Arete Healthcare LLC
Independent Practices | Shared Infrastructure | Clinician-Led Governance
Arete Healthcare LLC
Independent Practices | Shared Infrastructure | Clinician-Led Governance
Arete Original Research
Original Medicare Chiropractic and Physical Therapy Service-Unit Denials, Reporting Codes, and Coverage Boundaries, 2015–2023
By David Elton, Arete Healthcare LLC · Published October 23, 2026 · Version 2.1
These seven charts carry the paper’s argument. The first three show that chiropractic’s denial rate depends heavily on which service units are counted. The fourth and fifth show where chiropractic denials come from and how manipulation denials changed. The last two show why headline denial rates fell faster than clinical-code rates. All rates are denied service units divided by submitted service units in Original Medicare Part B data; they are not counts of patients or visits.
Figure 1. Denial rates for 15 specialties, 2023, all submitted codes (gray) versus clinical codes (colored). Chiropractic (dark blue) is highest and physical therapy (light blue) lowest on both measures. Clinical codes exclude reporting-only codes (quality measures and therapy functional reporting) and codes with no Physician Fee Schedule match. Specialties are ordered by clinical-code rate. Pain Medicine combines CMS specialty codes 09 and 72 by summing units before calculating rates. This is not a quality ranking and is not adjusted for differences in what Medicare covers for each specialty. Source: CMS Physician/Supplier Procedure Summary (PSPS) via Mimilabs, SQL1. Visualization was human generated.
Figure 2. The 2023 denial rate depends on what is counted. Chiropractic: 37.7% of all submitted units (n=28.9 million), 33.9% of clinical-code units (n=27.2 million), and 5.4% of spinal manipulation units (CPT 98940–98942) billed with the AT modifier in either modifier position (n=19.0 million). Physical therapy: 6.5% (n=178.9 million) and 5.7% (n=176.4 million). The bracket marks the chiropractic clinical-code rate that Figure 3 breaks down by code. Each rate is correct for its own denominator; they are not interchangeable performance measures. Source: PSPS via Mimilabs, SQL1 and SQL7. Visualization was human generated.
Figure 3. Denial rates by code or code family and specialty, 2023. Spinal manipulation rates (98940–98942) include all modifiers; Figure 5 separates units with and without AT. Extraspinal manipulation (98943) and chiropractor-billed evaluation and management (992xx) were denied 100% of the time. Physical therapy evaluation codes (97161–97164) are a different code family from 992xx, not a clinically equivalent service. For the eight shared treatment codes, the same code was denied 100% when billed by chiropractors and 4.6% to 9.0% when billed by physical therapists. “N/A” marks comparisons outside a profession’s benefit; small volumes do exist (physical therapists billed 451 manipulation units and chiropractors 396 PT evaluation units in 2023, all denied). Source: PSPS via Mimilabs, SQL7. Visualization was human generated.
Figure 4. Clinical-code service units denied each year, 2015–2023. Chiropractic fell from 14.1 million to 9.2 million (−34%); physical therapy was nearly flat (10.3 million to 10.0 million) while its clinical-code submissions grew 59%. In 2023, 81% of chiropractic’s denied clinical-code units were services other than spinal manipulation. Source: PSPS via Mimilabs, SQL1 and SQL7. Visualization was human generated.
Figure 5. Chiropractic spinal manipulation (98940–98942): share of units billed with the AT modifier (upper line) and denial rate for those units (lower line), 2015–2023. AT use held between 95.3% and 96.3%; the denial rate for AT units fell from 11.7% to 5.4%. Annual n is all spinal manipulation units. Manipulation without AT was denied 99.7% to 100% of the time in every year. Missing AT does not by itself prove the service was maintenance care, that an Advance Beneficiary Notice was signed, or that the patient paid. Source: PSPS via Mimilabs, SQL7. Visualization was human generated.
Figure 6. All-code versus clinical-code denial rates, 2015–2023. Chiropractic’s all-code rate fell from 58.4% to 37.7% (−20.7 points) while its clinical-code rate fell from 40.5% to 33.9% (−6.6 points). Physical therapy fell from 24.6% to 6.5% (−18.2 points) versus 9.3% to 5.7% (−3.6 points). The two rates use different denominators, so the difference between their declines is descriptive, not a precise breakdown of causes. Source: PSPS via Mimilabs, SQL1. Visualization was human generated.
