OIG: Medicare improperly paid $15.2M for sacroiliac joint injections; 72 of 100 sessions missed the LCD

Sacroiliac joint injections just became the newest line item HHS-OIG says Medicare paid for without the documentation to back it. On August 5, 2026, OIG posted audit OAS-25-09-021: of 100 sampled sacroiliac joint injection sessions, 72 did not comply with Medicare coverage requirements, and on that basis OIG estimated Medicare improperly paid physicians $15.2 million for these injections in FY2024. The finding is about documentation and coverage rules, not fraud, and it lands squarely on any practice that bills interventional pain.

What OIG looked at

The audit covered Medicare Part B payments for 186,842 sacroiliac joint injection sessions with dates of service from October 1, 2023 through September 30, 2024, for which Medicare paid about $22 million. OIG pulled a random sample of 100 sessions and reviewed each physician's supporting documentation against two things: the Medicare requirements in the applicable Local Coverage Determination (LCD), and the billing guidance in the matching LCD Reference Article.

Two codes carry these injections. CPT 27096 is for an injection made directly into the sacroiliac joint. CPT 64451 is for an injection targeting the nerves that innervate the joint. OIG did not perform a medical-necessity review here, and it did not allege fraud. Its determinations were limited to whether the record met the LCD coverage and billing rules.

Where the documentation fell short

Of the 100 sampled sessions, 28 complied with the coverage requirements and 72 did not. Because 41 of those sessions had more than one deficiency, the categories below add up to more than 72.

The indications-of-pain misses were the documentation basics: no record of at least three positive provocative maneuvers (26 sessions), pain that was not low back pain without radiculopathy (21), and no minimum four weeks of conservative therapy before the injection (18). Imaging and pain scores fared no better. In 31 sessions the pain level was never captured at the end of the visit; in 17, fluoroscopy ran without contrast (and ultrasound without a documented contrast allergy); in 16, the starting pain level was missing.

The diagnostic-versus-therapeutic trap

The billing-guidance finding is the one every biller should read twice. The LCDs limit coverage to a maximum of two diagnostic sessions and then a maximum of four therapeutic sessions per enrollee in a rolling 12-month period. The LCD Reference Articles say that when a diagnostic injection is administered, the physician appends modifier KX to distinguish it from a therapeutic one, and does not use KX on therapeutic lines.

Of the 100 sampled sessions, 25 were billed as therapeutic when they should have been billed as diagnostic with KX, and OIG estimated 46,711 of the 186,842 sessions were misclassified this way. Billing therapeutic instead of diagnostic does not change what the physician is paid. What it changes is the patient's math: a diagnostic session mislabeled as therapeutic burns one of the four therapeutic slots, so the patient can run out of covered, medically necessary injections earlier than they should.

The thresholds OIG measured against

If you code or audit these procedures, these are the LCD bars the sampled charts had to clear, and often did not:

• A diagnostic injection has to produce at least 75 percent relief of primary pain before a therapeutic injection is reasonable and necessary.
• A repeat therapeutic injection is not reasonable and necessary unless the prior therapeutic injection gave at least 50 percent consistent relief or functional improvement for at least three months.
• The injection has to be performed under CT or fluoroscopic guidance with contrast, unless a contrast allergy is documented, in which case ultrasound without contrast may be considered.
• Other musculoskeletal injections in the lumbosacral spine are not billed in the same session as the sacroiliac joint injection during the efficacy-assessment window.

Two MACs had no LCD at all

Only five of the seven Medicare Administrative Contractors had issued an LCD and a companion LCD Reference Article for sacroiliac joint injections during the audit period: CGS Administrators, National Government Services, Noridian Healthcare Solutions, Palmetto GBA, and Wisconsin Physicians Service. The other two MACs paid physicians $12.3 million for these injections, roughly 36 percent of the national total, with no LCD framing the coverage rules at all. OIG flagged that money as a further risk area it did not audit. If your jurisdiction is one of the two, the looser payment edits do not lower the bar the record has to clear on review. Same documentation standard, just fewer front-end checks catching a gap before it pays.

What OIG told CMS

OIG made three recommendations to CMS: work with the MACs to build education specific to the sacroiliac joint injection requirements, develop guidance to prevent these improper payments, and develop solutions to stop the incorrect billing of diagnostic injections as therapeutic. CMS concurred with the first two and did not concur with the third.

The claim lines here are from FY2024, so this is not a fresh recoupment notice. What is current is the recommendation: OIG asked CMS and the MACs to build education and guidance on these exact sacroiliac joint injection requirements, and CMS agreed to two of the three. That is the cue to self-audit now, the same response that answers the Novitas nursing-home Part B audit and the WPS billed-charges review: pull your own sacroiliac joint injection charts, confirm the pain indications, imaging, and pain scores are in the note, and check that every diagnostic session carries the KX modifier before the next request for records arrives. It is the same insufficient-documentation failure mode wearing a different CPT code.