Bill 95249 when the patient owns the device and 95250 when the practice lends one out and takes it back, each once per qualifying event. Add 95251 for the interpretation and report, once a month. All three need at least 72 hours of recording, and 95251 needs no modifier 25.
Everyone treats this as a code-selection problem. It is mostly an omission problem: in our claim audit, 95251 was never billed in months where a CGM download was documented on 23% of professional CGM encounters, and nobody appeals a line that was never submitted. Picking the wrong start-up code costs a practice about seventy-five dollars an encounter. Never billing the interpretation costs a three-physician practice roughly $17,900 a year, and it never shows up on a denial report.
Methodology:Luxen figures on this page come from three datasets: the Luxen claim audit of 61,400 claims audited from January 2025 to June 2026, Luxen billing reviews covering 410 practice billing reviews from January 2025 to June 2026, and Luxen client data across 38 client practices from January 2024 to June 2026. CGM figures are drawn from the endocrinology and primary care claims inside that audit. Code descriptors and parenthetical instructions are taken from the AMA coding handbook, coverage and supply rules from CMS LCD L33822 and policy article A52464, conversion factors from the CY2025 and CY2026 Medicare Physician Fee Schedule final rule fact sheets, and the published payment amounts from the source named beside them.
Whose equipment it was decides between 95249 and 95250. What you did decides whether 95251 belongs on the claim at all. Running those two questions together is the commonest reason a CGM claim is coded wrong.
95249 is the start-up on a device the patient already has: patient-provided equipment, sensor placement, hook-up, calibration, training, and printout of recording. Note what is absent, removal of sensor. The patient walks out still wearing it, because it is theirs.
95250 is the start-up on a device the practice owns and lends out. Its descriptor reads physician or other qualified health care professional (office) provided equipment, and it does include removal of sensor, because the recorder comes back. That word is the cleanest tell in the family.
95251 is neither. It is analysis, interpretation and report, with no equipment element and no training. It is the read, not the capture, and the line most practices never bill.
Steps two and three are independent. Placed your own recorder and read the data: 95250 and 95251. Hooked up a patient-owned sensor and read the data: 95249 and 95251. Only pulled a report off an existing sensor, with no placement encounter: 95251 alone, which is an ordinary claim and not a red flag. Getting this right at the coding stage costs nothing; getting it wrong produces either a refund or a line that was never billed.
Yes, and almost every published guide gets this wrong by omission. The phrase for a minimum of 72 hours sits in the common portion of the descriptor, which all three codes inherit. It is not a 95250 and 95251 rule with 95249 exempted. A start-up that produced 48 hours of usable data is not a billable 95249.
In our claim audit, CGM claims were submitted on recordings shorter than the 72-hour minimum 9% of the time. That is not a denial pattern, because the recording length is not on the claim. It is an audit pattern, which is worse.
95250 and 95251 are each reportable once per month. 95249 is governed by a different rule entirely, and it is the one nobody quotes correctly.
The CPT instruction is not once per lifetime and not once per patient. It is do not report 95249 more than once for the duration that the patient owns the data receiver. The trigger is receiver ownership, not the patient, not the sensor, and not the calendar.
The clock restarts when the receiver changes, and only then. A patient swapping sensors every ten days on the same receiver gets one 95249, ever. A patient who moves to another manufacturer, or takes a new receiver generation, has a new data receiver and a new billable start-up. A warranty replacement of the same model is the grey case, worth a written payer answer before it becomes a pattern.
The error runs both ways. We found 95249 rebilled against the same receiver inside twelve months on 6% of claims, and practices treating 95249 as once per patient forever, never rebilling a legitimate manufacturer switch.
CPT sets a ceiling of once a month. A payer may set a lower one, and some do. Wisconsin ForwardHealth caps 95250 and 95251 at four times per year, maximum once monthly. A practice reading the CPT rule as permission to bill twelve interpretations against that plan will collect four and write off eight.
The monthly cap is the starting point, not the answer; the answer is in the contract. Coding and modifier errors caused 21% of denials across the claims we audit, and frequency assumptions carried from one payer to another are a steady contributor.
No, and this is where the most widely circulated guidance is wrong. Noridian, the Medicare Administrative Contractor for jurisdictions E and F, answered it in a published Q and A on 7 May 2025: asked whether modifier 25 is needed with an E/M billed alongside CPT 95251 on the same day, the contractor said no, since the NCCI does not show any coding combinations, there should be no reason to append modifier 25 on the E/M.