Figure 7. Reporting-only submissions, 2015–2023, in millions of submitted units. Measurement codes are quality-reporting codes (Physician Fee Schedule status M, plus CPT Category II codes carrying status I); functional reporting codes are the therapy G-codes (status Q). CMS discontinued mandatory therapy functional reporting for dates of service on and after January 1, 2019; the 4.8 million physical therapy functional-reporting units still recorded in 2019 may reflect claim timing or residual reporting, which these aggregate data cannot separate. Reporting-only units carry no payment and are denied by design. Source: PSPS via Mimilabs, SQL7 (identical totals in SQL6). Visualization was human generated.
The Denial Paradox builds on an earlier Arete Original Research paper, A Decade of Divergence (https://arete.healthcare/research/a-decade-of-divergence), which documented substantial differences in Original Medicare utilization and spending between chiropractic and physical therapy. The Denial Paradox examines another dimension of that divergence: what service denials reveal about the underlying Medicare benefit.
One finding is not surprising. Medicare's chiropractic benefit is extraordinarily narrow. Medicare reimburses chiropractors for medically necessary spinal manipulation, but not evaluation and management services or commonly used therapeutic interventions.
These limitations raise important questions about the adequacy of the chiropractic benefit. Clinical evaluation precedes appropriate treatment, and evidence-informed management of musculoskeletal conditions may involve interventions beyond spinal manipulation. This does not establish that Medicare beneficiaries receive inadequate care. It does, however, raise questions about access, continuity, and the extent to which the chiropractic benefit supports a comprehensive approach to musculoskeletal care.
What was more surprising is the millions of services submitted by chiropractors despite their exclusion from the Medicare benefit. One possibility is that a Medicare denial provides documentation needed to obtain reimbursement from secondary insurance or to establish that a service is the patient's financial responsibility. A planned Arete Analytic Brief will examine why chiropractors continue to bill non-covered services.
The larger picture being revealed by the Arete Original Research series (https://arete.healthcare/research) is becoming increasingly interesting. Chiropractic is an organizational, digital, and research-funding outlier. A Decade of Divergence and The Denial Paradox add Medicare utilization, reimbursement, and benefit design to that picture.
Where might the Arete Original Research series go next?
Commercial reimbursement, provider contracts, contracting intermediaries, network concentration, and administrative requirements are logical areas for further investigation.
Our objective remains straightforward: use publicly available and practice-level data to examine narrow, consequential questions affecting independent outpatient practices; make the methods and limitations transparent; and allow readers to evaluate the evidence.
Table 1. At a Glance: chiropractic and physical therapy, Original Medicare Part B
| Domain | Chiropractic | Physical therapy | Interpretation boundary |
|---|---|---|---|
| All submitted service units, 2023 | 28.9M units; 37.7% denied | 178.9M units; 6.5% denied | Mixes reporting codes, services outside a benefit, and covered services |
| Clinical-code units, 2023 | 27.2M units; 33.9% denied | 176.4M units; 5.7% denied | Not adjusted for what Medicare covers for each profession |
| Spinal manipulation (98940–98942), all modifiers, 2023 | 19.8M units; 9.0% denied | Not applicable | The only chiropractic service Medicare covers |
| Spinal manipulation with AT modifier, 2023 | 19.0M units; 5.4% denied | Not applicable | AT found in either of the two modifier positions PSPS reports |
| Clinical-code units other than spinal manipulation, 2023 | 7.47M units; 100% denied | Not comparable | Services outside the chiropractic benefit; reason for submission not shown |
| Therapeutic exercise (97110), 2023 | 0.87M units; 100% denied | 61.2M units; 4.9% denied | Same code, different coverage by profession |
| Reporting-only share of submitted units, 2015 → 2023 | 30.1% → 5.8% | 16.9% → 0.5% | Reporting codes carry no payment |
| Change in all-code denial rate, 2015 → 2023 | −20.7 points (58.4% → 37.7%) | −18.2 points (24.6% → 6.5%) | Partly reflects fewer reporting-only codes |
| Change in clinical-code denial rate, 2015 → 2023 | −6.6 points (40.5% → 33.9%) | −3.6 points (9.3% → 5.7%) | Different denominator from the row above |
Rates are denied service units divided by submitted service units (M = million). Service units are not patients, visits, or claims. Source: CMS PSPS via Mimilabs, SQL1 and SQL7.
