Chapter 22: Medical Necessity: NCDs, LCDs, ABNs, and Proving the Service Was Needed

"The physician decided the patient needed it. That is a different question from whether the plan agreed to pay for it, and confusing the two is the most expensive misunderstanding in this book." — constructed

Overview

Chapter 20 §20.10 listed four reasons a valid code does not pay. Three of them are settled before the claim goes out. This chapter is the fourth — the only one where the medical record can change the answer.

Chapter 19's Case Study 1 opened a category the book had not used before: a service that should not have happened, as distinct from a service described wrongly. Chapter 20's Case Study 2 supplied the sharpest available statement of what governs it:

Clinical correctness and payability are different questions. In most of this book they run together closely enough that people forget they are different. In DMEPOS they came apart completely — an item can be genuinely needed, correctly fitted, and entirely unpayable.

This chapter is about the gap between those two questions, what fills it, and what a coder can and cannot do about it.

It also closes two of the book's open questions. Since Chapter 1 you have been carrying:

• Q5 — what is the knee actually wrong with?
• Q3 — should an ABN have been obtained for the March 14 injection?

Both are answered here, and the second one is answered no, for a reason that surprises most practices.

In this chapter, you will learn to:

• Say why medical necessity is a coverage word
• Trace "reasonable and necessary" to its source
• Tell an NCD from an LCD from a billing and coding article
• Find the policy governing your claim, in order
• Link a diagnosis to a service, and see when the link fails
• Apply a frequency limitation against a calendar
• Know when an ABN is required and when giving one is improper
• Apply GA, GX, GY, and GZ
• Tell a claim that was coded wrong from one that was never going to be covered

22.1 Medical necessity is a coverage word, not a clinical one

Start by separating two things that share a name.

A clinician's judgment that a patient needs a service is a CLINICAL determination. It is made under a professional standard of care, it is the physician's to make, and nothing in this chapter questions it.

A payer's determination that it will pay for that service is a COVERAGE determination. It is made under a contract or a statute, against published criteria, and it is a different question with a different answer.

"Medical necessity," as the phrase is used in billing, means the second one.

Three consequences, and every one of them is counterintuitive the first time.

A service can be necessary and not covered. Chapter 20's braces. A treatment that is the standard of care and is excluded from a benefit package. A service supported by evidence that the plan's policy has not adopted.

A service can be covered and not necessary for this patient. An annual screening in a patient who had one last month; an imaging study before the required conservative therapy.

And the physician's opinion is not the criterion. This is the sentence that causes the most resistance, and it is worth saying plainly: a physician's belief that a service was necessary is relevant evidence and is not the standard. The standard is the policy's criteria, applied to the record.

⚠️ Where Claims Die

"But the doctor ordered it" is the most common and least effective thing said in a medical necessity appeal.

It appears in appeal letters constantly, usually in the first paragraph, and it does nothing — because the payer already knows the doctor ordered it. That is why there is a claim.

What the payer does not know is whether the record contains the specific facts the policy requires, and that is the only thing an appeal can supply.

The reframe that makes appeals work: stop arguing that the service was needed. Show that the record contains what the policy asks for — the covered indication, the documented duration, the failed conservative therapy, the specific finding. Chapter 30 builds the letter and this is its foundation.

22.2 "Reasonable and necessary" and where that phrase comes from

The phrase is statutory, and knowing that changes how you read everything built on top of it.

The Social Security Act, at § 1862(a)(1)(A), excludes from Medicare payment items and services that are not "reasonable and necessary for the diagnosis or treatment of illness or injury or to improve the functioning of a malformed body member."

Read the exclusion carefully, because its structure is instructive.

It is written as an exclusion, not as a grant. Medicare does not list what it pays for; it pays for covered benefits and then excludes what is not reasonable and necessary. That is why coverage determinations mostly describe when a service is covered rather than whether it exists.

