Provider Departure Rescheduling in a Cardiology Practice

Provider departure rescheduling is the least automated expensive job in a cardiology practice. A cardiologist gives notice, works out the ninety days, and leaves behind a panel that is still on the schedule: follow-ups after a device check, patients waiting on a stress test, post-procedure visits that were booked eight months out. None of it moves on its own.

That panel is revenue the practice has already earned and cannot collect until someone rebooks it. Most groups work it the same way, by exporting the departing provider’s future appointments into a spreadsheet and handing it to whoever has the fewest calls that week. The list is long, every row has a different reason for existing, and the person working it has no way to know which patients can even be moved to which remaining cardiologist.

The departing panel is not one list

Physician turnover is not a rare event to plan around once. The annual physician turnover rate in the United States rose to 7.6% by 2018, up from 5.3% in 2010, according to national estimates built from Medicare claims. In a practice with eight cardiologists, that is a departure roughly every year and a half.

The first instinct is to sort the panel by provider and reassign in bulk. That breaks immediately, because the field most systems sort on is wrong. The chart’s primary provider value is stale almost everywhere and cannot be trusted for routing. A patient may be listed under a cardiologist they last saw in 2022 while their actual care has run through someone else in the group for two years.

The usable signal is who has actually seen the patient recently. Reading the encounter history rather than the chart header separates the panel into patients who genuinely belong to the departing provider and patients who were only nominally assigned to them. In most departures that second group is large, and every one of those rows is a call nobody needed to make.

Payer enrollment decides who the patient can move to

Once the real panel is identified, the constraint that matters is not who has an open slot. It is who the patient’s plan will pay for.

Credentialing resolves per provider, per payer, per state, and per treatment type. All four have to be true before a slot can be offered. Cardiology feels this harder than most specialties because the payer mix is wide and the procedure list is long, so a patient who was fine with the departing cardiologist may not be covered with the colleague who inherits them.

The provider-to-payer enrollment matrix almost never lives in the EHR. It sits in a spreadsheet in the credentialing coordinator’s folder, which means nothing in scheduling knows about it. Book a patient with a cardiologist who is not enrolled in their plan and you get a visit that happens and a claim that dies weeks later.

Practices usually cannot hand over a clean enrollment grid, because the full version is too messy to export. The workable artifact is the inverse: a per-provider list of the plans that provider does not take. That is a short document a credentialing coordinator can actually produce, and it is enough to keep the automation from offering a slot that will not pay. Feeding that list into the rescheduling pass, alongside a real-time eligibility check on the patient’s current coverage, is what turns a panel transfer into bookings that survive to payment.

Rescheduling silently breaks the authorization link

The rows on a departing cardiologist’s panel that cost the most are the ones attached to an approved authorization. Stress tests, catheterizations, device checks, and imaging all commonly carry one.

Rescheduling breaks that link. The new appointment is not attached to the existing authorization, and the practice does not find out at booking. It finds out at checkout, or later, when the claim is denied for a visit that was authorized under a date and a provider that no longer apply.

The rule that protects the claim is directional and worth stating plainly to whoever is working the list: an appointment whose authorization has not come back can be pushed later, never pulled earlier. A panel transfer that moves patients up into newly opened slots, which is exactly what a well-meaning scheduler does when the departing provider’s calendar frees up capacity, walks straight into this.

Automating the pass means checking the authorization window on every row before offering a date, flagging the ones where the approved provider is the person leaving, and routing those to the auth team rather than booking them. That is the difference between a transferred panel and a transferred panel that bills.

The receiving providers have caps, and the caps expire

The remaining cardiologists cannot simply absorb the panel. Availability is carved, and the carving has rules that no static scheduling template expresses well.

A provider may cap new patients at four a day, or one, while a large follow-up queue waits behind them. The schedule is built so that a thirty minute follow-up hole cannot absorb a patient who needs a longer new-visit slot, even when the calendar looks open. If the practice hired a replacement, that new cardiologist ramps with caps that expire on a date, for example a maximum of four new intakes a day for the first two weeks and unlimited after. The cap has to stop applying on its own, because nobody remembers to remove it.

