Ask a practice how long it takes to bill a visit and you will usually hear a confident "a day or two." Then pull the data. The median might be two days. The 90th percentile is often two weeks, and the tail includes encounters from six months ago that are still not billed. Both numbers are true, and only one of them is the one the practice quotes.
Charge lag is the count of days from the date of service to the date the claim was first submitted to the payer. It is the simplest measure of how well the front half of the revenue cycle works, because every delay in the process (unsigned notes, missing charges, coding queues, hold queues, batch schedules) shows up in it. It is also a direct measure of risk: every day of lag is a day closer to a timely filing limit and a day of cash the practice has not received. November is a good month to measure it, because the fixes take about eight weeks to show and the practice wants January's cash in January.
Key takeaways
- Measure charge lag as days from date of service to first claim transmission, and report the median and the 90th percentile, never the average alone.
- Include encounters that still have no claim; excluding them hides the worst part of the tail.
- Cut the numbers by provider and encounter type. Most of the tail usually belongs to one or two causes.
- The five fixes, in order: same-day note signing with a daily list, electronic charge capture, daily claim submission, a coding queue standard, a hold queue worked to zero.
- Charge lag leads days in AR by about six weeks, so it is the earlier warning.
How to measure it properly
You need two dates for every billable encounter in a period: the date of service and the date the claim was first transmitted to the clearinghouse. Most practice management systems can export both; if yours only stores the "claim created" date, use it and note the difference. Choose a full calendar month that ended at least 45 days ago so that nearly every encounter has had a chance to be billed, and include encounters that still have no claim, assigning them today's date as a placeholder so they are not silently excluded.
The spreadsheet work is short. Export the encounter list with the two dates, add a column that subtracts one from the other, and sort it. The median is the middle value. The 90th percentile is the value at the row that sits 90 percent of the way down the sorted list; in a month with 2,000 encounters, that is row 1,800. Any spreadsheet's percentile function does the same thing. Exclude encounters that should never produce a claim (no-charge nurse visits, cancelled appointments that were accidentally checked in) before you compute, and write down what you excluded so next month's number is comparable.
Then compute three numbers, not one:
| Measure | How to compute | What it tells you |
|---|---|---|
| Median | Sort all encounters by lag; take the middle value | How the process works on a normal day |
| 90th percentile | The lag value below which 90 percent of encounters fall | How bad the tail is; where the risk sits |
| Unbilled count | Encounters in the month with no claim at all | Revenue that has not entered the cycle |
Averages hide the problem. A practice with a 2-day median and a 21-day 90th percentile has an average around 5 days, which sounds fine and describes nobody's experience. Report the median and the 90th percentile side by side and the conversation changes immediately.
Cut the numbers by provider, by location and by encounter type. Office visits, procedures that wait for an operative report and encounters that wait for pathology behave differently and should be judged separately. A surgeon's 90th percentile will be longer than an internist's for legitimate reasons; a surgeon whose office visits also lag has a different problem.
What good looks like
For office visits, a median of one to three days and a 90th percentile under seven days is achievable in most practices where notes are signed the same day. For procedures dependent on an outside report, a 90th percentile of ten to fourteen days is realistic. We are cautious about universal benchmarks; the more useful comparison is your own practice month over month. If the 90th percentile drops from 19 days to 8, that is a real change regardless of what a national survey says. The unbilled count should trend toward zero for any month more than 45 days old, and any encounter still unbilled at 60 days should have a name next to it.
The five fixes, in the order we apply them
1. Same-day note signing, with a daily list
Unsigned notes are the largest single source of lag in electronic practices. Most systems will not generate a charge from an unsigned encounter, and providers rarely see how many they have open. Set a practice standard (signed by end of the next business day is common), run the unsigned encounter report every morning, and send each provider their own list. Providers respond to a list of six notes; they ignore a monthly count of ninety. We covered the reconciliation reports behind this in why completed encounters never become claims.