"For the diagnosis or treatment of illness or injury" is doing work. A service that is neither diagnostic nor therapeutic is outside the phrase entirely — which is the statutory basis for excluding much of what patients think of as preventive care, and the reason each preventive benefit Medicare does cover required its own statutory addition. Chapter 34 is that story.

And the statute does not define "reasonable and necessary." The content comes from coverage determinations — §22.3 and §22.4 — which is why those documents exist and why they are the operative authority rather than the statute itself.

Commercial plans use their own language, in their contracts and medical policies, and it frequently resembles this one without being identical. The mechanism is the same: a general standard, filled in by specific published policies.

🎓 Exam Watch

Medical necessity is heavily tested and the items cluster into five shapes.

"Which document contains the diagnosis codes that support a service?" — the billing and coding article, not the LCD and not the NCD. Candidates reliably answer LCD.

"An NCD and an LCD conflict." — the NCD governs. An LCD cannot contradict one and generally does not exist where one applies.

"A service is statutorily excluded. Is an ABN required?" — No. The ABN exists for services that would be covered but for the reasonable-and-necessary determination. GY, and GX if a voluntary notice was given.

"An ABN was signed at checkout with 'Medicare may not pay' as the reason." — defective. Not in advance, and no specific reason. GA is not supportable.

"Once per calendar year versus once every 12 months." — different rules, and the exam will give you dates chosen to separate them.

And the framing question that shows up as a short answer: is this claim coded wrong, or was it never going to be covered? §22.11.

22.3 National coverage determinations

A national coverage determination (NCD) is a nationwide determination by CMS as to whether Medicare will pay for a specific item or service.

Four things to know.

They are binding nationwide. A Medicare Administrative Contractor cannot pay for something an NCD excludes, and cannot refuse to cover something an NCD covers when its conditions are met.

They are relatively few. There are a few hundred, covering items and services that were contentious, expensive, or new enough to warrant a national decision. Most services have no NCD, which is exactly why local determinations exist.

They come in three shapes: covered, not covered, or covered with conditions — and the third is the most common and the most operationally important. "Covered when [these criteria] are met" is an instruction, and the criteria are what your documentation has to satisfy.

And they are public and searchable. The Medicare Coverage Database holds them, free.

Two features of NCDs worth knowing by name

Coverage with evidence development. For some items and services — typically newer ones with promising but incomplete evidence — CMS covers them only when the beneficiary is enrolled in an approved clinical study or registry. The service is covered and conditioned on data being collected.

Operationally this means a claim carries requirements a coder would not otherwise expect — specific registry participation, specific reporting, sometimes specific codes or modifiers. If a service is covered under this mechanism, the policy will say so, and the requirements are not optional.

And NCD reconsideration. Like LCDs, NCDs can be reconsidered, through a formal process with a public docket. It is slower and higher-stakes than LCD reconsideration and is generally driven by manufacturers, specialty societies, and CMS itself rather than by individual practices — but the docket is public, which means you can see what is coming.

22.4 Local coverage determinations and billing and coding articles

A local coverage determination (LCD) is a determination by a Medicare Administrative Contractor as to whether an item or service is reasonable and necessary within that contractor's jurisdiction.

LCDs are where most of the action is, for one reason: there are far more of them than NCDs, and they cover the services practices actually bill.

Three facts that matter operationally.

They are jurisdictional. A contractor covers a defined geographic area, and its LCDs apply there. The same service can be governed by different LCDs in different states — which is genuinely uncomfortable and is a real feature of the system rather than an accident.

They cannot contradict an NCD, and they may not exist for a service an NCD already governs.

And they are developed through a public process with comment periods and published rationales.

The process, and why a coder should care about it

LCDs are not handed down. The process, in outline:

   A proposed LCD is published
        ▼
   A COMMENT PERIOD opens — anyone may comment
        ▼
   Open meetings and a Contractor Advisory Committee
        ▼
   The final LCD is published with a RESPONSE TO COMMENTS
        ▼
   A NOTICE PERIOD before it takes effect
        ▼
   And afterward: a RECONSIDERATION process — anyone may
   request that an existing LCD be revised, with evidence

Three reasons this matters to someone doing the work.