Background. Medicare denial rates are often compared across professions as if they measured how well claims are prepared. A single specialty-wide rate, however, combines four different things: reporting-only codes that carry no payment, services outside a profession’s Medicare benefit, codes Medicare does not accept, and potentially covered clinical services. As a companion to A Decade of Divergence, this paper examines how separating these changes the picture for chiropractic and physical therapy.
Methods. Retrospective, descriptive analysis of the CMS Physician/Supplier Procedure Summary (PSPS), 2015–2023, for 15 specialty groups, accessed through Mimilabs. Each procedure code and year was matched to that year’s Medicare Physician Fee Schedule status to separate reporting-only codes from clinical codes. A detailed extract for chiropractic and physical therapy, by code and by the two modifier positions PSPS reports, was used for code-level and modifier-level results. All rates are denied service units divided by submitted service units.
Results. In 2023, chiropractic had the highest denial rate of the 15 groups (37.7% of all units; 33.9% of clinical-code units) and physical therapy the lowest (6.5%; 5.7%). Spinal manipulation billed with the AT modifier was denied 5.4% of the time, down from 11.7% in 2015. All 7.47 million 2023 chiropractic clinical-code units for services other than spinal manipulation were denied, accounting for 80.8% of chiropractic clinical-code denials. Identical therapy codes were denied 100% of the time for chiropractors and about 5% for physical therapists. From 2015 to 2023, all-code rates fell much more than clinical-code rates (chiropractic 20.7 versus 6.6 points; physical therapy 18.2 versus 3.6 points), alongside a steep decline in reporting-only submissions.
Conclusions. Medicare denial comparisons between chiropractic and physical therapy are highly sensitive to what is counted. Most of chiropractic’s denials reflect the narrow scope of its Medicare benefit and submissions of services outside it, not denials of covered manipulation. These data do not show why noncovered services were submitted, whether the care was provided, or who paid for it.
This is a national, repeated cross-sectional, descriptive analysis of Original Medicare Part B physician and supplier claims summarized in the CMS Physician/Supplier Procedure Summary (PSPS), performance years 2015 through 2023. PSPS reports, for each specialty, procedure code, and modifier combination, the number of service units submitted and denied. Data were accessed through the Mimilabs mirror of the CMS file (table mimi_ws_1.datacmsgov.psps). The unit of analysis is a submitted service unit, not a claim, visit, patient, or provider. Some unit counts are not whole numbers in the source; all calculations use the unrounded values.
Fifteen specialty groups were selected in advance, defined by the CMS provider specialty code on the claim: chiropractic (35), physical therapy (65), orthopedic surgery (20), pain medicine (09 and 72, combined), physical medicine and rehabilitation (25), family medicine (08), internal medicine (11), emergency medicine (93), nurse practitioner (50), physician assistant (97), optometry (41), clinical psychologist (62), clinical social worker (80), podiatry (48), and psychiatry (26). Pain medicine rates were calculated from summed units, not by averaging the two codes’ rates. Physical therapy here means therapists billing on Part B professional claims under specialty 65; therapy billed by hospitals and other facilities on institutional claims is not in PSPS.
Each PSPS code and year was matched to the same year’s Medicare Physician Fee Schedule relative value file (mimi_ws_1.cmspayment.rvu_pprrvu), using the status indicator from the latest file released within that year. The status indicator describes a code’s general payment status under the fee schedule. It does not say whether a service is covered when a particular profession provides it. Table 2 shows how each status was treated. This grouping is referred to as classification v4 in the study files (labelled v3.1 inside the SQL documentation; the content is the same).
Table 2. How fee schedule status was used to group codes
| Fee schedule status | Treatment in this analysis | What to keep in mind |
|---|---|---|
| M (measurement); Q (therapy functional reporting) | Reporting-only; excluded from clinical codes | Carry no payment; denied by design |
| I, when the code is a CPT Category II quality code (four digits followed by F) | Reporting-only; excluded from clinical codes | Same purpose as status M codes |
| I, all other codes (not valid for Medicare) | Included in clinical codes | Can be a real clinical service billed with a code Medicare does not accept (for example, 97014), a supply, or an administrative item |
| T (paid only if no other fee schedule service is paid that day) | Included in clinical codes | Very small volume (254 chiropractic and physical therapy units over nine years) |
| A, B, C, D, E, J, N, P, R, X | Included in clinical codes | Mix of payable, bundled, noncovered, and excluded services |
| No fee schedule match | Excluded from clinical codes; kept in all-code rates | Largest example: COVID-19 home test kits (K1034) billed by physical therapists in 2023 |
“Clinical codes” is therefore a practical grouping, not a claim that every unit in it was clinical treatment, and not a list of services Medicare covers for a given profession. A narrower “potentially payable” subset (status A, C, D, E, J, P, R) is kept in the internal reproducibility record for completeness; it has the same limitation.
