Screening Recall That Actually Reaches the Patient

Screening recall fails quietly. The order exists, the patient is due, everyone agrees it should happen, and eleven months later nothing has. Preventative medicine practices carry more of this than anyone, because almost all of their volume starts as something a clinician wrote down rather than something a patient called about.

That is the structural difference and it changes the whole front office. In most specialties the patient initiates. Here the practice does, which means the work is outbound, list-driven, and easy to postpone.

It is also frozen revenue. A recall list is a set of visits that have already been earned clinically and simply have not been booked. Nobody disputes that they should happen. There are just no hours in the week that belong to making them happen.

So the list ages. Staff work the top of it when the phones are quiet, which is never, and the practice ends up measuring outreach by whether anyone got to it rather than by whether patients came in.

And when someone does get to it, the list turns out not to be a list at all.

The recall list is two queues that disagree

What looks like one worklist is usually an open order sitting on the chart and a follow-up task sitting in a bucket, created at different moments by different people for the same patient.

In athenaOne both are readable. Outstanding orders can be pulled against the encounter, the order document itself lives on the patient chart, and follow-up tasks live in their own queues with their own routing. The trouble is that they do not know about each other.

An engineer working a live outreach rollout put it as a direct question: if a patient has both an open follow-up task and an open order at the same time, what should the workflow do? Working the task does not necessarily close the order, so the order stays alive and somebody calls the patient a second time. Compounding it, staff move items between queues by hand, so an item that should be in one bucket ends up in the one the automation reads from, or does not.

The fix is dull. Decide, per practice, what satisfies what: which queue is authoritative, whether booking the appointment closes the order, and what happens to the sibling item when it does. Write it down before any outreach runs. A recall program without that rule generates duplicate calls at exactly the volume it was built to handle.

The clinician decides what is due, the front office decides how it gets booked

Keeping that line sharp is what makes an automated recall program safe to run. It also makes it easier to build.

Screening intervals are set by guideline and by the clinician. For colorectal cancer, the U.S. Preventive Services Task Force recommends that adults age 45 to 75 be screened, with the decision between 76 and 85 made individually. None of that is a front-office judgment call, and none of it needs to be. The order already carries the answer.

What the front office owns is everything after the order exists: reaching the person, explaining what the appointment is, handling the objection about timing or cost, offering a slot, booking it, and confirming it. That is logistics, start to finish, and it is the part that fails.

The practical benefit of drawing the line clearly is that the automation never has to rank patients by anything other than how long the order has been open and what the practice’s own outreach policy says. It works the queue in the order the practice defined. Anything that reads as a question about whether the screening is right for this patient goes to clinical staff, unanswered.

Reaching the patient is a channel problem

The message is rarely why recall fails. The channel is.

One call during business hours to a number that goes to voicemail is the default recall program at most practices, and it is close to the weakest possible design. The patients who most need the reminder are the ones least able to answer a daytime call, and a voicemail asking them to call back moves the work to the person least likely to do it.

Digital self-service has not filled the gap either. A July 2025 MGMA poll found that 71% of medical groups have fewer than one in four patients using digital tools to schedule appointments. Sending the whole recall list to a portal link is a plan that reaches a quarter of them at best.

What works is running the channels together and letting the patient choose. Voice for the ones who answer the phone, secure message for the ones who live in the portal, text for the ones who will reply to a text and never call back, and the same content in the patient’s language rather than in English with an apology. The point is not the number of channels. The point is that a single attempt on a single channel is a coverage decision disguised as an outreach program.

If it does not end in a booked appointment, nothing closed

A contacted patient is not an outcome. It is a step that feels like one, which is why so many recall programs quote contact rates.

The outreach has to carry the booking. Open slots in athenaOne can be read live, so the person on the phone or in the message thread can be offered real times against the right appointment type and department rather than told that someone will call them back to schedule. The callback is where recall dies. Every handback adds a day and a chance to forget.

Closing the loop also has a documented failure mode on the other side. Reviews of outpatient practice have found that patients are not always informed of significant test results after the fact, which means the loop can break after the visit as easily as before it. A recall program that books the visit and then goes quiet has solved half a problem.

So the definition of done is narrow and worth defending: an appointment exists, the patient knows when it is, the order or task is satisfied according to the practice’s own rule, and the sibling item was handled so nobody calls again next Tuesday.

Where the person takes over

The automation reads the queue, dials or messages, explains what the appointment is, answers the ordinary questions about timing, location and cost, offers real slots, books, confirms and closes the item.

It stops the moment the conversation becomes clinical. A patient asking whether they still need the screening, describing a new symptom, or pushing back for a reason that is not about the calendar goes to clinical staff with the context attached. So does anything the automation cannot map to an order it can read. Care coordination in the sense the research literature uses it, deliberately organizing care between participants, remains a human function. What the front office contributes to it is that the appointments actually get made.

This is the split that practices describe when they say they want an extra team rather than another product. The clinical side keeps every decision. The AI absorbs the queue nobody has hours for, which is the reason the list was aging in the first place.

One number tells you whether any of this is working, and it is not contact rate. It is the share of open orders that turned into a kept appointment, measured monthly, with the age of the queue next to it.

Key Takeaways

Treat the recall list as two queues, open orders and follow-up tasks, and write down which one is authoritative before running any outreach.
Define what satisfies what: whether booking closes the order, and what happens to the sibling item, or expect duplicate calls at scale.
Keep interval and eligibility decisions with the clinician who wrote the order, and let the automation work the queue in the practice’s stated order.
Stop designing recall around one daytime phone call, since that reaches the patients who least need help and misses the ones who do.
Assume digital self-scheduling covers a minority of your panel, given that 71% of groups have under one in four patients using it.
Offer real open slots inside the same contact instead of promising a callback, because the callback is where recall dies.
Run outreach in the patient’s language on the channel they answer, and treat single-channel outreach as a coverage decision.
Measure the share of open orders that became a kept appointment, alongside the age of the queue, rather than contact rate.

Screening recall is not a communication problem, it is a queue problem with a communication layer on top. Reconcile the two queues, keep the clinical call where it belongs, reach people on the channel they actually answer, and refuse to count anything as closed until an appointment exists. The list stops aging the week somebody owns it full time, and nobody has to hire for that.

Sources