You get advance warning. A proposed LCD affecting a service you bill is published before it governs, and the notice period exists so you can prepare. Practices that read them are not surprised; practices that do not, are.

The response to comments is the most useful document in the set. It states what objections were raised and why the contractor did or did not change the policy — which tells you exactly what arguments have already been tried, and is therefore the best possible preparation for an appeal.

And the reconsideration process is genuinely open. A practice, a specialty society, or an individual may request revision with supporting evidence. It is slow and it works sometimes, and it is the only mechanism in this entire book by which the person doing the coding can change the rule rather than comply with it.

That last one is worth saying plainly, because nothing else in this book offers it. Everywhere else the rules are given. Here there is a door, it is public, and almost nobody in the field knows it is there.

Billing and coding articles

And here is the document most coders actually need, which many have never heard of.

A billing and coding article accompanies an LCD and contains the operational detail — the specific CPT and HCPCS codes the policy applies to, the ICD-10-CM codes that support medical necessity, coding guidance, documentation requirements, and utilization expectations.

The LCD says what is covered and why. The article says which codes to use and which diagnoses support it.

If you have ever wanted a list of which diagnosis codes will support a given service, this is that list, and it is free.

Two things follow.

A great many "why did this deny for medical necessity" questions are answered by opening the article and checking whether your diagnosis is on the list. That takes about two minutes and it is not a skill.

And the article is where the codes changed. LCDs are revised; the code lists in the articles are revised more often. A practice working from a printed list from three years ago is working from a document that has moved.

22.5 Finding the policy that governs your claim

A procedure, in order, because doing it out of order wastes time.

   1. WHO IS THE PAYER?
        Medicare  → continue to 2
        Commercial → skip to 22.10 (medical policy)
        Medicaid   → the state's own policy, plus its manual

   2. IS THERE AN NCD?
        Search the Medicare Coverage Database by code or topic.
        If yes, it governs nationwide. Read its conditions.

   3. IS THERE AN LCD in YOUR jurisdiction?
        Search by code AND by state. Jurisdiction matters.

   4. READ THE BILLING AND CODING ARTICLE attached to it.
        ► THE CODE LISTS ARE HERE
        ► THE SUPPORTING DIAGNOSIS LIST IS HERE
        ► THE DOCUMENTATION REQUIREMENTS ARE HERE

   5. NO POLICY FOUND?
        Then no published criteria govern, and the ordinary
        reasonable-and-necessary standard applies. Document
        the clinical rationale.

Step 5 is worth dwelling on, because coders assume a policy always exists and become paralyzed when they cannot find one.

Most services have no NCD and no LCD. That does not mean they are automatically covered; it means there are no published criteria, the general standard applies, and the record's clinical rationale is what supports it. The absence of a policy is not a problem to be solved. It is the normal case.

🔢 Code It

Two searches, five minutes, and most medical necessity questions are answered.

Search one — the Medicare Coverage Database, by CPT code. It returns any NCD, any LCD in each jurisdiction, and the associated articles. Filter to your state.

Search two — inside the billing and coding article, find the diagnosis list. Is the diagnosis you assigned on it?

text Diagnosis IS on the list ....... the service is supported; a denial is likely something else Diagnosis is NOT on the list ... the claim will deny, and it will deny correctly ► Is there a DIFFERENT, TRUE diagnosis that IS supported and IS documented? If yes: the coding was wrong. If no: the SERVICE was never covered for this patient. §22.11.

That last fork is the whole chapter, and §22.11 is its worked example.

22.6 Linking the diagnosis to the service

Coverage is established by the relationship between a diagnosis and a service, and the claim carries that relationship explicitly.

