Interventional spine procedures and the Medicare coverage determinations: what the record has to show

1 Two coverage determinations, read as a documentation specification

In Tennessee, Medicare coverage of medial branch blocks, facet radiofrequency ablation and epidural steroid injections is governed by two local coverage determinations issued by Palmetto GBA, the Medicare Administrative Contractor for Jurisdiction JJ: L38765, Facet Joint Interventions for Pain Management, and L38994, Epidural Steroid Injections for Pain Management.^1,2 Each has a billing and coding article that carries the claim rules.^3,4 This paper sets out what the four documents require as of September 2026, the audit history around them, and what the evidence beneath their thresholds shows. It reports coverage requirements, not which patients should receive which procedure.

The determinations are less a verdict on whether the procedures work than a specification of what a record must contain when a decision is made: baseline scores on named scales, the same scale after each block, relief thresholds, and session counts over a rolling 12 months.^1,2 Each is a number or a date that exists only if someone recorded it at the time.

The audit record explains why. Of facet joint injection services allowed by Medicare in 2006, the HHS Office of Inspector General found that 63% did not meet Medicare program requirements, but the leading errors were documentation (38%) and coding (31%); 8% were medical-necessity errors.⁵ Of 120 facet sessions sampled from late 2021, after the current LCD first took effect, 66 did not comply, 18 of them ablations without two documented qualifying blocks.⁶ The thresholds rest largely on consensus and observational series, and the randomized trials are contested. Coverage and evidence are separate questions; a practice is audited on the first.

2 Where the requirements come from, and which apply in Tennessee

Section 1862(a)(1)(A) of the Social Security Act excludes from Medicare payment items and services that “are not reasonable and necessary for the diagnosis or treatment of illness or injury”.⁷ Section 1869(f)(2)(B) defines a local coverage determination as a contractor’s determination, on a contractor-wide basis, of whether an item or service is covered “in accordance with section 1862(a)(1)(A)”.⁸ An LCD is a contractor’s reading of reasonable and necessary for its jurisdiction, neither a statute nor a clinical guideline.

Palmetto issues both LCDs under its Jurisdiction J and M contracts, covering Alabama, Georgia and Tennessee, and North Carolina, South Carolina, Virginia and West Virginia; the Tennessee contracts are 10311 for Part A and 10312 for Part B.^1,2 The facet LCD took effect for services on or after April 25, 2021, and its current revision applies from April 9, 2026. The ESI LCD took effect for services on or after December 5, 2021, and its current revision applies from April 16, 2026. The facet article, A58350, is at revision 9, effective May 21, 2026; the ESI article, A58695, is at revision 4, effective November 16, 2023.^3,4 Neither had a proposed revision in comment when this paper was reviewed. The numbers are contractor-specific even where LCDs share a title: L39240, for example, is Noridian’s Epidural Steroid Injections for Pain Management, not Palmetto’s.⁹

Medicare Advantage organizations must comply with the written coverage decisions of the local Medicare contractor with jurisdiction for the plan’s area, and may add internal coverage criteria only where coverage criteria are not fully established.¹⁰ Facet joint interventions have also required prior authorization as a hospital outpatient department service since July 1, 2023, under 42 C.F.R. § 419.83,¹¹ and were approved as a Recovery Audit topic for complex review of outpatient hospital claims on June 12, 2018, a listing CMS still posts.¹²

3 The facet joint LCD: indications, two blocks and a counter

Every facet joint intervention must meet general criteria that include moderate-to-severe chronic neck or low back pain, predominantly axial, “that causes functional deficit measured on pain or disability scale”; pain for at least 3 months with documented failure to respond to non-invasive conservative management, as tolerated; no untreated radiculopathy or neurogenic claudication, except radiculopathy caused by a facet joint synovial cyst; and no non-facet pathology on clinical assessment or imaging that could explain the pain.¹ The LCD names acceptable scales, including numerical rating and visual analog scales for pain and the Oswestry, Roland-Morris and PROMIS instruments for function, and requires that the scale used be documented in the record.

