The Defibrillator Claim Medicare Never Reads: Two Rejection Rules and a Ten-Code Gap (E0617, K0606–K0609)

Two ways to lose the claim, and only one of them involves a reviewer

DMEPOS is the worst-performing category in Medicare fee for service by a wide margin. CMS put the 2025 DMEPOS improper payment rate at 24.1%, roughly $2.3 billion projected, against an overall Medicare FFS rate of 6.6%. Almost four times the program average, and documentation failures drive most of it.

The external defibrillator policy is worth reading against that backdrop, because it has an unusual structure. LCD L33690 and Policy Article A52458 are short documents. The clinical criteria are tight and largely objective. And yet two of the three ways this file dies never involve clinical judgment at all.

A52458 spells out both mechanisms in the same section. Suppliers add KX only when every criterion in the LCD coverage section has been met. When they have not been met, the line needs GA if a properly executed ABN is on file, or GZ if it is not. Then the sentence that matters: claim lines billed without a GA, GZ, or KX modifier will be rejected as missing information. Immediately after it, a second and independent rule. All five codes in this policy carry an FDA Class III designation, so E0617, K0606, K0607, K0608 and K0609 must include the KF modifier, and lines billed without KF are rejected as missing information too.

"A rejection never reaches a reviewer. The chart is never read, the coverage criteria are never tested, and the line lands in a work queue that most teams clear once a quarter."

Those are two separate attestation stacks on the same line. A supplier who has drilled KX discipline into the billing team and never heard of KF will pass the first gate and fail the second, on every claim, silently.

The wearable test is one of four. The non-wearable test is two of two.

This is the distinction that most often gets flattened in an intake workflow, because both devices look like external defibrillators and both get filed under the same policy.

K0606, the garment type device with integrated ECG analysis, is covered when the beneficiary meets one of four criteria. A documented episode of ventricular fibrillation or a sustained ventricular tachyarrhythmia lasting 30 seconds or longer. A familial or inherited condition carrying high risk of life-threatening ventricular tachyarrhythmia, such as long QT syndrome or hypertrophic cardiomyopathy. Documented prior myocardial infarction or dilated cardiomyopathy with a measured left ventricular ejection fraction at or below 0.35. Or a previously implanted defibrillator that now requires explantation. Any one of those, and the medical necessity question is answered.

E0617, the non-wearable device, works differently and the difference is expensive. The beneficiary qualifies either by meeting both criterion A and criterion B, or by meeting criterion C on its own. Criterion A is a list of eight qualifying cardiac conditions, some of them quite detailed. Criterion B is one line: implantation surgery is contraindicated. Criterion C is the explantation path.

So for every E0617 file that is not an explantation, a qualifying arrhythmia or cardiomyopathy establishes half the case. The chart also has to say, somewhere a reviewer can find it, that the patient cannot have a device implanted. That sentence is rarely written, because the cardiologist who ordered the external device already knows why and has no reason to narrate it. Nobody asks for it at intake, because the diagnosis looks sufficient on its face.

The qualifying conditions carry their own definitions, and they are stricter than the narrative

L33690 defines its terms in a way that turns three common chart phrases into non-qualifying language.

Myocardial infarction is defined by elevated cardiac enzymes or Q-waves on an electrocardiogram. A discharge summary that says "history of MI" with no enzyme or ECG evidence anywhere in the record has not established the criterion. Ejection fraction must be measured by angiography, radionuclide scanning, or echocardiography, so an EF quoted in a clinic note without an underlying study attached is a number without a source. And the sustained arrhythmia criterion excludes dysrhythmias due to a transient or reversible cause, with the LCD naming drug toxicity, severe hypoxia, acidosis, hypokalemia, hypercalcemia, hyperkalemia, systemic infections and myocarditis as examples, and it excludes events occurring during the first 48 hours of an acute myocardial infarction.

That 48-hour exclusion is the one that catches suppliers on hospital discharge referrals. A patient codes in VF on day one of a STEMI, is resuscitated, and goes home with a wearable device while the team decides about an ICD. Clinically that is textbook. Against criterion I.1 as written, the arrhythmia occurred inside the excluded window, so the file has to qualify on a different criterion, usually the EF path, which means an EF study has to be in the record and it has to read at or below 0.35.

The LCD also opens with an expectation rather than a criterion: it is expected that the treating practitioner be experienced in the management of beneficiaries at risk for sudden cardiac death. Not a checkbox, but a useful thing to notice when a referral arrives from a generalist.

The closed list, and the ten codes that only one device gets

All diagnosis codes for this policy live in A52458, not in the LCD. They were moved there in 2020 under CR 10901 and the LCD has carried no ICD-10 fields since.

The article publishes two groups. Group 1 applies to E0617 and holds 67 codes. Group 2 applies to K0606 through K0609 and holds 77. Then the section that most teams skip entirely, headed "ICD-10-CM Codes that DO NOT Support Medical Necessity," whose Group 1 paragraph reads in full: all ICD-10 codes that are not specified in the previous section.

That single sentence turns the covered list into a closed list. There is no unlisted-but-reasonable path. A cardiologist can be entirely correct about the patient and the claim still has no route to payment if the code submitted is not on the group that matches the HCPCS.

The ten-code difference between the groups is where this becomes a real operational trap. Group 2 is Group 1 plus A18.84 and the dilated and restrictive cardiomyopathy family: I42.0, I42.3, I42.4, I42.5, I42.6, I42.7, I42.8, I42.9 and I43. Those support the wearable device. They do not appear on the E0617 list.