On the CMS-1500, diagnosis pointers connect each service line to the diagnoses that justify it — Chapter 25 §25.5 covers the mechanics. The pointer is the linkage, and it is not decorative: a service pointed at a diagnosis that does not support it will deny even when a supporting diagnosis is elsewhere on the claim.

The two ways linkage fails

The diagnosis is wrong. The service is supported by some diagnosis the patient genuinely has, and the claim reported a different one. This is a coding error and it is fixable — by correcting the claim, if the correct diagnosis is documented.

The diagnosis is right and it does not support the service. The patient's condition is what it is, and the policy does not cover this service for it. This is not a coding error and it is not fixable, and the only honest options are a different service, additional documentation establishing a covered indication that genuinely exists, or an acknowledgment that the service is not covered.

⚖️ Compliance Check

The pressure to "find a diagnosis that works" is the most direct compliance risk in this chapter, and it arrives in an entirely sympathetic form.

A service was performed. The patient needed it. The claim denied because the diagnosis is not on the list. Somebody asks whether there is another diagnosis that would be covered.

That question is legitimate exactly once, and only in this form: "is there a condition this patient genuinely has, that is documented in this record, that supports this service?" If yes, the original coding was incomplete and correcting it is correct.

It becomes something else the moment the answer is no and the question is asked again. Selecting a diagnosis because it is on a covered list, rather than because the record supports it, is coding to the policy rather than to the record — and Chapter 5 §5.3's standard reaches it without difficulty.

The tell is direction. A coder who reads the record and then finds the code is coding. A coder who reads the covered-diagnosis list and then goes looking in the record is doing something else, and the difference is visible in nothing except the sequence of the two acts.

The Encounter's documentation gap

Figure 4.2 contains three documented gaps, frozen since Chapter 4. The first one belongs to this section.

The assessment never states that conservative therapy failed.

Six weeks of intermittent ibuprofen with partial relief is documented — in the history of present illness, not in the assessment. The information exists and is in the wrong section.

Why it matters here. Many payer policies for joint injection require documented failure of conservative therapy. Northfield Mutual's does — §22.11 works a related case in detail.

The information is in the record. A reviewer reading the whole note will find it. But it is in the narrative rather than in the assessment, which means:

• A reviewer skimming the assessment will not see it
• A structured review looking for it in the plan will not find it
• And the coder cannot assert it, because the coder's job is not to relocate the physician's reasoning

This is not an error. The note is complete and truthful. It is a documentation improvement opportunity of the most ordinary kind — the right fact, in the wrong place — and Chapter 38 §38.3 is the conversation that fixes it. One sentence in the assessment, and the note is unambiguous to every reader.

22.7 Frequency limitations and the calendar

Some services are covered only so often, and the limitation is a real coverage criterion rather than a suggestion.

Frequency limitations come in several shapes, and the shape decides the arithmetic:

"Once per calendar year" and "once every 12 months" are NOT the same rule, and the difference

is the source of a specific, recurring, entirely preventable denial.

Under a calendar-year rule, a service on December 20 and again on January 5 is two calendar years and both are payable.

Under a 12-month rule, the same two dates are sixteen days apart and the second one denies.

Three operational facts.

The clock runs from the last date of service, not from the last claim. A service performed elsewhere and paid by the same payer starts the clock, and your system does not know about it.

You frequently cannot verify it. The patient may have received the service from another provider, and the only party who knows is the payer. Eligibility and benefit inquiries can sometimes answer it — Chapter 24 §24.3 — and are underused for exactly this.

And "the patient says they haven't had one" is not verification. People misremember, and a service performed during a hospital stay may not register with them as the same service at all.

🧮 Run the Numbers

Same patient, same service, four dates, two rules.

Last service: December 20. The patient presents on each of the dates below.

Date	"Once per CALENDAR YEAR"	"Once every 12 MONTHS"
January 5	PAYABLE — new calendar year	DENIES — 16 days
June 30	PAYABLE	DENIES — 6 months
December 19 (next year)	PAYABLE	DENIES by one day
December 21 (next year)	PAYABLE	PAYABLE

Under a calendar-year rule, three of the four are payable on the same reasoning. Under a 12-month rule, three of the four deny, and the third one denies by a single day.

