A therapy visit is minutes before it is money.
Every dollar an outpatient therapy practice earns starts as a number written on a flowsheet. Between that number and the remittance sit a unit ladder, a modifier set, a dollar threshold, a signature deadline, and a visit counter. Miss any one of them and the treatment still happened, the note still exists, and the money still does not arrive.
Module 01 · the unitThe billable unit is a pool, not a per-code rounding.
Outpatient therapy is one of the few places in healthcare where the revenue-bearing object is measured in minutes by a clinician mid-treatment and then converted to money by an arithmetic rule that most billing systems implement incorrectly. Everything else in this manual sits downstream of that conversion. Get it wrong and no amount of denial work recovers what was never billed in the first place.
Medicare pays timed therapy procedures in fifteen minute units. The conversion from minutes to units is set out in the CMS Medicare Claims Processing Manual, chapter 5, section 20.2, and it is universally known as the 8-minute rule. The critical word in the rule is total. The count is the sum of all timed minutes furnished to one patient, by one discipline, on one date of service. It is not each CPT code rounded on its own and then added up. That single distinction is where most missing units go.
The ladder itself is unambiguous. Each successive unit opens at the previous unit's fifteen minute block plus eight minutes, which is why the entry points are 8, 23, 38, 53, 68 and 83 rather than the round numbers a scheduler would expect. Below eight total timed minutes there is no billable timed unit at all.
| Total timed minutes | Units | Entry point arithmetic | What it means on the floor |
|---|---|---|---|
| 0 to 7 | 0 | below the first entry | No timed unit is billable. Untimed codes are unaffected. |
| 8 to 22 | 1 | (15 x 0) + 8 | The first unit opens at 8 minutes, not at 15. |
| 23 to 37 | 2 | (15 x 1) + 8 | A 22 minute visit and a 23 minute visit differ by a full unit. |
| 38 to 52 | 3 | (15 x 2) + 8 | The most commonly under-billed band in a mixed-code visit. |
| 53 to 67 | 4 | (15 x 3) + 8 | A standard hour-long treatment lands here. |
| 68 to 82 | 5 | (15 x 4) + 8 | Extended sessions. Expect payer scrutiny above four units. |
| 83 to 97 | 6 | (15 x 5) + 8 | The pattern continues in fifteen minute steps from here. |
Two boundary conditions deserve their own note. First, the pool is per discipline. If a patient receives physical therapy and occupational therapy on the same day, those are two separate pools, counted separately, laddered separately, and billed with different discipline modifiers. Combining them is not generosity, it is a billing error. Second, the pool is per date of service. Two visits in one day for the same discipline combine into one pool, which is precisely how a practice with a morning and afternoon appointment accidentally bills eight units where the ladder supports seven.
There is also a standard that is not Medicare's. CPT applies a midpoint convention, under which a fifteen minute unit is reported once the service passes the midpoint of the time increment. APTA notes that this differs from Medicare's 8-minute rule, and that individual payer policies vary. In practice that means a therapy practice needs its unit engine to be payer-aware rather than hard-coded to one rule. A commercial plan on the CPT midpoint standard and Medicare on the total-timed-minutes ladder will not always produce the same unit count for the same visit.
Module 02 · allocationKnowing the unit count is half. Assigning it is the other half.
The ladder tells you how many units the visit supports. It does not tell you which CPT line each unit belongs on, and that is the question the claim actually asks. Allocation is a mechanical procedure: full fifteen minute blocks first, then leftover units to the codes holding the largest remaining minutes, with a documented convention for ties. Every step is auditable, which is exactly why it should never be improvised.
Three visits, worked end to end
The arithmetic below is deliberately dull. That is the point. Every one of these visits is ordinary, and every one of them is billed wrongly by a system that rounds per code.
