results. For example, a practice might review the actions approved after a defined set of AWVs and report how many reached each outcome by the review date. The appropriate window depends on the action and clinical context.
Keep these outcomes distinct:
- Care completed, with the source and date of confirmation.
- Patient declined, with the documented disposition.
- Clinician closed the action for a recorded reason.
- Still open, including scheduled, pending, unreachable or unconfirmed work.
Report the denominator and observation period alongside any percentage. Review overdue work, missing owners and time to clinical review as operational measures. Evaluate patient outcomes, unnecessary work, errors and differences in access separately. A rising task-closure rate alone cannot establish better care.
Start with one workflow the team can review
Choose a bounded referral or follow-up process, agree on the statuses and responsibilities, and review unresolved cases with the people who perform the work. Compare the recorded status with the supporting evidence. Document where information is missing and where staff have to work outside the system.
That gives a practice a concrete way to evaluate whether its process makes follow-through more reliable. The question is whether the agreed next step happened, with evidence the team can inspect.
For the policy context, read Caesar Health’s submitted CMS comment. To discuss your practice’s coordination workflow, see our pilot program.