The same answer carries the real constraint, and it is not a modifier problem: CPT 95251 services cannot overlap with the separate E/M visits. The exposure is double-counting, not under-modifying. If the physician's review of the tracing is also driving the medical decision making behind the E/M level, that work has been billed twice, and a modifier fixes nothing.
What survives review is a separate interpretive report with its own findings: time in range, hypoglycemic episodes and their durations, the pattern identified, and what changed in the plan because of it. If the CGM report reads like a paragraph copied out of the visit note, the two services are the same service. 95249 and 95250 are different: the patient must be present, and modifier 25 on a same-day E/M there is a defensible convention rather than a sourced requirement.
99091 and 0446T are clear noes. Everything else is less settled than the billing blogs suggest.
CPT instructs that none of the three codes be reported in conjunction with 99091 or 0446T, and an NCCI edit independently prohibits 99091 with either 95250 or 95251. Wisconsin ForwardHealth states the same from the payer side. That one is not arguable.
The 99457 question is genuinely unresolved, and nobody ranking for this query says so. The AMA parenthetical directs that 95250 and 95251 not be reported with 99457 or 99458. The Endocrine Society coding FAQ states the opposite. Both are credible and they contradict each other, so check the current code book parenthetical before billing the pair.
The claim that CGM and remote patient monitoring device codes 99453 and 99454 cannot be billed in the same period is repeated constantly and we could not source it to CMS. The parentheticals name 99091, 0446T, 99457 and 99458, not 99453 or 99454. A practice conceding those lines on the strength of a blog post is conceding something no regulation appears to require. Our page on billing chronic care management and remote patient monitoring together covers the RPM clock in its own right.
Different claim, different contractor, different biller. This is the confusion that costs practices the most, because the two halves of CGM revenue look like one topic and are not.
95249, 95250 and 95251 are CPT codes on a professional claim, paid under the Physician Fee Schedule. E2103, A4239, E2102, A4238 and E2104 are HCPCS Level II codes billed by the DME supplier or pharmacy to a DME MAC. The ordering practice does not bill the supply codes. It writes the standard written order and lives with the consequences if that order is incomplete.
The definitions come from CMS policy article A52464, revised effective 18 February 2025. E2103 is a non-adjunctive CGM or receiver, one that can be used to make treatment decisions without a stand-alone blood glucose monitor. E2102 is adjunctive, requiring the user verify glucose levels with a BGM before making treatment decisions. A4239 is the supply allowance for a non-adjunctive CGM at one month per unit, and it bundles the home meter and its test strips, lancets, lancing device, calibration solution and batteries. Billing those separately alongside A4239 is how suppliers end up in a recoupment conversation.
Three more rules belong at the desk. No more than ninety days of supply allowance may be billed or dispensed at once. KX marks an insulin-treated beneficiary and KS a non-insulin-treated one, never together. A smartphone as the sole display, with no DME receiver, is non-covered and reports as A9270.
The coverage criteria in LCD L33822 govern this side, not the office codes, and that distinction is missing from every page ranking for CGM billing. A practice that believes L33822 gates 95250 will refuse to bill a payable professional study on a patient who fails the DME criteria. Verifying the DME benefit is separate work from verifying the visit, which is why eligibility and prior authorization checks on this population run twice.
Start with the published amounts, and the caveat that they are already out of date. Dexcom's 2025 coding guide puts the Medicare physician office amounts at $139.41 for 95250, $64.05 for 95249 and $33.32 for 95251.
Those are CY2025 figures at a $32.35 conversion factor. For CY2026, CMS finalized two conversion factors for the first time, $33.57 for qualifying APM participants and $33.40 for everyone else. It also finalized a 2.5% efficiency adjustment to the work relative values of every service that is not time based, which includes all three. The 2026 numbers move up on the conversion factor and back down on the efficiency adjustment, and any figure quoted without a year attached is worthless.
Take a three-physician endocrinology practice with 260 established patients on personal CGM, a documented download and interpretation for a median of nine months a year each, 70 new patients started on their own devices annually, and 90 professional CGM studies on practice-owned recorders.
Coded to the amounts above, that book is worth about $94,999 a year: 2,340 interpretation opportunities at $33.32, 70 start-ups at $64.05, and 90 office studies at $139.41. Now apply the error rate. In our claim audit, 95251 was never billed in months where a CGM download was documented on 23% of professional CGM encounters. That is 538 interpretations a year that happened, were documented, and never reached a claim, worth about $17,926. The practice bills $77,073 instead of $94,999.
The gap holds across practice sizes: $63,333 documented against $51,382 billed at two providers, $94,999 against $77,073 at three, and $189,998 against $154,146 at six.