Example C is the one that produces overpayment rather than underpayment, and it is therefore the one that survives internal review the longest. Nobody audits a visit that paid. It surfaces later, in a probe review, as a pattern of unit counts that the treatment notes cannot support. That is the worst possible way to find it.
Enter the timed minutes per code for a visit and the calculator returns the total pool, the units the 8-minute rule supports, and the line by line allocation including the largest-remainder assignment. Useful for settling a charge-entry dispute in about ten seconds.
Module 03 · the code mapTimed and untimed codes obey different physics.
Before any minute can be laddered, every code on the visit has to be sorted into one of two buckets. The test is the CPT descriptor itself. A descriptor that carries a time designation, most commonly the phrase each 15 minutes, is a timed code and produces units from minutes. A descriptor with no time designation is a service-based or untimed code, billed once per day regardless of how long it took.
Speech-language pathology sits almost entirely on the untimed side, which is why SLP revenue behaves differently from PT and OT inside the same practice. Most SLP treatment and evaluation codes carry no time designation in the descriptor and are billed once per day irrespective of session length. An SLP session that runs fifty minutes and one that runs twenty five produce the same claim line. That does not make documentation less important, it makes it more important, because duration is no longer the variable the payer can see and skilled content is the only thing left to review.
Two operational consequences follow. First, a practice that measures productivity in units will systematically understate SLP output and overstate PT output, which distorts staffing decisions in a multidisciplinary clinic. Measure SLP in visits and PT and OT in units, or measure everything in net revenue per treating hour, but do not mix the two on one dashboard. Second, group therapy under 97150 is untimed and is not the same object as individual treatment, so it does not enter the individual pool at all. A therapist who runs a thirty minute group and a twenty minute individual session has one untimed group line and a twenty minute individual pool, not a fifty minute pool.
Module 04 · CQ and COThe assistant modifiers are a staffing decision priced into every line.
CQ marks an outpatient therapy service furnished in whole or in part by a physical therapist assistant. CO marks the same for an occupational therapy assistant. Effective January 1, 2022, a line carrying either modifier pays at 85 percent of the otherwise applicable Physician Fee Schedule amount. That fifteen percent differential is not a penalty for poor billing, it is a payment policy, and the only way to manage it is to know in advance which units will carry it.
Example D: a split treatment, priced
A patient receives 40 minutes of 97110, of which the PTA furnished 30 minutes and the physical therapist independently furnished 10 minutes, plus 13 minutes of 97140 furnished entirely by the therapist. The total timed pool is 53 minutes, which the ladder converts to 4 units. Allocation: 97110 takes 2 full blocks and holds 10 leftover minutes, 97140 takes 0 full blocks and holds 13 leftover minutes. Two units remain to assign. The largest remainder is 97140 at 13 minutes, so it takes one, and 97110 at 10 minutes takes the other. The visit bills 97110 x 3 and 97140 x 1.
Now the modifier. The 97140 unit is entirely the therapist's, so it carries no CQ. Of the three 97110 units, the therapist independently furnished 10 minutes, which satisfies the carve-out of at least 8 minutes of a fifteen minute unit, so one 97110 unit bills without CQ. The remaining two 97110 units carry CQ and pay at 85 percent. If those two units happened to carry the same allowed amount as the two unmodified units, the visit would pay 92.5 percent of what an all-therapist visit would have paid, since two full units plus two units at 0.85 gives 3.7 of 4.
One nuance frequently gets lost. Minutes an assistant furnishes concurrently with the therapist count as the therapist's minutes, not the assistant's. That matters enormously in a clinic that routinely staffs a therapist and an assistant on the same patient. The scheduling model that looks most expensive on paper may be the one that produces the fewest modified units. The point is not to avoid assistants, who deliver a large share of appropriate and skilled care. The point is that the differential should be a decision, budgeted and forecast, rather than a surprise discovered in a variance report.
Module 05 · the second codeThe multiple procedure reduction quietly reprices every extra unit.