There is a second trap in the slot mechanics. Generic template slots, the athenaOne “Any 15” and “Any 30” types, are returned when you search for a specific appointment type, but a generic fifteen minute slot may need to be an hour for a patient establishing with a new cardiologist. Which specific types a generic slot is genuinely eligible for, per provider and per department, is the whole mapping problem, and getting it wrong produces a booked schedule that the provider cannot actually work.

This is where the work is a poor fit for a person with a spreadsheet and a good fit for automation. The rules are knowable, they are just numerous, and they have to hold on every one of several hundred rows.

What the AI does, and the point where it stops

In practice the pass runs in a defined order. Read the departing provider’s future appointments and open orders in athenaOne. Separate the real panel from the nominal one using encounter history. Check each patient’s current eligibility, screen the receiving provider against the do-not-take plan list, confirm the appointment type maps to a slot the receiving provider can genuinely work, and check whether an authorization is attached. Then call the patient, offer dates that satisfy all of it, book, and write the outcome back to the tickler queue so the row is closed rather than merely touched.

Outbound calls carry the same rules as inbound ones, which matters here because a panel transfer is a campaign, not a queue. Patients who do not answer get worked again on a schedule instead of falling to the bottom of a spreadsheet.

The handoff is specific and it is not a fallback for when the automation fails. Three categories go to a human every time: any row where the authorization is attached to the departing provider, any patient whose coverage no longer matches an available cardiologist, and any patient who asks a question about their care rather than their appointment. The last one is a hard line. Patients call about a departing physician with real clinical anxiety, and the correct response to “should I be worried about waiting another month” is a warm transfer to clinical staff with the chart context attached, not an answer.

What the practice gets is not a replaced scheduler. It is the volume worked in days instead of quarters, with the exception rows arriving on a person’s desk already sorted and already carrying the reason they are exceptions.

Why the list ages badly if you leave it

There is a reason to work a departing panel quickly that has nothing to do with tidiness. Research on primary care physician exits found that in the year after their physician left, patients had 18.4% fewer primary care visits and more specialty, urgent, and emergency visits than patients whose physician stayed. The pattern in that study is what a panel does when nobody rebooks it: patients do not wait patiently, they go somewhere else or they go nowhere.

The cardiology version has a sharper edge because a fair share of the panel is on a follow-up interval that exists for a reason. A patient who drops off a recall list after a device check is not a lapsed marketing contact. They are a scheduled follow-up that stopped happening, and getting them rebooked is both the clinically appropriate outcome and the revenue outcome at the same time.

Time is the variable the practice controls. The panel is the same size whether it is worked in two weeks or five months, but its yield is not.

Key Takeaways

Build the panel from encounter history, not the chart’s primary provider field. That field is stale almost everywhere, and sorting on it puts patients on the call list who never needed to be there.
Get the inverse credentialing list. A per-provider list of plans that provider does not take is short enough for a credentialing coordinator to actually produce, and it stops the automation offering slots that will never pay.
Check the authorization on every row before offering a date. A rescheduled appointment does not carry its old approval, and an appointment whose auth has not returned can be pushed later but never pulled earlier.
Map generic slots to real appointment types before the pass runs. An “Any 15” that gets booked as a new-patient establish visit produces a schedule the receiving cardiologist cannot work.
Make ramp caps self-expiring. A new provider’s four-intakes-a-day limit for the first two weeks has to lift on its own, because nobody remembers to go turn it off.
Define the three handoff categories before go-live: auth attached to the departing provider, coverage with no matching cardiologist, and any question about care rather than scheduling. The last one goes to clinical staff every time.

A provider departure is one of the few events where a practice knows months in advance that several hundred appointments will need to move, and still ends up working the list by hand in the weeks after the provider is gone. The rules that govern who can be booked with whom are all knowable. They are just too many to hold in a scheduler’s head across a few hundred calls, which is exactly the shape of work worth handing to an AI team that operates inside athenaOne rather than beside it.

Sources