The all-code denial rate is total denied units divided by total submitted units. The clinical-code denial rate uses the same calculation restricted to clinical codes. The reporting-only share is reporting-only submitted units divided by all submitted units. All sums are calculated before rounding.
A second extract (SQL7) kept the exact procedure code and the first two modifier positions for chiropractic and physical therapy, with no minimum volume. It reconciles to the specialty totals in SQL1 within rounding for every specialty and year. Spinal manipulation means CPT 98940–98942. A unit was counted as having the AT modifier if AT appeared in either modifier position. Services outside the chiropractic benefit were examined in two ways: (1) all clinical-code units other than spinal manipulation, and (2) four code families: therapy and modality codes (97xxx, which also includes acupuncture codes 97810–97814), G0xxx procedures (mostly G0283, electrical stimulation), evaluation and management (992xx), and extraspinal manipulation (98943).
Medicare Part B covers a chiropractor’s services only for manual manipulation of the spine to correct a subluxation; maintenance therapy is not considered medically reasonable and necessary (Medicare Benefit Policy Manual, Chapter 15). Medicare Administrative Contractor guidance requires the AT modifier on manipulation claims for active or corrective treatment and treats manipulation claims without it as not medically necessary. The same guidance states that chiropractors are not required to submit services outside the benefit, but may do so to obtain a denial needed to bill a secondary insurer. An Advance Beneficiary Notice is not required for services that are excluded by law. PSPS does not contain notices, denial reason codes, secondary payer payments, patient invoices, or collections.
The SQL queries, independent verification, and reconciliation record are kept in the internal AI Assistance and Reproducibility Record maintained under the Arete Research Library Governance Framework.
In 2023, chiropractic had the highest denial rate of the 15 specialty groups on both measures, and physical therapy the lowest (Table 3). Separating reporting-only codes changed the picture for some specialties more than others. Family medicine fell from 21.0% to 8.1% and internal medicine from 16.8% to 6.2%, because reporting codes made up 13.8% and 11.1% of their submissions. Chiropractic’s rate changed less (37.7% to 33.9%) because only 5.8% of its 2023 submissions were reporting codes.
Table 3. Denial rates by specialty, 2023, ranked by clinical-code rate
| Specialty | Submitted units | All codes | Clinical codes | Reporting-only share |
|---|---|---|---|---|
| Chiropractic | 28.9M | 37.7% | 33.9% | 5.8% |
| Optometry | 24.8M | 36.4% | 22.1% | 18.4% |
| Pain Medicine (09 + 72) | 20.3M | 15.4% | 10.6% | 5.3% |
| Podiatry | 38.7M | 14.6% | 8.7% | 6.2% |
| Clinical Social Worker | 8.0M | 11.2% | 8.5% | 3.0% |
| PM&R | 31.6M | 11.4% | 8.2% | 3.5% |
| Family Medicine | 154.2M | 21.0% | 8.1% | 13.8% |
| Emergency Medicine | 26.8M | 9.7% | 8.0% | 1.8% |
| Psychiatry | 11.1M | 13.8% | 7.3% | 4.1% |
| Nurse Practitioner | 211.7M | 14.1% | 7.1% | 7.1% |
| Orthopedic Surgery | 54.4M | 10.3% | 6.8% | 3.6% |
| Physician Assistant | 91.4M | 12.0% | 6.6% | 5.7% |
| Internal Medicine | 212.8M | 16.8% | 6.2% | 11.1% |
| Clinical Psychologist | 0.1M | 9.7% | 6.0% | 4.0% |
| Physical Therapy | 178.9M | 6.5% | 5.7% | 0.5% |
Not a quality ranking and not adjusted for differences in what Medicare covers for each specialty. Clinical psychologist volume is small (111,670 units). Source: PSPS via Mimilabs, SQL1.
The 2023 chiropractic denial rate was 37.7% across all submitted units, 33.9% across clinical-code units, 9.0% across all spinal manipulation units, and 5.4% across spinal manipulation units billed with the AT modifier. The physical therapy rates were 6.5% and 5.7%. Each figure is accurate for its own denominator. None alone measures how often clinically appropriate care was denied, or how many patients lost access to care.