A second differential sits underneath the assistant differential and is far less visible because it never generates a denial, a modifier, or a message on the remittance. The multiple procedure payment reduction applies a 50 percent reduction to the practice expense value of the second and each subsequent always-therapy procedure furnished to the same patient on the same day. Work and malpractice values are untouched. Congress made the 50 percent reduction permanent in the American Taxpayer Relief Act of 2012, and CMS implemented it on April 1, 2013.
Because it lands on the practice expense component rather than the whole payment, MPPR does not reduce a four unit visit by half. It reduces one of the three inputs to the fee schedule amount, and the size of that reduction varies code by code according to how practice-expense-heavy the code is. That variability is why MPPR is genuinely hard to intuit and why a therapy practice should model it against its own top ten code combinations rather than trying to reason about it in the abstract.
The practical implication is straightforward and slightly uncomfortable. The marginal revenue of the fourth unit in a visit is materially lower than the marginal revenue of the first, and the marginal revenue of a second visit on a different day is higher than a fourth unit on the same day. None of that should drive a clinical decision. All of it should inform how a practice forecasts revenue from a schedule and how it interprets a drop in revenue per visit that has nothing to do with billing quality.
Module 06 · the thresholdThe KX threshold is a checkpoint, not a ceiling.
For calendar year 2026 the threshold is $2,480 for physical therapy and speech-language pathology services combined, and a separate $2,480 for occupational therapy services. Crossing it does not stop payment and does not require an appeal. It requires the KX modifier, which attests that the services remain medically necessary and that the medical record supports that judgment. A separate targeted medical review threshold sits at $3,000, applies for 2018 through 2028, and makes claims above it eligible for review by the Supplemental Medical Review Contractor.
The operational question is when to look. The answer is not once a month. A therapy episode can move a patient several hundred dollars in a week, so the accrual has to be evaluated at scheduling time, before the visit rather than after the claim. Practices that discover a crossing on the remittance are, by definition, discovering it one to three visits too late.
To make the timing concrete, the table below converts the two thresholds into a number of visits at several illustrative average allowed amounts. These are illustrations, not benchmarks. Substitute your own realized allowed amount per visit by discipline and the arithmetic will be exact for your book.
| Average allowed per visit | Visits to $2,480 | Visits to $3,000 | Visits between | 80% alert visit |
|---|---|---|---|---|
| $80 | 31.0 | 37.5 | 6.5 | 24.8 |
| $100 | 24.8 | 30.0 | 5.2 | 19.8 |
| $120 | 20.7 | 25.0 | 4.3 | 16.5 |
| $140 | 17.7 | 21.4 | 3.7 | 14.2 |
Read the fourth column carefully. At $140 per visit there are fewer than four visits between the KX line and the medical review line. A practice running an intensive episode can cross both inside a single week. That is the case for putting the threshold check in the scheduling workflow rather than the billing workflow, and it is one of the controls that drove the results in our therapy authorization turnaround case study, where documented ABN-on-file discipline moved from 46 percent to 98 percent.
Module 07 · certificationA plan of care has two dates, and both are unforgiving.
A physician or non-physician practitioner must certify the outpatient therapy plan of care within 30 days of the first day of treatment, and recertify at least every 90 days. Both intervals come from the CMS Medicare Benefit Policy Manual, chapter 15, section 220.1.3. Neither is negotiable and neither is enforced by the EMR unless somebody configures it to be. Visits delivered after a lapse are unbillable until the signature exists, and the signature is held by someone outside the practice.
Certification lapse is the purest form of therapy revenue leakage because absolutely nothing about it is clinical. The patient attended, the therapist treated, the note is complete, and the claim is unbillable because a document is missing a signature. It is also the most recoverable defect in the entire manual, since a lapse identified quickly is usually curable by obtaining the signature, whereas a lapse identified at the timely filing deadline is not curable at all.
