The Authorization Queue: Catch the Lapse Before the Visit Happens

The denial code says the authorization lapsed, and it lands on the biller’s desk. But rewind the tape: the lapse did not happen in billing, and it did not happen this week. It happened three weeks ago, at the front desk, in the four seconds it took to book a thirteenth visit against a twelve-visit approval — or to schedule into the week after an authorization’s end date. Everyone did their job. The therapist treated, the front desk scheduled the standing slot, the biller submitted a clean claim. The system failed because no part of it was watching the two numbers that decide whether a visit is payable before it happens: visits remaining and days remaining. This article rebuilds prior authorization as what it operationally is — a scheduling constraint managed as a tracked queue — so the practice finds out about a problem while it can still be fixed with a fax instead of a write-off.

Follow the timeline of a typical lapse and the misdiagnosis becomes obvious. Day one: a payer approves twelve visits over ninety days. Week six: visit twelve happens, unremarked, because nobody’s workflow surfaces a running count at booking or check-in. Weeks seven and eight: visits thirteen and fourteen happen — delivered in good faith, documented, and unbillable. Week ten: the remittance arrives and billing discovers the problem, files for a retroactive authorization if the payer allows one, and starts an appeal if it does not. The practice experiences this as a billing problem because billing is where it surfaces. But billing is the smoke detector, not the fire. The fire is a schedule that will happily book a patient past either of their authorization’s two ceilings, because the person booking cannot see them.

The misdiagnosis matters because it decides where practices invest. Treat lapses as a billing problem and the fixes are downstream: faster appeals, retro-auth request templates, a spreadsheet the biller reconciles after claims bounce. All of that is cleanup. Treat lapses as a scheduling problem and the fix moves upstream to the only place it prevents anything: the moment of booking, and the daily check of what is already booked. A practice that cannot answer “how many approved visits does this patient have left, and when does the window close?” at the front desk — in the scheduler, not in a binder — does not have an authorization process. It has an authorization archive.

Prior authorization is a payer-by-payer fact, not a category-wide one, and the pattern surprises people who are new to running a caseload. Original Medicare Part B generally does not require prior authorization for outpatient therapy services: it manages utilization after the fact, through the annual threshold amounts — above which the KX modifier attests continued medical necessity — and a targeted medical review process at a higher threshold. Medicare Advantage plans, by contrast, are run by private insurers and frequently do require prior authorization for the very same services, a distinction that catches practices when a patient switches from one to the other. In recent years large national carriers have expanded authorization requirements for outpatient therapy under their Medicare Advantage products, a shift ASHA and APTA have both tracked and pushed back on. Commercial plans and Medicaid managed-care organizations sit across the whole spectrum: some authorize episodes generously after an evaluation, some cap tightly and demand renewal paperwork every handful of visits, and a few require nothing at all.

The operational consequence: whether authorization is required is itself a data point your intake and verification step must capture per patient, per plan — not something anyone should assume from the payer’s name. A practice’s authorization queue therefore starts at eligibility verification, where “does this plan require prior auth for these codes, and what are its renewal rules?” gets asked and the answer gets written down where scheduling can see it.

A queue is not a spreadsheet that lists authorizations. It is a discipline: every authorization is always in exactly one state, every state has an owner and an exit condition, and the queue gets reviewed on a schedule rather than when something breaks. Six states cover the whole life of an authorization. Run them in whatever tool you have — a capable EHR tracks most of this natively, and a shared sheet can carry a small practice — as long as the states, owners, and triggers are explicit.

1. 01

   Flag at verification

   During eligibility verification, before the first visit, determine whether this plan requires prior authorization for the services you expect to deliver — evaluation and treatment codes separately, since plans often treat them differently. Record the answer either way: “no auth required” written down is what keeps someone from re-litigating the question at claim time. If authorization is required, the patient enters the queue in this state until the request goes out.

2. 02

   Submit and log

   Send the request the way this payer requires — portal, form, fax — and log four things the moment it leaves: the date, the channel, the payer’s reference or case number, and exactly what was requested (codes, visit count, date span). The reference number is what turns every future phone call from an argument into a lookup. An authorization request that is not logged is, operationally, a request that was never sent.

3. 03

   Chase while pending

   Every pending request gets a follow-up date the day it is submitted, calibrated to this payer’s decision clock — for Medicare Advantage, Medicaid, and CHIP plans, federal rule caps standard decisions at seven calendar days and expedited ones at seventy-two hours beginning in 2026, which tells you exactly when silence has become a problem worth a phone call. Chasing is a scheduled activity, not a reaction to a patient sitting in the waiting room without an approval.

4. 04

   Activate with both ceilings visible

   When the approval arrives, transcribe it into the scheduler — not just the chart. The approved visit count, the start date, and the end date must be visible, or better, enforced, at the point of booking. This is the step most practices skip: the approval gets filed as a document instead of installed as a constraint, and from that moment the front desk is booking blind.

