Audit & compliance

Your internal chart audit fills a binder every month. What did it actually fix?

Most small programs audit charts because an accreditor or a compliance plan says to. Done well, it is the cheapest early warning you have. Here is how to size the sample, build the checklist, assign reviewers and close the loop in a month.

Samuel Jean, Co-Founder22 October 20268 min read
The short version
  • Pick a sample small enough to finish every month, and keep it consistent so this month can be compared with last month.
  • Use one fixed, one-page checklist with yes, no and not applicable answers, each defined in writing so two reviewers score the same chart the same way.
  • Nobody audits their own charts, and the person who checks the claim side should know billing, not just clinical documentation.
  • Sort findings by cause (knowledge, workflow, system, individual) before you sort them by clinician, and fix the cause.
  • Every finding gets an owner and a date, and the item that failed this month is re-checked next month.

Somewhere in most small programs there is a binder, or a shared folder, holding a year of internal chart audits. Each month has a completed checklist, a score, maybe a short summary. Ask what changed because of the March audit and the room goes quiet. The same unsigned-note finding appears in March, May and August. The audit happened. It just never connected to anything. This post is about running a monthly internal audit at a program of a handful to a few dozen clinical staff so that it produces fixes, not paper.

Why internal audits drift into binders

The pattern is predictable. Someone builds an ambitious checklist with forty items. The first month, the clinical director reviews twenty charts and it takes two full days. The second month it slips. By the third month the sample has quietly shrunk, the checklist has been answered from memory, and the summary says what it said last time. Findings are written up, filed, and never assigned to anyone.

Three design choices prevent most of that: a sample you can actually finish, a checklist short enough to score honestly, and a meeting where every finding leaves with a name attached. None of these require software or a consultant. They require deciding in advance and then not renegotiating every month.

Pick a sample you can finish every month

There is no universal sample size for an internal audit, and you should check whether your accreditor, state licensing rules or any payer contract specify one for your program. Absent a requirement, size it to the hours you can protect. A consistent small sample run every month beats an impressive one run twice a year.

A workable structure is a fixed number of charts per clinician, split between random and targeted selection. Say a hypothetical outpatient and IOP program has eight clinicians. Two charts per clinician per month is sixteen charts. If each chart review takes twenty to thirty minutes with a good checklist, that is roughly a day of reviewer time, spread across two or three people.

  • Random charts tell you what normal looks like. Pull them from the active census by a method you write down, such as every nth patient on a sorted list, so nobody hand-picks the tidy ones.
  • Targeted charts go where risk lives: new clinicians in their first months, patients who changed level of care, recent discharges, high-utilization episodes, anything that came back denied, and any clinician with a prior finding.
  • A defined window. Audit services from a specific date range, typically the prior month, so findings are recent enough to fix and you are not re-auditing the same encounters.

Keep the structure stable for at least a quarter. If you change the sample every month, you lose the ability to tell whether a problem is improving or whether you simply looked somewhere else.

What to check: one page, defined answers

The checklist should follow one encounter, or one episode, from start to claim. Each item gets yes, no or not applicable, and each has a one-line written definition of what yes means. "Treatment plan current" is not a definition. "A signed treatment plan with a review date on or after the date of service" is. Your definitions should come from the rules that actually apply to you: payer contracts, state regulations, accreditation standards and your own policies.

Clinical record

  • Required consents present and signed, including any 42 CFR Part 2 consent covering disclosures that were made.
  • Assessment completed and supporting the diagnosis and level of care billed.
  • Treatment plan signed and current on the date of service, with goals the notes can be traced back to.
  • Progress note signed within your policy window, by the person who delivered the service.
  • Note content specific to this patient on this date, not carried forward from a prior session or copied across a group.
  • Start and end times, or units, consistent with what was billed.

Eligibility and claim

  • Clinician credentialed, licensed and enrolled with that payer on the date of service, with supervision documented where required.
  • Authorization in place and covering the date, level of care and units.
  • Attendance record matching the note (present, late, left early).
  • Claim matching the note: service, code, units, place of service, rendering provider.
  • Any edits made to the note after signing handled as dated addenda, not silent changes.

That is around a dozen items. Resist adding more until the program is consistently passing these. If a new risk appears, such as a payer that started requesting a specific element, swap an item rather than growing the list.

Who checks it

The single firm rule: nobody audits their own charts, and nobody audits a chart where they co-signed or supervised the service. In a small program that takes some arranging, but it is doable.

  • Clinical items go to peer reviewers on a rotation, with the clinical director reviewing the targeted charts and reviewing a few of the peer-reviewed charts as a calibration check.
  • Claim items go to someone who reads claims for a living, usually the billing lead. Clinicians often cannot tell whether a note supports the code billed; billers often cannot tell whether a note is clinically adequate. Split the checklist accordingly.
  • The process belongs to the compliance lead, or whoever holds that role. They own the sample, the log, the meeting and the follow-up, but they do not have to review every chart.
  • If the program is very small, consider having an outside reviewer audit the clinical director's own charts periodically, so the most senior clinician is not the one person nobody checks.