2. Electronic charge capture tied to the encounter
Paper superbills and separate charge entry add days and lose encounters. Where the EHR can generate charges from the signed note and its orders, use it, and map every commonly performed service to a charge trigger. Where a procedure or a nurse visit has no trigger, add one. The test is simple: can a visit be completed and signed without a charge existing? If yes, the gap is a process, not a person.
3. Daily claim submission, not weekly batches
Some practices still transmit claims once or twice a week because that is how it was set up years ago. A weekly batch adds an average of three days to every claim for no benefit. Set transmission to every business day and make the rejection report the first task of the following morning. If your clearinghouse charges per batch rather than per claim, ask; most do not, and the ones that do will usually change the terms if asked.
4. A coding queue with a turnaround standard
If coders review encounters before billing, the queue needs an owner and a standard, typically 24 to 48 hours for office visits. Sort the queue by date of service so the oldest is worked first, and report the queue age daily alongside the count. A queue of 40 encounters that are all from yesterday is healthy; a queue of 40 with ten from last month is not.
5. A hold queue that is worked to zero every day
Charges held for a missing diagnosis, an inactive insurance, a missing referral or an unenrolled provider are charges with a clock running. Assign each hold reason to a named person: eligibility holds to the front desk lead, coding holds to the coding lead, enrollment holds to credentialing. Report the count and the oldest date every day. A hold queue that nobody owns is where lag goes to become a write-off.
A worked example
A four-provider primary care practice measured charge lag for a completed month: 2,880 billable encounters, median 3 days, 90th percentile 16 days, 41 encounters unbilled. By provider, three physicians had a 90th percentile under 6 days; one had 27 days and 34 of the 41 unbilled encounters. That provider signed notes in weekend batches. The practice set a next-business-day signing standard, started the daily unsigned list, and changed claim transmission from Tuesdays and Fridays to daily.
Two months later the median was 1 day, the 90th percentile was 5 days and the unbilled count for the measured month was 4. Nothing was purchased. The cash effect was visible too: with roughly $95 in average payment per encounter, moving the 90th percentile from 16 days to 5 pulled about eleven days of collections for the slowest tenth of encounters forward, which for 288 encounters a month is around $27,000 arriving a week and a half sooner, every month. The figures are illustrative but the pattern is what we see repeatedly: most of the tail belongs to one or two causes, and they are visible as soon as the number is cut by provider.
Keep it on the wall
Report the median and the 90th percentile monthly, by provider, on the same page as denials and days in accounts receivable. Charge lag is a leading indicator; AR days is a lagging one. When lag rises, AR days will follow in six weeks, so the practice that watches lag has six weeks of warning. Put the provider cut on the page even when it is uncomfortable. In our experience the provider with the long tail usually did not know, and fixes it within a month of seeing it.
Questions we hear
Our system reports "days to bill" already. Is that enough?
Check what it measures. Many built-in reports use the charge entry date rather than the transmission date, and most report an average. If you cannot get the median and the 90th percentile from the built-in report, export the two dates and compute them in a spreadsheet; it takes ten minutes.
Should we measure lag to payment instead?
Measure both, but separately. Days from claim to payment is mostly the payer's behavior and your denial rate. Days from service to claim is entirely inside the practice, which is what makes charge lag the number the practice can act on this month.
Does outsourced billing fix charge lag?
Only the part after the signature. A billing company can transmit daily and work holds, but unsigned notes and missing charge triggers are inside the practice. When Revelrex provides medical billing, the monthly report shows lag by provider precisely so the practice can see which side of the line the delay is on. An RCM audit measures it for a full quarter as a starting point.
What to do this week
- Export date of service and first transmission date for every encounter in a month that ended at least 45 days ago, including unbilled encounters.
- Compute the median, the 90th percentile and the unbilled count; then cut by provider and encounter type.
- Show each provider their own 90th percentile and unbilled list, privately, with the practice standard for note signing.
- Check the claim transmission schedule and change it to daily if it is not already.
- Assign an owner to every hold queue reason and start the daily count and oldest-date report.
