Occupational therapy billing sits in the same units-based, prior-auth-heavy world as physical and speech therapy, but OT denials have a distinct profile. The GO modifier requirement catches practices off guard (many teams assume OT uses GP like PT, or GN like SLP, and get it wrong). Medical necessity disputes hinge on activities-of-daily-living documentation that is harder to objectify than range-of-motion measurements. And authorization paths for evaluation versus ongoing treatment are often on separate tracks that billers collapse into one. Four patterns account for the majority of OT claim denials. Working them as a cluster is faster than treating each one as its own problem.
The GO modifier and CO-4
Medicare uses therapy-type modifiers to identify which discipline provided a service. Physical therapy uses GP. Speech-language pathology uses GN. Occupational therapy uses GO. Without GO on every OT procedure line, Medicare denies the claim as CO-4. The corrected-claim fix is simple: resubmit with GO on every line. The durable fix is a billing rule that appends GO automatically to all OT procedure codes before submission, so this cannot recur from a manual oversight.
Commercial payers do not uniformly require GO. Some do; some identify OT through taxonomy code and do not require a therapy modifier at all. Verify by payer. A CO-4 from a commercial carrier on an OT claim often has a different cause than a missing modifier.
Evaluation plus treatment on the same day, and CO-97
When an occupational therapist performs an initial evaluation (97165, 97166, or 97167) and a therapeutic procedure on the same date of service, payers commonly bundle the procedure into the evaluation as CO-97. The evaluation is treated as the primary service and the procedure is considered included.
- Check the NCCI modifier indicator for the specific evaluation-plus-procedure code pair. A modifier indicator of 1 means the bundling can be overridden; indicator 0 means it cannot.
- For pairs that allow an override, append modifier 59 (or a more specific X-modifier) to the therapeutic procedure line and submit a corrected claim. The documentation must support the procedure as a distinct, separately identifiable service beyond the evaluation itself.
- If the same-day procedure was genuinely part of the evaluation, the bundling is correct and the CO-97 stands.
Medical necessity documentation and CO-50
CO-50 is the hardest OT denial to win because the line between skilled occupational therapy and supervisory or maintenance care is subjective, and payers exploit that ambiguity. The core question a utilization reviewer asks is whether the service required the clinical skills of a licensed OT, or whether it could have been carried out by a caregiver following written instructions.
Documentation that supports a successful necessity appeal covers four things:
- Baseline functional deficit. Specific ADL or IADL limitations measured at the start of the episode, for example the patient's ability to dress, bathe, manage household tasks, or return to work tasks, stated in functional terms, not just diagnostic codes.
- Skilled-care rationale. Why this specific intervention required an OT rather than a trained caregiver. Cognitive retraining, sensory processing interventions, upper-extremity therapeutic exercises requiring clinical judgment, and adaptive equipment prescription and training are examples that tend to support necessity.
- Objective progress. Measured improvement across sessions tied to the functional baseline, not just "patient tolerated session well" or "patient followed cues appropriately." Functional independence measure scores, grip strength, standardized ADL scales, or work-capacity measurements provide the data reviewers are looking for.
- Discharge endpoint. A treatment plan with a defined functional goal and an expectation of when the patient will reach it. Open-ended treatment plans invite continued-stay denials.
Request the payer's clinical criteria before writing the appeal letter. You are entitled to them, and appeals built around the criteria the payer actually applied win at a materially higher rate than appeals that resubmit the original note.
Authorization gaps between evaluation and treatment
A structural billing problem specific to therapy practices: many commercial payers authorize the initial OT evaluation and ongoing therapy sessions on separate authorization tracks. A practice that obtains an evaluation authorization and then starts treatment sessions under the same auth number is billing treatment without a valid treatment authorization. That produces CO-197 or CO-15 denials on every treatment session, sometimes going back weeks before anyone notices.
The fix is procedural. When the evaluation authorization is approved, submit the treatment authorization request before the first treatment visit. Many payers require the evaluation report and treatment plan as supporting documentation for the treatment auth, so build the treatment auth request into the workflow that generates the evaluation report. Treating it as a separate follow-up step is where the gap opens.
Also confirm the payer's authorization categories at scheduling. Hand therapy, cognitive rehabilitation, low-vision OT, and driver rehabilitation often require separate authorization tracks from standard ADL-focused outpatient OT. One authorization number commonly does not cover all OT service types for a single patient.
Starting the denial audit
Pull the last 90 days of OT denials and sort by CARC code. A CO-4 cluster from Medicare almost always resolves with a GO modifier billing rule. A CO-97 cluster that tracks to evaluation dates points to a consistent modifier-59 gap on same-day procedures. A CO-50 cluster points to documentation templates that are not generating the functional outcome data payers look for. A CO-197 or CO-15 cluster tracking to treatment sessions points to an authorization workflow that does not separate evaluation and treatment auth tracks.
Once the cluster patterns are clear, each fix is a single upstream change rather than individual claim rework. For a broader view of the revenue loss that never appears in a denial queue, see the guide on silent denials and the one on small-balance revenue leakage.