Calibrate early. In the first month, have two reviewers score the same three charts independently and compare. Where they disagree, the definition is not clear enough. Fix the definition, not the reviewer.

Sort findings by cause before you sort them by person

A raw list of failed items invites the wrong conversation: who did badly this month. The more useful question is why each finding happened. Tag every finding with one cause.

  • Knowledge. The clinician did not know the requirement. Fix: targeted training, a clearer template prompt, a short written standard.
  • Workflow. The requirement was known but the process made it easy to miss, such as treatment plan reviews nobody was reminded about. Fix: change the process or who owns the step.
  • System. The template, form or software allowed or encouraged the error, such as a note type that defaults to the wrong service. Fix: change the form.
  • Individual. One person, repeatedly, despite knowing and despite a workable process. Fix: supervision, with documentation.

If the same finding shows up across four clinicians, it is not four performance problems. It is one workflow problem with four names on it.

In practice, most recurring findings at small programs turn out to be workflow or system causes. That is good news, because those are the cheapest to fix and they stay fixed.

Turning findings into fixes

Run a thirty-minute meeting each month with the reviewers, the clinical director, the billing lead and the compliance lead. The output is not a score. It is a short findings log with these columns: finding, chart reference (internal only), cause, fix, owner, due date, and the month it will be re-checked.

  1. Correct the record properly. Where documentation can legitimately be completed, it is completed as a dated late entry or addendum by the person who delivered the service, following your policy. Never backdate or rewrite. See late entries, addenda and amendments.
  2. Route claim-affecting findings to compliance the same week. A paid claim the record does not support is no longer just a documentation finding. Compliance decides, with counsel where needed, what has to be corrected, refunded or reported.
  3. Choose one systemic fix per month. Pick the finding with the widest reach and change the workflow or form behind it. A program that makes twelve small process fixes a year is in a very different place than one that wrote twelve summaries.
  4. Re-check what failed. Next month's targeted sample includes the item, and the clinicians, behind last month's findings. That is how you find out whether the fix worked.
  5. Report one page upward. Leadership or the board gets trend by checklist item, open fixes and their owners, and anything routed to compliance. Not the charts.
When an audit finds a paid claim the record does not support

Federal rules require providers to report and return identified overpayments in Medicare and Medicaid, and commercial contracts often carry their own refund terms. For Medicare, see 42 CFR 401.305. Do not set your own deadlines from memory: confirm the obligation and timing with the applicable regulation, payer contract and counsel, and log when the finding was first identified.

A monthly rhythm that survives a busy month

  • Week one: compliance lead pulls the random and targeted sample and assigns reviewers.
  • Week two: reviewers score charts against the checklist and enter findings in the log.
  • Week three: thirty-minute findings meeting; every finding leaves with an owner and date.
  • Week four: corrections, the one systemic fix, and the one-page report.

If a month is genuinely impossible, shrink the random sample but keep the targeted charts and the re-checks. Those are where the fixes are verified. Skipping the meeting is the one thing that turns the process back into a binder.

Software can take some of the manual pulling off the reviewers' plate. In Unisona Health, the documentation clock shows late and unsigned notes live, so timeliness stops being something the audit has to discover, and the activity log shows every edit and signature with a timestamp when a reviewer needs to know whether a note changed after signing. Neither replaces the review itself. The checklist, the meeting and the owner on every finding are what make the audit worth the hours, whatever system you use. For how auditing fits a wider compliance program, see the seven elements, built for behavioral health.

Frequently asked questions

How many charts should a small behavioral health program audit each month?

There is no universal number, so first check whether your accreditor, state rules or payer contracts set one. Otherwise, size the sample to hours you can protect every month, for example a fixed number of charts per clinician split between random and targeted selection, and keep it stable for at least a quarter.

What should an internal chart audit checklist include?

A one-page list that follows an encounter from consent to claim: consents, assessment, current signed treatment plan, timely and individualized progress notes, times or units, credentials and enrollment on the date of service, authorization, attendance, and whether the claim matches the note. Each item needs a written definition of what passing means.

Who should review charts in an internal audit?

Nobody should audit their own charts or services they supervised. Peers on a rotation can review clinical items, the clinical director takes targeted charts and calibration checks, the billing lead checks claim items, and the compliance lead owns the process and follow-up.

What do you do with chart audit findings?

Log each finding with its cause, fix, owner, due date and re-check month. Correct records properly with dated addenda, route anything affecting paid claims to compliance, choose one systemic fix each month, and re-check failed items in the next month's sample.

What happens if an internal audit finds an overpayment?

A paid claim the record does not support should go to compliance the same week. Federal rules require reporting and returning identified Medicare and Medicaid overpayments, and commercial contracts may have their own terms, so confirm obligations and timing with the regulation, payer contract and counsel.

Samuel Jean

Co-Founder at Unisona Health, the behavioral health platform built for the audit. More about us →

Get started

Start today, or take a look first.

Create your workspace in minutes and try the whole platform free for 14 days — no credit card. Or book a one-hour walkthrough first.

Start your free trial