Every January, practices make promises about their revenue cycle. The rejection report will be worked daily. Notes will be signed within 48 hours. Patient balances will be statemented on a 30-day cycle. By May, nobody has checked whether any of that happened, and the people who made the promises are busy with something else.
This is why we push the practices we work with to run a mid-year leakage review in May, not in December. A May review covers a window (January through March) that is old enough for most claims to be adjudicated, and it leaves seven months to fix what it finds. A December review produces a report that nobody acts on until the following spring.
Here is how we run it. It takes one person about two working days if the reports are easy to pull, longer if they are not, and that difficulty is itself a finding.
Key takeaways
- Use one window, January 1 to March 31, 2026, for every report. Mixed windows produce confident, wrong conclusions.
- Six reports cover almost all leakage: arrived visits, signed encounters, accepted claims, denials by reason code, payments against expected allowed, and patient AR by last statement date.
- Every finding gets a dollar figure and an owner, even a rough one. Findings without both are ignored.
- Check the January promises against data, not memory. Usually one of three held.
- Carry three numbers into the second half and update them monthly; they take twenty minutes.
Fix the window first
Use dates of service from January 1 to March 31, 2026. Do not use April: commercial payers commonly take 14 to 30 days to adjudicate, and denials often arrive later, so an April claim may still be moving. Pull every report for the same window. The single most common mistake in these reviews is comparing a schedule report for one quarter with a claims report for another and drawing confident, wrong conclusions.
Two more rules about the window. Run the reports by date of service, not by posting date or submission date; a claim from March that was submitted in April belongs to the review, and a claim from December that posted in January does not. And freeze the reports on the day you pull them. AR reports change every hour, and a finding that cannot be reproduced is a finding that will be argued about instead of fixed.
The six reports
| Report | What it shows | The leak it finds |
|---|---|---|
| Arrived appointments by provider | Visits that actually happened | Baseline for everything else |
| Signed encounters by provider | Visits with a closed note | Unsigned or missing documentation |
| Claims with payer acceptance (277CA accepted or an ERA) | Charges that reached a payer | Held charges and unworked rejections |
| Denials by CARC code and dollar | What payers refused and why | Preventable denials, unappealed write-offs |
| Payments vs. expected allowed by payer | Contract variance | Underpayments posted as paid in full |
| Patient AR by last statement date | Balances nobody asked for | Unstatemented and stalled patient balances |
If your system cannot produce one of these, write that down. A practice that cannot see its own denial dollars by reason code is not managing denials; it is experiencing them. The reports that are hardest to pull in our experience are the third (most systems can show claim status but not a clean count of encounters that never became an accepted claim) and the fifth (which needs an expected-payment table loaded, and most practices have not loaded one). If the third is missing, build it by hand: export signed encounters and accepted claims to a spreadsheet and match on encounter number. It is tedious and it is where the money is.
Sizing each finding
A finding without a dollar figure gets ignored. Attach one to each line, even a rough one. For unbilled encounters, multiply the count by your average charge per visit. For rejections never resubmitted, add up the billed amounts. For denials written off without an appeal, sum the write-off adjustments by reason code. For underpayments, take a 50-line sample from your two largest payers, compare the allowed amount to the contracted rate for each CPT code, and apply the variance rate to the quarter's volume for that payer.
A worked example from the kind of practice we see most often: a four-provider family medicine group with about 5,200 arrived visits in the quarter. Signed encounters came to 5,146, so 54 visits had no closed note. Accepted claims came to 5,071, so another 75 signed encounters never reached a payer. At an average charge of $165, that is roughly $21,000 in charges that never left the building. The denial report showed $48,000 in denials, of which $9,400 had been adjusted off with a generic write-off code and no appeal on file. The remittance sample found the largest commercial payer applying a multiple-procedure reduction to CPT 96372 (therapeutic injection) when billed with a 99213, which the contract did not allow. Small on any one claim, about $14 each, but it happened 312 times in the quarter.