Diagnosis is by block, and the block is judged by a number

Diagnosis is by medial branch block; intra-articular blocks count as diagnostic only where medial branch blocks cannot be performed because of specific documented anatomic restrictions. After the first diagnostic procedure there must be “a consistent positive response of at least 80% relief of primary (index) pain”, with the duration of relief consistent with the agent used, before a second, confirmatory procedure at the same level. The second may only be performed at least 2 weeks after the first; exceptions “must be clearly documented in the medical record.” Pain is to be documented at baseline and after each diagnostic procedure “using the same pain scale for each assessment”, with a disability scale obtained at baseline.¹

Initial thermal radiofrequency ablation requires at least 2 diagnostic medial branch blocks, “with each 1 providing a consistent minimum of 80% sustained relief of primary (index) pain”. A repeat ablation at the same site requires at least 50% improvement in pain for at least 6 months, or 50% improvement in the ability to perform previously painful movements and activities of daily living against baseline on the same scale. Therapeutic facet joint injections are covered only after two qualifying diagnostic procedures; later injections at the same site need at least 50% relief for at least 3 months or a 50% gain in previously painful movements and activities, and the record must document why the patient is not a candidate for ablation, such as established spinal pseudarthrosis or an implanted electrical device.¹

Per spinal region, no more than 4 diagnostic, 4 therapeutic and 2 radiofrequency sessions are reimbursed per rolling 12 months. A session is limited to one region and one or two levels, unilateral or bilateral; three- and four-level procedures are non-covered. Interventions without CT or fluoroscopic guidance are not reasonable and necessary, and diagnostic blocks at the level of a previously successful ablation are not covered.¹ The article adds the claim rules: diagnostic injection lines carry the KX modifier, which “in most cases” will be used only for the 2 initial diagnostic injections and whose aberrant use may trigger focused medical review, and bilateral procedures are one unit per level with modifier 50.³

4 The epidural steroid injection LCD: four weeks, a baseline and 50% for three months

ESIs are covered for radiculopathy, radicular pain or neurogenic claudication from listed structural causes, for post-laminectomy syndrome and for acute herpes zoster pain, where history, examination and concordant imaging support the diagnosis. The condition must be “severe enough to greatly impact quality of life or function”, and “an objective pain scale or functional assessment must be performed at baseline (prior to interventions)”, with the same scale repeated at each follow-up. Pain must have lasted at least 4 weeks, with inability to tolerate, or documented failure of, 4 weeks of noninvasive conservative care; acute herpes zoster refractory to conservative management is exempt from the wait.²

A repeat ESI requires that the record document at least 50% improvement in pain for at least 3 months, or 50% improvement in the ability to perform previously painful movements and activities of daily living. If the first injection fails, a repeat after 14 days using a different approach, level or medication is allowed, with its rationale documented. Coverage is capped at 4 sessions per spinal region in a rolling 12 months, and a predetermined series is not reasonable and necessary. A session is limited to one region, transforaminal injections to 2 nerve root levels, and caudal or interlaminar injections to one level and not bilaterally. Guidance must be CT or fluoroscopy with contrast, unless a documented contrast allergy or pregnancy permits ultrasound without it.²

The article specifies the procedure note. For selective nerve root blocks and transforaminal injections the procedural report should document the indications, the medical necessity and the baseline pain score; for selective nerve root blocks, also the percent relief immediately after the injection. “Films that adequately document (minimum of 2 views) final needle position and contrast flow should be retained and made available upon request.” A diagnostic selective nerve root block, coded identically to an epidural injection, takes the KX modifier.⁴

5 The audit record is mostly about paper

In 2008 the Inspector General reported that Medicare Part B payments for facet joint injections rose from $141 million in 2003 to $307 million in 2006, while claims rose 76%.⁵ In a stratified random sample of 646 services from 2006, reviewed by physicians experienced in facet injections and a certified coder, 63% did not meet Medicare program requirements, approximately $96 million in improper payments. Services in an office had an error 71% of the time, against 51% in a facility.

The 63% is easy to read as a measure of overuse. The report’s own breakdown does not support that reading. Thirty-eight percent of services had a documentation error, 27% being undocumented and 11% insufficiently documented; 31% had a coding error, and 61% of the miscoded services that were overpaid had add-on codes billed for bilateral injections instead of modifier 50; 8% had a medical-necessity error; and 14% had overlapping errors.⁵ The finding is principally about records and claims.