Read that against the coverage criteria and the logic holds. Criterion I.3 for K0606 names dilated cardiomyopathy directly. But criterion II.A.7 for E0617 references nonischemic dilated cardiomyopathy as a qualifying condition, and the codes a coder would naturally reach for to describe it are not in Group 1. A file that qualifies clinically under II.A.7 needs a diagnosis code that survives the Group 1 list, and I42.0 is not one of them.

A52458 also adds the caution that saves nobody from the first trap but matters for the second: the presence of an ICD-10 code listed in this section is not sufficient by itself to assure coverage. The list is necessary. It is not enough.

The 2025 diagnosis expansion is a reason to re-check old crosswalks

Revision 11 to A52458, also effective 09/04/2025, added a substantial block of heart failure codes to both groups: I25.5, the I50.1 through I50.9 family in full, plus Q20.5 and Q21.3. If your coding crosswalk for this policy was built before September 2025, it is rejecting or downgrading files that are now covered. That is the rarer failure, revenue left on the table rather than clawed back, and it never shows up as an error anywhere.

What the policy does not say, which is the part that costs the most

Suppliers reasonably assume the LCD tells them when to stop. It does not.

Neither L33690 nor A52458 contains a rental rule, a capped rental classification, a length of need limit, or any statement about what happens when the patient finally receives an implantable device. The LCD's utilization guidelines section says only: refer to coverage indications, limitations and medical necessity. There is no month count anywhere.

This matters because the wearable device is, in practice, a bridge. Patients wear it for the weeks or months between a qualifying event and a decision about an ICD, or through a waiting period after revascularization, or while a newly diagnosed cardiomyopathy is given time to recover on medical therapy. The clinical endpoint is real and it arrives. The policy just does not name it.

What governs instead is the general documentation article, A55426, and it is less forgiving than it sounds. For rented equipment, continued medical need is established by a recent order, a recent repair order, a recent order change, or timely documentation in the medical record showing usage, where timely means within the preceding 12 months unless a policy says otherwise. Separately, suppliers are responsible for monitoring utilization of rental items, and must discontinue billing Medicare when the item is no longer being used by the beneficiary.

So the obligation to know that the patient got an ICD in March sits with the supplier, and the LCD offers no trigger, no recertification date and no checkpoint to hang it on. Here is how that plays out. A 61-year-old woman with newly diagnosed nonischemic cardiomyopathy and an EF of 0.28 goes home on a wearable device in January. She is on guideline-directed therapy. In April her EF has recovered to 0.42 and the ICD conversation is dropped. She stops wearing the vest in May and tells her cardiologist, who has no reason to call the supplier. Billing continues, because nothing in the billing system knows what a vest is for.

The false comfort of no prior authorization

K0606 sits on the CMS Master List of DMEPOS Items Potentially Subject to Conditions of Payment. Prior authorization: no. Face-to-face encounter and written order prior to delivery: no. Effective dates for both: not applicable.

Teams read that and relax, and it is the wrong conclusion. Prior authorization is a front-loaded review that returns an affirmation you can rely on. Without it, nothing in the workflow forces anyone to look at the file before the device goes out. The review happens later, on a post-payment pull, against a chart that was assembled in a hospital discharge rush eighteen months earlier.

The ordinary rules still apply in full, and A52458 restates them with teeth. A Standard Written Order must be communicated to the supplier before the claim is submitted, or the claim is denied as not reasonable and necessary. For items on the Required List under Final Rule 1713, delivery ahead of a signed WOPD is unrecoverable: if the order is not obtained before delivery, payment will not be made for that item even if the supplier obtains a WOPD afterwards. Proof of delivery must be produced on request or the service is denied. And the one policy-specific documentation requirement in A52458 is a single line that is easy to treat as clerical: the diagnosis code that justifies the need for these items must be included on the claim.

Coding, briefly, because two of these rules end in a denial rather than a rejection

A52458's coding guidelines are short. Non-wearable automatic external defibrillators with integrated ECG capability are E0617. Wearable, automatic external defibrillators with integrated ECG analysis are K0606. Replacement supplies and accessories for K0606 are K0607, K0608 and K0609 as appropriate, while replacement supplies and accessories for E0617 go on A9999. Other types of defibrillator are coded A9270, and no separate payment is made for carrying cases or mounting hardware.

A9270 is the code for a statutorily non-covered item, so routing a device there is a coverage decision rather than a pricing one. When a new product does not map cleanly, the answer is the PDAC contractor, which A52458 names directly.

How a clean file goes bad

What to check before the file goes out

This is practical triage at setup and at each monthly cycle. It is not the full policy, and it does not replace reading L33690, A52458 and A55426 against your own book of business.

Where this actually goes wrong

None of these failures is a cardiology problem. The patients are genuinely at risk of sudden cardiac death, the devices are genuinely indicated, and the treating practitioners are usually electrophysiologists who know exactly what they are doing.

What breaks is the translation layer. A modifier rule expands in a revision note that nobody in billing subscribes to. A coder picks the obvious I42 code without knowing that the obvious code lives on the other group. An intake coordinator reads a discharge summary describing an ICD that cannot be placed and does not realize that the absence of the phrase is the entire second half of the coverage test. A rental keeps running because no field in the system was ever designed to hold the date a vest came off.

Every one of those was knowable before the claim went out. Most of them are machine-checkable against the published text of L33690, A52458 and A55426. Reading the file against the rule before it ships is the whole job, and it is the part DocuFindr automates.