Add the lines up and the quarter's visible leakage for that practice was about $35,000, before counting the patient balances. Annualized, that is well over $100,000 for a practice whose owners believed the revenue cycle was fine because cash was roughly where it was the year before. Cash being flat is not evidence that nothing is leaking; it is evidence that volume grew enough to cover the leak.
Give every finding an owner
The second reason findings get ignored is that nobody is named. Put a column on the findings sheet for the person who will fix the process, not the person who will work the backlog. Unsigned notes belong to the medical director, not the billing manager. Unworked rejections belong to whoever runs the clearinghouse queue. Underpayments belong to the person who holds the contracts. Patient statements belong to the front office manager if statements are triggered there, or to billing if they are not. When the same name appears against four findings, that is a staffing finding, and it goes on the sheet too.
What January's promises look like in the data
This is the part of the review that matters most, and the part everyone skips. Take each process change the practice made at the start of the year and find the number that would prove it worked.
- "Notes signed within 48 hours." Pull encounter date versus signature date for the quarter. Report the median and the 90th percentile by provider. A good median with a 90th percentile of 11 days means one provider is carrying the problem.
- "Rejections worked daily." Pull the clearinghouse rejection report and the resubmission date for each line. Median days from rejection to resubmission should be two or fewer. If it is nine, the report is being worked weekly, whatever anyone says.
- "Statements every 30 days." Pull patient AR with the last statement date. Count balances over $25 with no statement in 45 days. That count should be near zero.
Honestly, most practices find that one of the three promises held and two did not. That is normal. What is not normal is finding out in December.
Timely filing: the clock is already running
By May, January dates of service are 120 days old. Many commercial contracts allow 90 to 180 days from the date of service; some Medicaid managed care plans allow as little as 90 days for corrected claims. Sort every unbilled encounter and every unresubmitted rejection from the review by date of service and work the oldest first. Anything from January at a 90-day payer is probably lost already, and the right move is to record the write-off with a reason, not to let it age silently into next year.
Medicare gives you 12 months, which is why practices with heavy Medicare volume feel leakage later and less sharply. Do not let that comfort you about the commercial book. Build a one-page grid of your top ten payers and their filing limits, and keep it next to the hold queue; the person releasing held claims should be able to see which ones expire this week.
Three numbers to carry into the second half
When the review is done, put three numbers on the wall and check them monthly: unbilled encounters older than 10 days (target: single digits), denial rate on first submission by payer (watch for anything above 8 to 10 percent), and net collection rate against expected allowed (anything under 95 percent means either underpayments or bad write-offs). These three catch most of what the six reports catch, and they take twenty minutes to update.
If you would rather have someone outside the practice run the review, our RCM audit uses the same six reports over the same kind of 90-day window and returns findings by owner and dollar. Rates are on the pricing page.
Questions we hear
Our first quarter always looks bad because deductibles reset. Should we skip patient AR?
No. Deductible resets change who owes the money, not whether it was asked for. The review checks whether balances were statemented, not whether they were paid. A high patient balance with a recent statement is a collections question; a high balance with no statement is leakage.
We changed billing companies in February. Is the review still useful?
More useful, not less. Hand-offs are where encounters disappear. Run the January to March window and pay particular attention to encounters from the last two weeks of the old vendor and the first two weeks of the new one.
How often should this be repeated?
Twice a year for the full six-report review, monthly for the three wall numbers. Practices that do the full review quarterly usually stop after two rounds because the findings shrink to a page.
What to do this month
- Pick the puller and the two days. Block them on the calendar before the summer schedule fills.
- Pull the six reports for January 1 to March 31, 2026, by date of service, and save them with the pull date.
- Reconcile arrived visits to signed encounters to accepted claims, and list every encounter that dropped out.
- Size each finding in dollars and name an owner for the process fix.
- Sort the unbilled and unresubmitted items by payer filing limit and release the oldest first.
- Post the three wall numbers and put a monthly reminder on the practice manager's calendar.