Three later audits tested only whether frequency limits were enforced. For January 2017 to May 2019, the Inspector General found $748,555 in improper payments for 3,249 facet injection sessions beyond the five-session limit that 11 of 12 jurisdictions then applied.¹³ For January 2019 to August 2020 it found $9.5 million paid for denervation sessions beyond session and facet-joint limits.¹⁴ For 2019 and 2020, of $52.8 million paid for 303,408 ESI sessions, it found $3.6 million improperly paid for 20,838 sessions, given to 15,362 beneficiaries, beyond the limits then in force; afterwards all 12 jurisdictions moved to 4 ESI sessions per spinal region in 12 months.¹⁵ Each report found oversight inadequate to prevent or detect the payments.^13–15

An audit under the 2021 LCDs

An audit of facet sessions under the 2021 LCDs sampled 120 of 218,421 sessions from August through October 2021. Fifty-four complied and 66 did not, for $18,084 in sampled improper payments, projected to an estimated $29.6 million.⁶ The itemized deficiencies read as a list of the LCD’s fields: 18 ablations without two diagnostic sessions each providing 80% sustained relief; 9 without moderate-to-severe axial pain causing a functional deficit measured on a scale; 5 without 3 months of pain and documented failure of conservative management; 4 second diagnostic blocks within 2 weeks of the first with no documented exception; 7 therapeutic injections with no documentation of why denervation was not an option; and 10 sessions billed for fewer levels, and 6 for more, than the record showed. Separately, 33 of 43 sessions billed as therapeutic should have been billed as diagnostic, an error without payment effect because the two are reimbursed identically. An audit of sacroiliac joint injections, under a different LCD, found 72 of 100 sampled 2023–2024 sessions noncompliant.¹⁶

6 The thresholds rest on consensus; the trials are contested

The facet LCD’s rationale leans on society guidance. It states that most studies have used a relief cut-off greater than 80% to call a medial branch block positive, and cites guidelines advocating dual blocks because of the false-positive rate of a single block.¹ A 2020 multispecialty, international consensus guideline on lumbar facet interventions recommends a 3-month trial of conservative treatment first, graded C on a low level of evidence, and concludes that medial branch ablation may benefit well-selected individuals.¹⁷ It did not settle the two parameters the LCD fixes most firmly: one society dissented on exactly two questions, the number of blocks and the cut-off for a positive block before ablation.

A randomized comparison of block paradigms in 151 patients shows the trade.¹⁸ Among patients who went on to ablation, success at 3 months was 33% with no block, 39% after one block and 64% after two comparative blocks. Because blocks screened patients out, success across everyone randomized was 33%, 16% and 22%, and the cost per successful treatment $6,286, $17,142 and $15,241. Success was defined as at least 50% relief with a positive global effect lasting 3 months. Dual blocks raise the yield of ablation among patients who pass them and lower the number who reach it; the LCD has chosen yield. The direct evidence reviewed here for the 80% dual-block paradigm itself is observational: a cross-sectional cohort of 85 patients selected that way, surveyed by telephone, found 63.2% (95% CI 41% to 85%) reporting at least 50% pain reduction at 6 to 12 months and 44.1% (95% CI 27% to 61%) beyond 24 months, with no comparison group.¹⁹

The MINT trials point the other way, with a selection caveat. Three pragmatic, nonblinded trials in 16 Dutch pain clinics randomized 681 patients to ablation plus a standardized exercise program or exercise alone.²⁰ In the facet trial, 251 patients, the 3-month difference in pain on a 0–10 scale was −0.18 (95% CI −0.76 to 0.40), against a prespecified minimal clinically important difference of 2 points. Participants needed one positive diagnostic block, not two at 80%, and the comparator was not a sham, so the trial did not test the population the LCD selects.

For ESIs the randomized evidence is larger. The double-blind LESS trial assigned 400 patients with lumbar spinal stenosis to glucocorticoid plus lidocaine or lidocaine alone; at 6 weeks the adjusted difference was −1.0 points on the Roland-Morris scale (95% CI −2.1 to 0.1; P = .07) and −0.2 in leg pain (95% CI −0.8 to 0.4).²¹ A 2015 meta-analysis of placebo-controlled trials found, for radiculopathy, a greater immediate-term reduction in pain (weighted mean difference −7.55 on a 0–100 scale, 95% CI −11.4 to −3.74) and a lower short-term risk of surgery (relative risk 0.62, 95% CI 0.41 to 0.92), but effects below its predefined minimal clinically important differences, no longer-term benefit and no clear effect in stenosis.²² Neurogenic claudication remains a covered indication under L38994.² At the far end, a 2025 guideline in The BMJ made strong recommendations against joint radiofrequency ablation, joint-targeted injections and epidural injections for chronic axial spine pain, and against epidural injections for chronic radicular pain, finding “no high certainty evidence of important pain relief” for any intervention it assessed.²³ Coverage has not followed it.