The leak is durable because it produces no denial: the revenue report reconciles perfectly against the claims that went out. The only way to see it is to count documented CGM downloads against 95251 lines billed, by provider, by month. That beats any denial dashboard here, and it is what a complete revenue cycle view surfaces.
The losses split into two piles, only one of which is a denial.
Across the CGM claims in our audit, 95251 was never billed on 23% of documented encounters, the wrong code was used for whose equipment it was on 11%, a recording under 72 hours was billed on 9%, 95250 was billed twice in one calendar month on 8%, and 95249 was rebilled against an unchanged receiver on 6%.
Only two of those five produce a denial. The frequency errors bounce and are correctable. The rest are silent: the underbilling never appears, and the 72-hour and receiver-ownership errors pay quietly until somebody pulls records.
On the supply side, CMS has reviewed CGM claims formally since Recovery Audit Issue 0189 was approved on 8 September 2020, a complex review at all DME MACs nationwide. Build the checklist around its named failure points: missing or invalid written order elements, missing face-to-face documentation, insufficient medical necessity, and missing proof of delivery. The L33822 face-to-face rule is specific and practices miss it constantly: within six months prior to ordering, the treating practitioner must have an in-person or Medicare-approved telehealth visit evaluating diabetes control, and continued coverage requires a visit every six months documenting adherence. In our billing reviews, 41% of practices had no process to confirm that six-month visit before a CGM order went out.
Which denials are worth working? Frequency denials against a payer whose published cap you can produce are not; the answer there is a calendar. Bundling denials on 95251 alongside an E/M are, with the Noridian answer attached. Appeals are overturned 68% of the time across our client practices, but 19% of denied claims are never reworked or appealed at all, and CGM lines sit disproportionately in that pile because no one owns them. That is what a denials and AR recovery process is for.
In house wins whenever one named person owns the code family and reads it monthly. The decision tree fits on an index card, the frequency rules are three sentences, and the 95251 count report takes an afternoon to build once.
The case for outside help is not the rules. It is that the knowledge sits in one person, and when that person leaves the practice goes back to never billing 95251. Fully loaded in-house billing cost 7.9% of collections for practices under $2M in our reviews, against an outsourced range of 3% to 6%.
What decides it is everything outside these three codes. Eligibility and coverage errors caused 24% of denials across our audit work, and the same front-end discipline lifts every line a practice bills. Across our client practices, first-pass denial rate fell from 14.2% to 6.1% within 90 days, clean claim rate rose from 89.6% to 97.3%, and median days in AR dropped from 54 to 33 within 120 days.
If you are comparing partners, ask one question before price: will they show you documented CGM downloads against 95251 lines billed, monthly, by provider? Most cannot. Our page on how to evaluate a medical billing company has the rest of the list, and full-service medical billing covers a complete engagement inside your existing system. To see your own last ninety days of CGM lines against what your notes documented, a billing review starts there.
Want to know how this applies to your practice? We will review your AR and denials, free, in 30 minutes.
Book the reviewThree questions settle every CGM claim: whose equipment was it, did the sensor come back, and did anybody write an interpretive report. This table answers all three at once.
| Question | 95249 | 95250 | 95251 |
|---|---|---|---|
| Whose equipment | Patient-provided | Office-provided | Either, not relevant |
| Sensor removal in descriptor | No, patient keeps wearing it | Yes, recorder comes back | Not applicable |
| What is included | Placement, hook-up, calibration, training, printout | Placement, hook-up, calibration, training, removal, printout | Analysis, interpretation and report |
| Minimum recording | 72 hours | 72 hours | 72 hours |
| Patient physically present | Required | Required | Not required |
| How often reportable | Once for the duration the patient owns the data receiver | Once per month | Once per month |
| Who may perform | Trained clinical staff under supervision, billed by the practitioner | Trained clinical staff under supervision, billed by the practitioner | Physician, NP or PA only |
| Modifier 25 on a same-day E/M | Conventional, not sourced to a MAC | Conventional, not sourced to a MAC | Not needed per Noridian, May 2025 |
| Barred with 99091 and 0446T | Yes | Yes | Yes |
| Published 2025 Medicare office amount | $64.05 | $139.41 | $33.32 |
One line worth memorizing: the equipment decides between 95249 and 95250, and 95251 is a separate decision that stands on whether an interpretive report exists.
