← All resources

21 compliance checks for therapy documentation

The compliance dimensions that decide whether SNF therapy documentation survives Medicare review — medical necessity to case-mix consistency.

When a Medicare Administrative Contractor, RAC auditor, or state surveyor opens a therapy chart, they are not reading it the way a clinician does. They are running a checklist — and in documentation reviews, denied claims and recoupment demands trace back to checklist items the record failed.

Skilled Nursing Facility (SNF) therapy documentation is evaluated across a surprisingly stable set of dimensions. Here is the practical map — what each dimension asks, and what “failing it” looks like in a real chart.

The big four: where most money is lost

1. Medical necessity. Does the record support that skilled care was necessary for this resident, this week? “Resident tolerated treatment well” is the classic failure — it describes attendance, not necessity. The standard, rooted in 42 CFR §409.32 and reinforced by Jimmo v. Sebelius, is that skilled care can be justified by the need to maintain function, not only to improve it — but the documentation must actually make that case.

2. Skilled service justification. Closely related but distinct: why did this intervention require a therapist rather than a caregiver or restorative aide? Notes that describe what was done without why it took a licensed clinician fail here.

3. Plan of care compliance. The plan says three visits per week for 4 weeks; the record shows 5 visits one week and 2 the next, with no order change. Frequency, duration, and intensity in the documentation must match the certified plan — and the certification dates must cover every visit billed.

4. Timeliness. Evaluations completed within the required window, Medicare Part B progress reports on their treatment-day clock, recertifications signed before the period lapses (delayed certifications are allowed only with a documented, acceptable reason per the Benefit Policy Manual). Documentation that misses its window with no compliant cure is treated as missing.

Quality of the clinical record

5. Accuracy and consistency. Contradictions between disciplines, between the note and the MDS, or between minutes documented and minutes billed.

6. Objective measurement and functional outcomes. “Improving” is not a measurement. Reviewers look for baselines, standardized test scores, assist levels, and distances.

7. Goal compliance. Goals must be measurable, functional, and time-bound. “Improve transfers” fails; “resident will complete bed-to-wheelchair transfer with contact-guard assist within 2 weeks” passes.

8. Progress notes. Under Medicare Part B, the required progress report is written by the therapist (not the assistant) on its treatment-day interval; Part A and other payers set their own documentation schedules.

9. Discharge documentation. A summary that accounts for the episode: status against goals, discharge disposition, and recommendations.

10. Frequency, duration and intensity alignment. The billed minutes and visit pattern must be clinically coherent with the documented condition — a flat, identical minutes pattern every week for every resident is a pattern reviewers notice.

Payment-system consistency

11. PDPM compliance. Under the Patient Driven Payment Model, the clinical record must support the case-mix classification — diagnoses, function scores, and therapy delivery all consistent with what was submitted.

12. Signature and authentication rules. The right credential, in the right order, at the right time. Unsigned notes and missing co-signatures for assistants where required are findings — and backdated signatures are prohibited outright (Program Integrity Manual), though properly labeled late entries, addenda, and corrections that meet the applicable requirements remain permitted.

13. Scope of practice. PTAs and COTAs documenting or performing only what their license and state practice act allow — and evaluations always by the therapist.

The rest of the checklist

14. Patient participation and tolerance — documented engagement, refusals handled and followed up. 15. Safety and risk — precautions documented and observed in the record. 16. Interdisciplinary communication — therapy consistent with nursing and physician documentation. 17. Regulatory audit readiness — the chart can stand alone without verbal explanation. 18. Fraud, waste and abuse safeguards — no cloned notes, inflated minutes, or fabricated entries. 19–21. Evaluation completeness, billing and coding compliance, physician orders and referrals — the paperwork chain from order to claim, intact.

Why manual audit programs miss these

Not because auditors don’t know the rules — because arithmetic is against them. A conscientious internal auditor covers a handful of charts per day. A 100-bed facility generates thousands of pages of documentation a month. Sampling is the only coping strategy, which means most documentation is never reviewed by anyone before a payer reviews it.

The failure modes follow directly: findings depend on who audited, feedback reaches therapists weeks after the note was written (if ever), and the same issues recur because nothing connects audit results to training.

What good looks like now

Facilities are increasingly moving to 100% audit coverage: every document checked against every dimension, automatically, as documentation is created — with findings that carry their evidence (document, page, section) and their regulatory citation, assigned to a person, and tracked to resolution. The role of the human compliance professional shifts from sampling charts to managing a program: reviewing findings, tuning thresholds, and closing the loop through targeted training.

That last step matters most. An audit that finds the same “goal lacks baseline” finding every month is measuring a training problem. The programs that work treat every recurring finding as a curriculum item — which is exactly how the audit-plus-coaching loop in CareDocs.ai Compliance Agent is designed.

Want to see your own documentation run through all 21 dimensions? Request a demo — bring a de-identified sample and watch the findings arrive with their evidence.

See it on your own workflows.

Bring a de-identified eval, schedule, or contract — most teams know within one demo.