5. 05

   Track the burn

   Each completed visit decrements the count; the queue always shows visits used, visits remaining, and days remaining for every active authorization. This is also where the renewal trigger lives: a per-patient threshold — calculated, not guessed, from visit frequency and the payer’s decision time — that flips the authorization into renewal the moment it is crossed. The arithmetic is in the next section.

6. 06

   Renew or close

   Crossing the trigger starts the renewal: updated progress documentation, the request, and a new cycle through pending and active. If the episode is ending instead, close the authorization with a disposition — completed, discharged, moved to a new plan year — so the queue only ever contains live constraints. A queue full of dead rows stops being read, and a queue that stops being read is a binder.

Every authorization ends in one of two ways, and your tracking has to watch both. The visit-count ceiling falls first for frequent attenders: a patient seen twice weekly consumes a twelve-visit approval in six weeks, however long the paper window runs. The date ceiling falls first for sporadic attenders: a patient who reschedules often can reach a ninety-day end date with visits still unused — and those visits do not roll over. The queue’s job is to project, for each patient, which ceiling arrives first at their actual attendance rate, and to start the renewal far enough ahead of that ceiling that the payer’s decision clock runs out before the visits do.

That lead time is arithmetic. Take the patient’s real visits per week — from attendance, not from the plan of care — multiply by the payer’s decision time in weeks, and add a buffer for gathering documentation and for the payer asking questions. The result is the renewal trigger, expressed in visits remaining: submit when the count falls to that number. A twice-weekly patient with a payer that decides in a week needs the renewal moving at roughly four to five visits remaining — two visits that will happen while the payer deliberates, plus buffer. The same patient with a payer that routinely takes three weeks needs the trigger at eight or nine. Set the trigger per authorization, write it into the tracker as a number, and the question “should we start the renewal?” stops requiring judgment at all.

The renewal trigger depends on knowing how long a payer takes to decide, and for a large class of plans that number is now bounded by federal rule rather than by habit. The CMS Interoperability and Prior Authorization final rule (CMS-0057-F, finalized January 2024) requires the payers it covers — Medicare Advantage plans, state Medicaid and CHIP programs, and Medicaid and CHIP managed-care plans — to answer standard prior authorization requests within seven calendar days and expedited requests within seventy-two hours, generally beginning January 1, 2026. The same rule requires those payers to give a specific reason when they deny a request, and to publicly report their authorization metrics — approval, denial, and appeal-overturn rates among them — with the first reporting due by March 31, 2026. Commercial plans outside the rule set their own clocks, subject to state law and accreditation standards, which is why the tracker records a decision time per payer instead of assuming one.

These ceilings are useful to a practice in two concrete ways. First, they calibrate the follow-up date in the pending state: for an in-scope plan, a standard request with no answer on day eight is not “still processing” — it is late, and the call you make can say so. Second, the published metrics give you, for the first time, payer-level denial and overturn rates to consult when deciding which panels are worth the administrative load their authorization behavior imposes.

A well-run queue does not prevent denials; it prevents surprises. When a denial does come, the default should be to look hard at appealing rather than writing off, because the evidence says payer prior-authorization denials are wrong often enough to matter. When the HHS Office of Inspector General reviewed a stratified random sample of prior authorization denials issued by fifteen of the largest Medicare Advantage organizations during one week of June 2019, it found that 13 percent of the denied requests met Medicare coverage rules — services that likely would have been approved under original Medicare. A common cause was plans applying clinical criteria beyond what Medicare coverage rules contain. That is one program, one sample week, and one line of evidence — but it is the federal watchdog’s own arithmetic, and it argues against treating any single denial as final.

The appeal itself is mostly assembled from things the queue already holds: the reference number and submission log from the pending state, the exact scope requested, the progress documentation that supported the renewal, and — under CMS-0057-F, for in-scope plans — the payer’s specific stated reason for the denial, which tells you precisely what the appeal must rebut. Practices that lose appeals usually lose them for missing paperwork and blown deadlines, not weak clinical cases. The queue is what makes the paperwork exist and the deadlines visible.

Queues fail socially before they fail technically: a tracker that belongs to everyone is updated by no one. Give the queue a single named owner — in a small practice usually the person who runs verification and scheduling, precisely because the queue is a scheduling instrument. The owner’s daily interaction is a glance, not a project: anything pending past its follow-up date, anything that crossed its renewal trigger, anything expiring inside the next two weeks. Ten minutes. The weekly review is the deeper pass, and it is short enough to survive contact with a real week:

“

“Billing discovers the lapse; scheduling causes it. Put the tracking where the cause lives, and the discovery never happens.”