The December 19 row is the one that costs money. A patient scheduled a year later, near the anniversary, is a routine scheduling act — and under a rolling-period rule it is a denial nobody anticipated, on a service everybody agrees the patient needed.

The fix is scheduling, not billing. If the rule is 12 months, the appointment goes on the calendar for the anniversary plus a margin, and the person who books it needs to know that. This is one of the few places in this book where the highest-value intervention belongs to the front desk.

22.8 The Advance Beneficiary Notice: when it is required, when it is improper

The ABN is the mechanism that shifts financial liability to a Medicare beneficiary when a service is expected to be denied — and the rules about when it is improper are as important as the rules about when it is required.

What it is

An Advance Beneficiary Notice of Noncoverage is a standardized written notice, on the CMS-approved form, given to a beneficiary before an item or service is furnished, when the provider believes Medicare may not pay because it is not reasonable and necessary.

Its purpose is to let the beneficiary make an informed choice about whether to receive the service and accept financial responsibility.

Four requirements for a valid ABN, and each of them fails in practice:

It must be given in advance — before the service, with enough time for the beneficiary to consider it. A form signed at checkout is not advance notice.

It must identify the specific item or service and state a specific reason the provider expects denial. "Medicare may not pay" is not a reason. "Medicare does not pay for this test more often than once every 12 months and you had one in March" is.

It must include a cost estimate.

And the beneficiary must be able to choose. The form offers options, and the beneficiary selects one and signs.

The three options — and who chooses

The form's option block is where the ABN stops being paperwork and becomes a decision, and it is the part staff most often fill in for the patient. They may not.

   OPTION 1  "I want the service. Bill Medicare."
               ► the service is furnished
               ► a claim IS submitted, with modifier GA
               ► the beneficiary may appeal the denial
               ► the beneficiary is liable if Medicare denies

   OPTION 2  "I want the service. Do NOT bill Medicare."
               ► the service is furnished
               ► NO claim is submitted
               ► the beneficiary pays and has NO APPEAL RIGHTS

   OPTION 3  "I do not want the service."
               ► the service is not furnished
               ► no claim, no liability

Three rules about the options.

The beneficiary selects one. Not the practice. A form arriving at the billing office with an option pre-checked is a defective ABN, and it is defective in a way that matters — the whole point of the notice is that the beneficiary chose.

Option 1 preserves appeal rights and Option 2 destroys them. Patients rarely understand this. Option 1 is almost always better for the patient, and a practice that explains the difference is doing something genuinely useful.

And only one option may be selected. A form with two boxes checked, or none, is defective.

⚠️ Where Claims Die

A defective ABN is worse than no ABN, and practices consistently believe the opposite.

With no ABN, the situation is clear: the service was expected to be denied, no notice was given, GZ, and the patient may not be billed. Expensive, correct, over.

With a defective ABN — routine, blanket, undated, reasonless, pre-checked, or signed at checkout — the practice has a document showing it knew a notice was required and produced one that does not work. It reports GA, asserting a valid ABN, which is not true.

And it will bill the patient, on the strength of a form that does not shift liability.

The audit finding writes itself, and the remedy is unusually simple: if your ABNs are collected at checkout on a stack of forms with a generic reason, you do not have ABNs. You have paper.

When it is IMPROPER

This is the half nobody teaches and it matters.

A ROUTINE ABN — given to every patient, for every service, as a matter of policy — is improper.

A BLANKET ABN — given without a specific, individualized reason to expect denial — is improper.

The reason is structural. An ABN's function is to inform a beneficiary of a specific, expected noncoverage so they can decide. A form everyone signs for everything informs nobody of anything, and it converts a protection for the patient into a waiver harvested by the provider.