The same procedure code had very different results depending on which profession billed it (Table 4). Every 2023 chiropractic unit for the eight shared treatment codes was denied, compared with 4.6% to 9.0% for physical therapy. Chiropractor-billed evaluation and management codes (992xx; 843,340 units) and extraspinal manipulation (98943; 780,180 units) were also denied 100% of the time. Physical therapy evaluation codes (97161–97164; 4,108,953 units) were denied 5.6% of the time; these are different codes from 992xx and are shown for context, not as an equivalent service.
Table 4. Selected codes billed by both professions, 2023
| Code | Service | DC units | DC denied | PT units | PT denied |
|---|---|---|---|---|---|
| 97110 | Therapeutic exercise | 873,917 | 100.0% | 61,153,405 | 4.9% |
| 97140 | Manual therapy | 849,334 | 100.0% | 28,399,618 | 4.9% |
| 97112 | Neuromuscular re-education | 191,409 | 100.0% | 29,126,674 | 4.6% |
| 97530 | Therapeutic activities | 188,794 | 100.0% | 32,630,411 | 4.6% |
| 97012 | Mechanical traction | 942,804 | 100.0% | 398,559 | 5.5% |
| G0283 | Electrical stimulation | 876,458 | 100.0% | 5,685,988 | 5.7% |
| 97124 | Massage | 226,031 | 100.0% | 282,310 | 9.0% |
| 97035 | Ultrasound | 286,879 | 100.0% | 986,095 | 5.9% |
DC = chiropractic; PT = physical therapy. Units are rounded to whole numbers for display. Source: PSPS via Mimilabs, SQL7.
Chiropractic clinical-code denials fell from 14.1 million units in 2015 to 9.2 million in 2023 (−34%). Physical therapy’s were nearly flat (10.3 million to 10.0 million) even though its clinical-code submissions rose 59%, from 110.7 million to 176.4 million units. Within chiropractic, the share of clinical-code denials coming from services other than spinal manipulation rose from 72.8% in 2015 to 80.8% in 2023 (Table 5). Those units were denied essentially 100% of the time in every year. The four code families account for most of them: 7,141,471 units in 2023 (97xxx 4.64 million; G0xxx 0.88 million, almost all G0283; 992xx 0.84 million; 98943 0.78 million), down from 9,865,978 in 2015. The remaining 2023 units were mainly X-ray codes (72xxx) and other low-volume codes.
Table 5. Chiropractic clinical-code denials: spinal manipulation versus all other codes
| Year | Clinical-code units denied | Spinal manipulation denied | Other codes denied | Other codes, share of denials |
|---|---|---|---|---|
| 2015 | 14.08M | 3.84M | 10.25M | 72.8% |
| 2016 | 13.30M | 3.23M | 10.07M | 75.7% |
| 2017 | 12.39M | 2.84M | 9.55M | 77.0% |
| 2018 | 11.23M | 2.83M | 8.40M | 74.8% |
| 2019 | 10.89M | 2.68M | 8.22M | 75.4% |
| 2020 | 8.74M | 2.15M | 6.59M | 75.4% |
| 2021 | 9.59M | 2.13M | 7.46M | 77.8% |
| 2022 | 9.33M | 1.94M | 7.39M | 79.2% |
| 2023 | 9.24M | 1.77M | 7.47M | 80.8% |
Spinal manipulation = 98940–98942, all modifiers. “Other codes” were submitted in almost exactly the same volume as they were denied in every year (99.99% or more). Source: PSPS via Mimilabs, SQL7.
About 95% to 96% of chiropractic spinal manipulation units carried the AT modifier in every year (Table 6). The denial rate for those units fell from 11.7% in 2015 to 5.4% in 2023, and denied AT units fell from 2.73 million to 1.02 million. Units without AT, 0.75 million in 2023 (3.8% of manipulation units), were denied 99.7% to 100% of the time. Most of the remaining 2023 units without AT carried GA (an Advance Beneficiary Notice is on file), carried no modifier, or carried GY or GZ. A missing AT modifier does not by itself show that the care was maintenance care, that a notice was properly delivered, or that the patient paid.