This is the specialty the codes were written for, and the one where the 95251 leak is largest because the volume is largest. A practice carrying 260 CGM patients has over 2,000 interpretation opportunities a year, and every percentage point of unbilled interpretations is roughly $780. The operational fix is a monthly report counting documented downloads against 95251 lines by provider, not more coding education. Practices running professional CGM on loaner recorders should also track recorder inventory against 95250 lines, since a recorder that never comes back is a 95250 that should have been a 95249. The wider code set sits in endocrinology billing.
Primary care starts more CGM patients than endocrinology does and bills the codes least often, usually because the superbill has no line for them. The commonest pattern we see is a nurse placing a patient-owned sensor, training the patient, and the encounter leaving the building as an office visit alone, with no 95249. The second is a physician reviewing a Clarity or LibreView report at a follow-up and folding that work into the E/M rather than billing 95251 beside it, which is the one case where the Noridian overlap warning cuts the other way. A primary care practice gains more from one added superbill field here than from any coding course.
Pharmacist-led CGM is real billable work in some states and invisible in others, and the difference is enrollment rather than clinical scope. North Carolina Medicaid enrolls Clinical Pharmacist Practitioners under taxonomy 1835P0018X to bill 95249 and 95250 directly, at rates set at $44.80 and $122.63 when the policy took effect on 1 February 2022. A pharmacy running a CGM service outside such a pathway is generally billing incident to a supervising practitioner, and the supervision documentation is what decides whether the claim survives.
Suppliers bill E2103 and A4239, never the 952xx codes, and their exposure is documentation rather than code selection. The order must carry the beneficiary name or MBI, order date, item description, quantity, practitioner name or NPI and signature, and proof of delivery has to exist separately. A4239 bundles the home meter and its strips, lancets and batteries, so the codes most likely to trigger a recoupment are the ones a supplier adds alongside it out of habit. DME billing lives or dies on the written order file.
Patients on dialysis with diabetes generate CGM interpretation work that routinely goes unbilled, because the monthly capitated dialysis billing pulls attention away from the professional claim. The interpretation is separately reportable, subject to the same monthly cap, and the practice reading the tracing is the practice that bills it regardless of who manages the dialysis. Where two groups are both reviewing the same data, only one 95251 per month is payable, and agreeing which group bills it beforehand avoids a duplicate denial that neither can appeal.
The encounter-based payment structure changes the question entirely. The professional CGM work is generally wrapped into the encounter rate rather than paid line by line, so the value of coding it correctly is reporting and wrap-around accuracy rather than a separate payment. Clinics that stop capturing the codes because they do not pay separately lose the data that supports rate setting later, and visits billed under the wrong rendering provider cause 8% of RHC and FQHC denials in our audit work regardless of the code on the line.
Yes. The two codes describe different work, the office study with practice-owned equipment and the analysis and interpretive report, and no CPT parenthetical bars the pair. Each carries its own once-per-month limit. What has to exist is a written interpretive report separate from the study record, because a 95251 supported only by a printout attached to the chart is the version that fails on review.
No. The descriptor covers analysis, interpretation and report, with no patient-contact element, and Wisconsin ForwardHealth lists the face-to-face requirement for 95251 as not required. That is separate from the DME coverage rule in LCD L33822, which requires a visit within six months before ordering a personal CGM. The first governs the professional claim, the second governs the supply claim.
Generally yes. The CPT limit is once for the duration that the patient owns the data receiver, so a different manufacturer means a different receiver and a new billable start-up. A new sensor on the same receiver does not qualify. A warranty replacement of the identical model is the grey case, and it is worth a written payer answer before it becomes a recurring pattern on your claims.
It does not need to be. No 952xx code appears on the Medicare telehealth services list, but 95251 is not a telehealth service in the first place. It is a non-face-to-face professional service billed like any other interpretation, with no originating site requirement and no telehealth modifier. 95249 and 95250 cannot be furnished by telehealth at all, because placement, hook-up and removal require the patient present.
Trained clinical staff working within their scope, with the service billed by the supervising practitioner. The AAFP states 95250 may be performed by any qualified staff member under the direct supervision of a physician, PA or NP, while 95251 may be reported only by a physician, NP or PA. No CMS or contractor document states the supervision level for these codes specifically, so confirm the supervision indicator in the fee schedule file before relying on it.
Not usually, because they are different claims to different contractors. A4239 is a DME supply allowance billed by the supplier or pharmacy to a DME MAC, while 95250 is a professional claim paid under the Physician Fee Schedule. A practice that is also an enrolled DME supplier bills them separately, never on one claim. A4239 also bundles the home meter and its strips, lancets and batteries.
A free 30 minute review of your AR ageing and denial reasons. We tell you what is recoverable and what it would take. No deck, no commitment, no fee.
Book a free billing review