Two design choices make the difference. First, the tracker keys off the first treatment date and the certification date, not off the referral date, because a referral can predate the first visit by weeks and produce a window that closes earlier than anyone expects. Second, ownership is named per patient rather than per clinic. A shared queue where everyone can chase means nobody does, and the 71 percent on-time recertification rate we found at the multi-site group in our therapy authorization case study was the direct output of a shared queue with no owner.
Module 08 · the countersVisit limits run on a counter, not a calendar.
Commercial and Medicare Advantage plans manage therapy volume with visit caps, authorized visit counts, and authorized unit counts. All three are consumed by delivery, which means the meaningful clock is a counter that ticks down every time a patient walks in, not a date on the authorization letter. A practice that tracks the expiry date and not the counter will reliably run out of visits in the middle of an episode with a valid authorization still on file.
The middle card is the instructive one. A units-based authorization consumed at four units per visit exhausts in roughly six visits, which at three visits a week is two weeks of runway. If that authorization also carries a sixty day expiry date, a date-driven tracker will show it as comfortably current right up to the visit that denies. This is the same units-versus-calendar problem we describe for ABA in the prior authorization command center, and the fix is identical: the trigger is a percentage of units consumed, with the expiry date serving only as a backstop.
Set the trigger at 80 percent of authorized units or visits consumed. Not at a fixed number of days before expiry, and not at the last visit. Eighty percent gives a reauthorization packet enough runway to clear a payer turnaround that has its own variance, and it produces a queue small enough that a named owner can actually work it. A practice with two thousand active authorizations and a 90 percent trigger has built itself a permanent emergency.
Two adjacent failures round out the module. Benefit maximums that reset on the plan year rather than the calendar year will silently break any tracker that assumes January. And an authorization is issued to a rendering credential as well as a patient, so a visit covered by an authorization naming one therapist and delivered by another can deny on a technicality that no clinical review will ever catch.
Module 09 · the recordDocumentation that survives review is comparative, not descriptive.
The single most common finding in a therapy medical review is not a missing note. It is a complete, well written, thoroughly detailed note that describes what happened and never demonstrates that the patient changed. A reviewer applying medical necessity is looking for skilled care producing measurable movement toward the goals in the certified plan. A note that could be copied to next week without editing has failed that test regardless of how long it is.
The same instrument, scored the same way, at defined intervals. A gait speed or a range of motion measurement is only evidence when there is a prior value to compare it against.
Goals stated as a target value with a target date. Improve strength is not a goal. Ascend twelve stairs with one rail by week six is a goal a reviewer can score.
Recorded at the time of treatment, per code, with total timed minutes and total treatment time stated separately. Reconstructed minutes are the fastest route to a unit denial.
Why this intervention required the judgment of a licensed therapist. Performed exercises describes a gym. Adjusted resistance in response to compensatory pattern describes skilled care.
Written by the therapist, tied to the certification cycle, and stating whether the plan continues, changes, or ends. An assistant's daily note does not substitute for it.
The signed plan is a billing document, not a clinical courtesy. Store it where the biller can see it, and treat a missing signature date the same as a missing signature.
Once KX is on the claim, the record must independently support continued necessity. The modifier is an attestation, and the attestation has to be true on the page.
Therapist and assistant minutes recorded separately per code. Without that split the CQ and CO decision cannot be made correctly, and cannot be defended afterward.
Item eight is the newest and the least well implemented. Most therapy EMRs were designed before the assistant differential existed, and many still record a single treating clinician per visit rather than minutes per clinician per code. That design makes the modifier decision a reconstruction rather than a record, and a reconstruction is exactly what a reviewer will disallow. Where the EMR cannot be changed, the practice needs a structured supplementary capture, ideally at the point of care rather than at charge entry.
The habit that separates a defensible book from a fragile one is simple to state and hard to sustain: every progress note answers the question of what changed since the last one. Not what was done. What changed. That is the sentence a reviewer is looking for, it is the sentence a KX attestation depends on, and it is the sentence that most notes never contain.