Table 6. Chiropractic spinal manipulation (98940–98942) by AT modifier
| Year | Units with AT | Denied, with AT | Units without AT | Denied, without AT | Share with AT |
|---|---|---|---|---|---|
| 2015 | 23.40M | 11.7% | 1.11M | 99.7% | 95.5% |
| 2016 | 22.76M | 9.4% | 1.08M | 99.9% | 95.5% |
| 2017 | 21.80M | 8.1% | 1.08M | 99.9% | 95.3% |
| 2018 | 21.83M | 8.2% | 1.03M | 100.0% | 95.5% |
| 2019 | 21.84M | 8.0% | 0.93M | 100.0% | 95.9% |
| 2020 | 18.12M | 7.5% | 0.78M | 100.0% | 95.9% |
| 2021 | 19.62M | 7.0% | 0.75M | 100.0% | 96.3% |
| 2022 | 19.10M | 6.2% | 0.76M | 100.0% | 96.2% |
| 2023 | 19.03M | 5.4% | 0.75M | 100.0% | 96.2% |
AT counted if present in either of the two modifier positions PSPS reports. Values shown as 100.0% are at least 99.98% before rounding. Source: PSPS via Mimilabs, SQL7.
Both professions’ all-code denial rates fell much more than their clinical-code rates between 2015 and 2023 (Table 7). Over the same period, reporting-only submissions dropped sharply. Chiropractic measurement-code units fell from 14.95 million to 1.67 million (30.1% to 5.8% of its submissions). Physical therapy functional-reporting units rose from 12.22 million in 2015 to 14.77 million in 2018, fell to 4.80 million in 2019 after CMS ended the reporting requirement, and were absent from 2020 onward; its measurement-code units fell from 10.35 million to 0.92 million. Because reporting-only units are denied by design, fewer of them pulls the all-code rate down without any change in how clinical services are handled. The two rates use different denominators, so the gap between their declines is a description, not an exact accounting of causes.
Table 7. Annual denial rates, 2015–2023
| Year | DC all codes | DC clinical codes | DC reporting-only share | PT all codes | PT clinical codes | PT reporting-only share |
|---|---|---|---|---|---|---|
| 2015 | 58.4% | 40.5% | 30.1% | 24.6% | 9.3% | 16.9% |
| 2016 | 58.4% | 39.2% | 31.5% | 23.8% | 8.3% | 16.9% |
| 2017 | 57.2% | 38.2% | 30.8% | 19.9% | 8.1% | 12.8% |
| 2018 | 54.8% | 35.9% | 29.5% | 18.9% | 7.8% | 12.1% |
| 2019 | 52.7% | 35.2% | 27.0% | 11.3% | 7.1% | 4.5% |
| 2020 | 50.5% | 34.3% | 24.6% | 7.6% | 6.5% | 1.1% |
| 2021 | 39.3% | 34.4% | 7.4% | 7.4% | 6.6% | 0.9% |
| 2022 | 38.5% | 34.2% | 6.5% | 6.8% | 6.2% | 0.6% |
| 2023 | 37.7% | 33.9% | 5.8% | 6.5% | 5.7% | 0.5% |
DC = chiropractic; PT = physical therapy. Source: PSPS via Mimilabs, SQL1.
Electrical stimulation: 97014 and G0283. Medicare does not accept CPT 97014 (status I) and uses G0283 for unattended electrical stimulation instead. In 2023, chiropractors submitted 329,117 units of 97014 and 876,458 units of G0283; all were denied. Physical therapists’ G0283 units were denied 5.7% of the time. G0283 is an active, payable code, so the chiropractic denials reflect the benefit, not the code.
COVID-19 home test kits: K1034. Physical therapists submitted 1,511,792 units of K1034 in 2023, of which 682,369 were denied. The code has no fee schedule match, so it is excluded from clinical codes but included in the all-code rate. Without it, physical therapy’s 2023 all-code rate would be 6.2% rather than 6.5%.
Sensitivity to the grouping rules. An earlier grouping (v3) left all status I and status T codes out of clinical codes. Under it, chiropractic’s 2023 clinical-code rate was 33.0% rather than 33.9%, mainly because 97014 moved into clinical codes in v4; physical therapy was 5.7% under both. The direction of every comparison is unchanged.
A single Medicare denial rate for a profession blends payment decisions on covered services with reporting codes, services the profession’s benefit does not include, and codes Medicare does not accept. For chiropractic the blend matters a great deal. The covered service, spinal manipulation billed as active treatment, was denied 5.4% of the time in 2023. The 33.9% clinical-code rate is driven mainly by 7.47 million units of services that Medicare does not cover when a chiropractor provides them, all of which were denied. Comparing chiropractic’s 33.9% with physical therapy’s 5.7% therefore compares two different benefits more than two levels of billing accuracy.