7 Image guidance, contrast and sedation

Guidance and contrast requirements have a safety history. In April 2014 the FDA required a Warning on injectable corticosteroid labels describing rare but serious neurologic events after epidural injection, including death, spinal cord infarction, paraplegia, cortical blindness and stroke, and noted that corticosteroids are not FDA-approved for epidural injection.²⁴ A multidisciplinary working group of stakeholder societies published 17 clinical considerations in 2015 aimed at preventing such injuries,²⁵ and the ESI LCD cites a 2015 workgroup recommendation that cervical and lumbar interlaminar injections be performed with image guidance and a test dose of contrast.²

Both LCDs restrict sedation. The ESI LCD calls moderate or deep sedation, general anesthesia and monitored anesthesia care “usually unnecessary or rarely indicated” and therefore not reasonable and necessary.² The facet LCD excludes them for intra-articular injections, medial branch blocks and cyst aspiration or rupture and excludes routine use for ablation, but will consider moderate sedation for ablation or cyst procedures in individual cases with documented medical necessity, such as a long-standing inability to cooperate or to remain motionless. Its article allows “moderate anesthesia or MAC” for ablation and cyst procedures if medical necessity is clearly established and explained in the record, and warns that frequent reporting may trigger focused review.^1,3 The 2020 consensus guideline adds a diagnostic reason: sedation should not be routine for diagnostic blocks, and patients who receive it should be told of the increased risk of a false-positive block.¹⁷ A Recovery Audit topic dated July 15, 2026, posted among CMS’s proposed topics but with its type given as approved, states that anesthesia claims under code 01992 will be denied when reported on the same date of service, for the same beneficiary, as facet injections, facet denervation or ESIs.²⁶

8 What a procedure record has to carry

The requirements translate into properties any record of these procedures would need, not features of a product.

1. Baselines fixed before the first procedure. A named pain scale and a named disability scale, with values and dates, recorded before any intervention.^1,2
2. The same scale after every diagnostic block, against the index pain, with its time course. Relief must be at least 80% of the primary pain for a duration consistent with the agent used, so a percentage without the agent and the timing is incomplete.¹
3. Session counters by region and category over a rolling 12 months. Three audits found claims beyond the limits paid without detection, so a practice cannot rely on a denial to tell it a limit has been reached.^13–15
4. Claim lines that match the note. Region, levels, laterality, the diagnostic or therapeutic classification, KX and modifier 50 have to agree with the procedure report;³ mismatches of levels, laterality and classification appeared in the most recent sample.⁶
5. Reasons recorded where the LCD asks for one. Why ablation is not an option, why sedation was necessary, why a second block fell inside 2 weeks, why a second ESI followed a failed first. An exception without its documented reason does not meet the criterion that allows it.^1–3
6. Images retained. For ESIs, at least 2 views of final needle position and contrast flow, retrievable on request.⁴
7. Relief figures that originate with the patient. The LCDs ask for a response the patient experienced, on a scale the patient completed. A relief percentage that is estimated, carried forward or pre-filled supplies the number an auditor tests without the observation behind it.

9 What survives

Four statements are supported. In Tennessee the governing documents are Palmetto GBA’s L38765 and L38994 with articles A58350 and A58695, and neither LCD has a revision in comment.^1–4 Their requirements are mostly requirements of the record: named scales, repeated measurement, thresholds, rolling counts and written reasons. The audit history locates failures mostly in what records document and claims report: in 2006 documentation and coding errors far outnumbered medical-necessity errors, 8% of services; three audits found claims beyond the session limits paid; and the gaps found under the 2021 LCDs map onto what they ask for.^5,6,13–15 And the thresholds are consensus-derived; the largest randomized trials did not enroll patients selected by them and found small or no effects, and a 2025 guideline recommends against the procedures for chronic spine pain.^17,20–23

Different things would settle the two questions. The evidence question needs a sham-controlled trial enrolling patients selected by the LCD’s own paradigm, two blocks at 80%, with outcomes past a year. The documentation question needs nothing new: a record in which every number the LCD asks for exists, was entered when it was observed, and agrees with the claim.