Module 10 · the taxonomyTherapy denials are ten problems wearing many code numbers.
Generic denial management assigns therapy denials to buckets built for surgical and hospital claims, where they are meaningless. Coding error is not a root cause in a therapy practice. A missing discipline modifier, a unit count that exceeds documented minutes, and an unbundled pair without the required modifier are three different failures with three different owners and three different fixes, and they all arrive as coding error. The taxonomy below sorts them by what actually broke.
| CARC | What the payer says | Therapy root cause | The control that prevents it |
|---|---|---|---|
| CO-4 | Procedure inconsistent with the modifier used | Missing GP, GO or GN discipline modifier, or a missing CQ or CO assistant modifier | Modifier set derived from the plan of care discipline and the minutes-per-clinician split, applied at claim build |
| CO-151 | Information does not support this many services | Unit count exceeds the timed minutes the note documents, usually from per-code rounding or a charge-entry override | Ladder validation that blocks release when billed units differ from the pool the minutes support |
| CO-50 | Not deemed a medical necessity | Documentation shows repetition rather than measurable progress against the certified goals | Progress-note template that requires a compared objective measure before the note can be signed |
| CO-197 | Precertification or authorization absent | Visit delivered after the authorized visit or unit count was exhausted, with a valid expiry date still on file | Counter-based trigger at 80 percent of authorized units, with the expiry date as backstop only |
| CO-119 | Benefit maximum reached | Commercial visit cap exhausted, frequently on a plan year that does not start in January | Benefit maximum tracked per plan year at eligibility, re-verified at the plan year boundary |
| CO-96 | Non-covered charges | Services above the CY 2026 threshold billed without KX, or a service outside the certified plan of care | Threshold accrual evaluated at scheduling, with KX or a signed ABN forced before the visit |
| CO-16 | Claim lacks information | Missing certifying practitioner NPI, missing therapy minutes on an institutional claim, or an absent onset date | Pre-submission completeness edit driven by a therapy-specific field list, not the generic scrubber |
| CO-B7 | Provider not eligible for this service on this date | Rendering credential does not match the authorization, or an assistant billed under a supervision arrangement the plan does not recognize | Authorization holds the rendering credential, and scheduling cannot assign a therapist outside it |
| CO-18 | Exact duplicate claim or service | Same timed code resubmitted after a corrected-claim cycle rather than replaced | Corrected claims submitted as replacements with the original claim number, never as new claims |
| CO-29 | Time limit for filing has expired | Claim held in a work queue waiting for a certification signature that never arrived | Hard rule: a claim never waits on a signature past the filing deadline, it goes out and gets corrected |
The last row is the one that costs the most and gets the least attention. A claim held for a signature is invisible, generates no denial, appears on no dashboard, and ages quietly past the filing deadline in a queue somebody set up years ago with good intentions. Audit the held queue by age monthly. Anything older than half the shortest filing limit in your payer mix is already a write-off waiting to be discovered.
Notice how few of these rows are fixed by an appeal. Seven of the ten are prevented at or before the visit, which is why a therapy denial program that consists mainly of appeal writing produces a lot of activity and very little improvement. The measurable version of this argument is in the therapy authorization turnaround, where the authorization-related denial rate moved from 3.2 percent of visits to 0.4 percent almost entirely through pre-visit controls.
Module 11 · the scorecardSeven numbers tell you whether any of this is working.
A therapy revenue dashboard that reports charges, collections and days in accounts receivable is measuring the wrong layer. Those are outcomes of the mechanics in this manual and they move too slowly to steer by. The seven measures below sit close enough to the defect to be actionable inside a week, and each one has a formula that a supervisor can compute by hand to check the report.