One documented reason is secondary coverage: Medicare contractor guidance states that chiropractors may submit services outside the benefit to obtain a denial needed to bill a secondary insurer. Other possible reasons include a patient’s request for an official coverage decision, practice billing software or workflows that send all services to Medicare, mistaken expectations of payment, and the choice to submit rather than bill the patient directly. PSPS cannot separate these. None of them shows that the patient ultimately paid.
For anyone comparing denial rates (payers, analysts, associations, journalists): state the denominator. All-code, clinical-code, and covered-service rates answer different questions and can differ several-fold for the same profession.
For chiropractic practices: the data are consistent with Medicare policy that the AT modifier is required for manipulation to be considered for payment; manipulation billed without it was almost always denied.
For policy discussions: the gap between chiropractic and physical therapy denial rates mostly reflects how the Medicare chiropractic benefit is defined, a benefit-design question, rather than how claims are processed.
The companion paper described Original Medicare payments, services, and code mix for 2013–2024 from CMS files that report paid services. This paper uses PSPS, which also records submitted and denied services that generated no payment, for 2015–2023. The two datasets have different inclusion and summarization rules, and their totals are not directly comparable. Together they separate the care Medicare paid for from billing activity visible only in submissions.
A separate Arete Analytic Brief is planned to examine, in participating practices’ own records, how often Original Medicare patients receive services outside the Medicare benefit and how those services are billed and paid. No practice-level data are used in this paper.
PSPS counts service units. It does not identify patients, visits, or episodes of care, and it does not show whether a denied service was provided, appealed, paid by another insurer, or billed to the patient.
PSPS has no denial reason codes. A denial may reflect a statutory exclusion, a medical-necessity decision, incorrect coding, or an administrative rule, and these cannot be separated.
Clinical codes are an operational grouping based on general fee schedule status. They include some supplies and administrative items, and they are not adjusted for what Medicare covers for each profession.
Fee schedule status was taken from the latest file released in each year, not matched to each date of service. A check using the earliest file in each year gave identical clinical-code rates to three decimal places; that check was run under the earlier v3 grouping.
Only the first two modifier positions are available. A modifier in a later position, claim-level sequencing, and Advance Beneficiary Notice records are not observable.
Physical therapy results cover therapists billing Part B professional claims under specialty 65. Therapy provided in hospital outpatient departments and other facilities is billed on institutional claims and is not included.
Specialty is the billing provider’s specialty on the claim. Services billed under another provider’s number are attributed to that provider’s specialty.
Codes in the 97xxx family use different unit conventions, and some are billed in multiple units per visit. Unit counts are not comparable to visit or patient counts.
PSPS reports data by performance year. The 2019 physical therapy functional-reporting units may reflect claim timing or residual reporting after the requirement ended; the data cannot tell which.
The 15 specialty groups were selected for comparison. The rankings are descriptive and are not a measure of quality of care or administrative performance.
This work is classified as a public-data reimbursement and coverage analysis under the Arete Research Library Governance Framework (§2); it is not PBRN-sponsored human-subjects research. This analysis was produced by Arete Healthcare without external funding; Arete supports independent outpatient chiropractic and physical therapy practices and has a commercial and professional interest in Medicare coverage, billing, and payer policy affecting those practices. AI tools assisted with data review, calculation, and drafting. Every factual assertion and calculation requires responsible human verification before publication, and AI reconciliation does not substitute for the named Independent Data/Methodology Reviewer or DAWG approval.
AI use disclosure. Drafts before Version 1.1 were prepared in October 2026 with assistance from ChatGPT (OpenAI; GPT-5.6 Sol, the same version used for this series’ other Original Research papers). The SQL queries were drafted by Mimibot, the Mimilabs AI assistant, and run in Mimilabs. Versions 1.1 through 2.1 were rebuilt in the standard Original Research format, and every statistic was independently recomputed from the source extracts, with assistance from Claude (Anthropic; Claude Opus 5.5) on October 21, 2026. All figures were produced by the Author without AI assistance. Arete’s Take was written by the Author; AI was used only to place the Author’s text in the manuscript. The Author takes responsibility for the accuracy of the content. Details are kept in the internal AI Assistance and Reproducibility Record maintained under the Arete Research Library Governance Framework.
Data. Aggregated public CMS data only. No patient-identifiable information was used.