Two more improper cases:

An ABN for a statutorily excluded service is not required — the ABN exists for services that would be covered but for the reasonable-and-necessary determination. A service that is not a benefit at all is never covered, and no notice is needed to bill the patient. A voluntary notice may be given as a courtesy — that is what modifier GX identifies — but it is not an ABN in the required sense.

And an ABN cannot be given after the fact. A notice signed after the service does not shift liability. It shifts nothing.

🗂️ The Encounter — Q3, ANSWERED

Should an ABN have been obtained for the March 14 injection?

No — and giving one would have been improper.

Two reasons, and both matter.

First, the patient is not a Medicare beneficiary. Account 10-4471 is covered by Northfield Mutual, a commercial PPO. The ABN is a Medicare instrument. It does not apply to commercial coverage, and a practice presenting a CMS-R-131 to a commercial patient is presenting a form that has no effect on that patient's liability. Commercial plans have their own notice mechanisms, defined in the provider contract — §22.10 — and they are not this form.

Second, and more instructively: even if this had been a Medicare patient, an ABN would have been improper. An ABN requires a specific, individualized reason to expect denial as not reasonable and necessary. There was none. 20610 for M25.561 in a symptomatic patient is an ordinary covered service, and the practice had no basis to expect a medical necessity denial.

The line 1 denial on day 17 was not a medical necessity denial. It was CO-97 with RARC N19 — an edit denial, asserting the E/M was incidental to the procedure. Chapter 21 §21.9 walked it. An ABN would not have addressed it, would not have shifted liability for it, and could not have been justified in advance.

What a routine ABN would have been: exactly the improper practice §22.8 describes — a form collected from everyone, against no specific expectation, converting a patient protection into paperwork.

Q3 is CLOSED. The answer is no, for two independent reasons, and the second one is the one worth remembering.

22.9 GA, GX, GY, and GZ in practice

Chapter 14 §14.11 introduced the four liability modifiers as a decision tree. This section is what they look like on real claims, now that you know what an ABN is.

   IS THE SERVICE A MEDICARE BENEFIT AT ALL?

   NO  → GY   statutorily excluded / not a benefit
              ► the patient MAY be billed; no ABN required
              ► add GX if a VOLUNTARY notice was given

   YES, but you expect denial as not reasonable and necessary:

        ABN obtained, valid ......... GA  ► patient MAY be billed
        ABN not obtained ............ GZ  ► patient may NOT be billed

Four things this section adds to Chapter 14's tree.

GA requires a VALID ABN, not merely a signed form. §22.8's four requirements are the test, and an ABN that was routine, blanket, retroactive, or reasonless does not support GA. Reporting GA on the strength of a defective ABN asserts something untrue.

GY is used affirmatively and it is underused. Reporting GY on a statutorily excluded service tells the payer we know this is not a benefit; we are submitting for the denial so the secondary payer or the patient can be billed. It produces a fast, clean denial, which is exactly what you want, and it is frequently the correct handling of a service everyone knows is not covered.

GX is a courtesy, not a requirement. A voluntary notice for a statutorily excluded service is good practice — patients should not be surprised — and GX records that you gave one.

And GZ is an admission that is better than the alternative. Chapter 14 §14.11 said it and it bears repeating with the ABN rules attached: reporting GZ says you expected denial and did not obtain a notice. The alternative — billing the patient anyway — is the actual violation. GZ costs you the payment. Silence costs you more.

📞 On the Phone

"You didn't have me sign anything and now you're billing me."

This call has two completely different correct answers, and getting the wrong one is expensive in opposite directions.

If the service was statutorily excluded — not a benefit at all — no ABN was required and the patient may be billed. "That service isn't covered by Medicare at all, for anyone — it's not a question of your specific situation. We should have told you in advance as a courtesy, and I'm sorry we didn't." Then find out why nobody told them, because that is a real failure even though the bill stands.