The seventh measure has no target because it is a control rather than a result: the age of the oldest claim sitting in a held queue. Report it every week as a single number. It is the only measure on this page that goes wrong silently, and it is the one that turns a recoverable defect into a permanent one.
Read the scorecard as a system rather than a set of dials. Unit accuracy without certification discipline produces perfectly computed units on unbillable visits. Certification discipline without threshold decisions produces billable visits that get written off above the KX line. Threshold discipline without authorization counters produces attested claims for visits nobody authorized. Each control protects the one after it, which is why partial implementations tend to show almost no financial improvement and get abandoned as ineffective.
In therapy, the money is decided in the treatment room and merely confirmed in the billing office. Every control worth building sits before the claim, not after it.
Aparna Suresh, CPB · President and Founder, ASP-RCM Solutions
Questions we getThe six that come up every time.
How does the 8-minute rule convert therapy minutes into billable units?
Medicare counts total timed minutes for the discipline for the day, then reads a ladder: 8 to 22 minutes is 1 unit, 23 to 37 is 2, 38 to 52 is 3, 53 to 67 is 4, 68 to 82 is 5, and 83 to 97 is 6. The count is the pool of all timed codes added together, not each code rounded on its own. Untimed service-based codes such as an evaluation never enter the pool. The rule is set out in the CMS Medicare Claims Processing Manual, chapter 5, section 20.2.
What happens when no single timed code reaches 8 minutes?
The minutes still combine. Under chapter 5, section 20.2, when two or more 15-minute timed services are each furnished for 7 minutes or less on the same day and their total is 8 minutes or greater, one unit is billed for the service performed for the most minutes. A visit of 7 minutes of manual therapy, 6 minutes of therapeutic activities, and 5 minutes of neuromuscular re-education is 18 total timed minutes, which is 1 unit, billed on the manual therapy line. Rounding each code separately would bill nothing at all.
When are the CQ and CO assistant modifiers required, and what do they cost?
CQ marks a service furnished in whole or in part by a physical therapist assistant and CO marks one furnished by an occupational therapy assistant. Effective January 1, 2022, lines carrying either modifier pay at 85 percent of the otherwise applicable Physician Fee Schedule amount. The de minimis standard is 10 percent of the total service, and it does not apply when the therapist independently furnishes at least 8 minutes of a 15-minute unit, so that unit is billed without the modifier.
What is the CY 2026 KX modifier threshold and what does crossing it do?
For calendar year 2026 the threshold is $2,480 for physical therapy and speech-language pathology services combined, and a separate $2,480 for occupational therapy services. Crossing it is not a cap and not a denial. It requires the KX modifier attesting that the care remains medically necessary and that the record supports it. A separate $3,000 targeted medical review threshold, which applies for 2018 through 2028, makes a claim eligible for review by the Supplemental Medical Review Contractor.
How often must an outpatient therapy plan of care be certified?
A physician or non-physician practitioner must certify the plan of care within 30 days of the first day of treatment, and recertify at least every 90 days. Both intervals are set out in the CMS Medicare Benefit Policy Manual, chapter 15, section 220.1.3. Every visit delivered after a lapse is unbillable until the signature is obtained, which is why certification belongs on a calendar with a chase that starts long before the deadline.
Which denials are specific to outpatient therapy?
Nine appear in almost every therapy book: a missing GP, GO, or GN discipline modifier, a missing CQ or CO assistant modifier, unit counts that exceed the documented timed minutes, a lapsed certification or recertification, an exhausted visit limit or authorization, services above the threshold without KX, documentation that shows repetition rather than measurable progress, an unbundled code pair billed without the required modifier, and a claim held so long waiting for a signature that it misses timely filing.
How many units did your practice earn but never bill?
Send us ninety days of therapy claims and treatment minutes. A senior partner returns a written audit: unit allocation variance against the ladder on a real sample, certification lapse exposure, threshold and ABN gaps for the current year, authorization counter risk, and a thirty day corrective plan. Yours to keep.