Review status. Approved for release. Two non-conflicted DAWG reviewers approved the Findings, Methods, and Results (Version 1.4), and at least two of the three Arete Executive Reviewers approved Arete’s Take, independently of DAWG (Arete Research Library Governance Framework §1.3 and §7). No findings, figures, tables, or Arete’s Take text changed at release.
Recommended Citation: Elton, D. The Denial Paradox. Arete Original Research ARETE-OR-2026-013, Version 2.1. Arete Healthcare LLC. Published October 23, 2026. https://arete.healthcare/research/the-denial-paradox https://doi.org/10.13140/RG.2.2.30472.58881
Centers for Medicare & Medicaid Services. Physician/Supplier Procedure Summary. https://data.cms.gov/summary-statistics-on-use-and-payments/physiciansupplier-procedure-summary
Centers for Medicare & Medicaid Services. Physician Fee Schedule Relative Value Files. https://www.cms.gov/medicare/payment/fee-schedules/physician/pfs-relative-value-files
Centers for Medicare & Medicaid Services. Status Indicators (Physician Fee Schedule). https://www.cms.gov/help/status-indicators
Centers for Medicare & Medicaid Services. Medicare Benefit Policy Manual, Chapter 15, §30.5 (Chiropractor’s Services) and §240 (Chiropractic Services). https://www.cms.gov/regulations-and-guidance/guidance/manuals/downloads/bp102c15.pdf
Centers for Medicare & Medicaid Services. Functional Reporting (therapy services; discontinued for dates of service on and after January 1, 2019, per CMS-1693-F). https://www.cms.gov/medicare/billing/therapyservices/functional-reporting
Wisconsin Physicians Service Insurance Corporation (Medicare Administrative Contractor). Billing and Coding: Chiropractic Services (A56273). https://www.cms.gov/medicare-coverage-database/view/article.aspx?articleid=56273
CGS Administrators, LLC (Medicare Administrative Contractor). Billing and Coding: Chiropractic Services (A56455). https://www.cms.gov/medicare-coverage-database/view/article.aspx?articleId=56455
Mimilabs. Mirrored tables mimi_ws_1.datacmsgov.psps and mimi_ws_1.cmspayment.rvu_pprrvu; queries SQL1, SQL6, and SQL7, extracted October 2026.
Elton, D. A Decade of Divergence. Arete Original Research ARETE-OR-2026-004, Version 15.1. Arete Healthcare LLC; October 19, 2026. https://arete.healthcare/research/a-decade-of-divergence https://doi.org/10.13140/RG.2.2.29102.88646
Terms used in this paper, in plain language.
| Term | Definition |
|---|---|
| ABN | Advance Beneficiary Notice: a form telling a Medicare patient, before the service, that Medicare may not pay and the patient may be responsible. |
| AT modifier | Code added to a chiropractic manipulation claim to show active or corrective treatment. Required for Medicare to consider paying for manipulation. |
| Clinical codes | In this paper, all codes except reporting-only codes and codes with no fee schedule match. A practical grouping, not a list of covered services. |
| CMT | Chiropractic manipulative treatment. Spinal CMT is CPT 98940–98942; 98943 is extraspinal (arms, legs, and other non-spinal areas). |
| CPT / HCPCS | Standard code sets used to bill medical services and supplies. CPT Category II codes (four digits and an F) are used for quality reporting. |
| Denial rate | Denied service units divided by submitted service units. |
| E/M (992xx) | Evaluation and management codes for office and other visits. |
| Fee schedule status | A letter on each code in the Medicare Physician Fee Schedule describing its general payment status (for example, A = active, I = not valid for Medicare, M = measurement). |
| Functional reporting | Therapy G-codes (status Q) once required on therapy claims to report a patient’s function. Discontinued for dates of service on and after January 1, 2019. |
| GA, GY, GZ | Modifiers: GA = an ABN is on file; GY = service excluded by law or outside a benefit; GZ = denial expected and no ABN on file. |
| MAC | Medicare Administrative Contractor: the company that processes Medicare claims for a region. |
| Original Medicare | Traditional fee-for-service Medicare (Parts A and B), not Medicare Advantage. |
| PSPS | Physician/Supplier Procedure Summary: a CMS file summarizing Part B professional and supplier claims by specialty, code, and modifier. |
| Reporting-only codes | Codes used to report quality or function. They carry no payment and are denied by design. |
| Service unit | One billed unit of a code. Some services are billed in several units per visit, so units are not visits or patients. |
Download: The Denial Paradox (PDF)