If the service was expected to be denied as not reasonable and necessary and no valid ABN was obtained — the patient may not be billed. Full stop. "You're right. We should have given you a notice in advance and we didn't, so this isn't your responsibility. I'm removing it."

What does not work: producing a form the patient signed at checkout with no specific reason on it. That is not a valid ABN, and asserting it is worse than having none — because now there is a document showing the practice knew a notice was needed and produced a defective one.

22.10 Commercial medical policy and prior authorization

Everything so far has been Medicare's machinery. Commercial plans have their own, and it is structurally similar with three important differences.

The document is called a medical policy — or a clinical policy, or a coverage policy — and it does what an LCD plus its billing and coding article do: state the criteria, list the codes, and frequently list the supporting diagnoses. Most large payers publish theirs.

The criteria are the plan's own. They may be stricter than Medicare's, looser, or simply different, and they can differ between products from the same payer.

And there is a mechanism Medicare mostly does not use: prior authorization.

Prior authorization

Prior authorization is a coverage determination made BEFORE the service, on submitted clinical information, resulting in an approval or a denial.

Three things a coder should understand about it.

An authorization is not a payment guarantee. It generally confirms that the service meets medical necessity criteria as presented. It does not confirm eligibility on the date of service, benefit availability, or that the service performed matched the service authorized.

The authorization is tied to specifics — a code, a date range, a number of visits or units, a facility, sometimes a rendering provider. A service that drifts from any of them may deny despite an authorization on file, and this is one of the more painful denial categories because everyone believed it was handled.

And a denial at prior authorization is information, not a defeat. It tells you, before the service, exactly which criterion is unmet — which is the most useful moment in the entire process to learn it.

Peer-to-peer review

One mechanism worth knowing about because it is underused and because it is frequently the fastest path.

A peer-to-peer review is a conversation between the treating physician and a physician reviewer employed or contracted by the payer, about a specific case.

Three facts.

It is generally available on request after an adverse determination, and there is usually a window.

It is a clinical conversation, not a billing one. The treating physician explains the clinical picture to another physician. A coder cannot conduct it and should not try.

And it frequently resolves cases that written appeals do not, for an unglamorous reason: a conversation surfaces facts that a form did not ask for. A physician can say "the patient failed two prior agents and cannot tolerate the third" in ten seconds; getting that into a written appeal requires someone to have known to include it.

The coder's role is real and narrow: know it exists, know the deadline, and get it scheduled. Practices lose cases every week because nobody asked for a peer-to-peer inside the window.

📋 Read the Chart

Source: referral and prior authorization request, Account 10-5502 Encounter: commercial patient, 44, referred for MRI lumbar spine (72148) What it says:

"Six weeks of low back pain. Four weeks of NSAIDs with partial relief. Patient reports trying physical therapy. No red flags. MRI requested to evaluate for disc pathology."

What the policy requires: six weeks of documented conservative therapy before advanced imaging.

What the record establishes: four weeks of one modality, and a patient report of another — with no dates, no provider, no documentation, and no outcome.

What it means: the authorization will be denied, and it will be denied correctly. CARC 197 — precertification/authorization absent.

What to do about it: §22.11.

Where it appears: on advanced imaging referrals every day, in every practice, and the missing piece is almost always the same one — the conservative therapy happened and was not documented as the policy requires it to be.

22.11 The MRI that was never going to be covered

Account 10-5502, worked through. (Constructed.)

This is the file that teaches the distinction the chapter exists for.

What a practice usually does

The authorization denies. The denial goes into a queue. Someone writes an appeal, explaining that the patient has had six weeks of low back pain, that the physician believes imaging is indicated, and that the patient has tried conservative measures.

The appeal is denied. They write another one, attaching the office note.

That one is denied too.

Somewhere in the second or third round, somebody says the payer is being unreasonable — which Chapter 21's Case Study 1 established is a hypothesis, and which is wrong here for a reason that is visible on the face of the record.

What is actually wrong

This claim was not coded wrong. This service was never going to be covered, on this record, on this date.

The policy requires six weeks of documented conservative therapy. The record documents four weeks of NSAIDs and a patient report of physical therapy.

Four is not six. And a patient's report of having tried something is not documentation of a course of therapy — there is no start date, no provider, no description, and no outcome.

No appeal can fix this, because an appeal argues about the record, and the record says what it says. Three rounds of appeals is three rounds of asking a payer to reach a conclusion its own published criteria forbid.

What actually fixes it

A phone call, four more documented weeks, and a resubmission.

The phone call — to the referring office — establishes what the policy actually requires and what the record currently shows. Ten minutes, and it converts an unwinnable appeal into a plan.

And it is a call worth being good at, because the referring office is not the adversary here and is frequently as frustrated as you are. What works: "The plan's policy requires six weeks of documented conservative therapy and the referral shows four weeks of NSAIDs plus a patient report of physical therapy. If Dr. — can document two more weeks with dates and outcome, we can resubmit and it should approve." Specific, actionable, and it names what is missing rather than that something is.

What does not work: "The MRI was denied." That sentence has been said to every referring office in the country and it produces nothing, because it does not tell them what to do.

Four more weeks of documented conservative therapy, which the patient needs anyway if imaging is not yet indicated, and which the physician can supervise and document properly. This is not a delay tactic. It is the clinical pathway the policy encodes.

And a resubmission with a record that meets the criteria, which is then approved.

The distinction, stated plainly

   CODED WRONG                    NEVER GOING TO BE COVERED
   ───────────                    ─────────────────────────
   The record supports the        The record does not contain
   service; the claim does        what the policy requires.
   not reflect the record.
                                  An appeal argues about a record
   FIX: correct the claim.        that says what it says.

   An appeal is appropriate       FIX: change the facts, then
   and frequently succeeds.       resubmit — or accept that the
                                  service is not covered.

The first question on any medical necessity denial is which of these you are in, and it is

answerable in the two minutes it takes to open the policy and compare it to the record.

A billing office that asks it first will stop writing a category of letters it has been writing for years.

🗂️ The Encounter — Q5, ANSWERED

What is the knee actually wrong with?

Since Chapter 1 the book has withheld this, and Chapter 4's note is why: "consistent with a degenerative process; no definitive diagnosis established today and no prior imaging of this knee is available."

Now there is imaging. (Constructed.) A radiograph obtained after March 14 demonstrates findings supporting unilateral primary osteoarthritis of the right knee — M17.11.

Which does NOT mean the March 14 coding was wrong.

M25.561 — pain in right knee — was CORRECT for March 14.

Chapter 12 §12.3's rule governs: in the outpatient setting, code the condition to the highest degree of certainty established at that encounter. On March 14 no definitive diagnosis was established and no imaging existed. M25.561 is exactly what the record supported.

Three things follow, and the third is the one worth carrying.

The later diagnosis does not reach backward. A claim is coded from the record as it existed. A diagnosis established in April does not make a March code wrong — and a coder who "corrects" the March claim after the imaging has changed a claim to say something the March record did not.

From this point forward, M17.11 is available — for encounters after the imaging, on records that support it.

And the medical necessity consequence is real. Many payer policies for joint injection list osteoarthritis among the supporting diagnoses and treat undifferentiated joint pain less favorably. The March 14 injection was covered anyway — Northfield Mutual paid it, and the day-17 denial was an edit denial on line 1, not a necessity denial on line 2. But had it been questioned, the practice would have been defending a symptom code where the policy expected a diagnosis, on a record where the workup genuinely had not been done yet.

That is not a coding failure. It is the ordinary cost of coding honestly on day one, and it is why Chapter 33 §33.10's documentation argument — and the clinical documentation integrity discipline Chapter 38 builds from it — are about completing records, not about choosing better codes.

Q5 is CLOSED.

What the two closed questions have in common

Q3 and Q5 look unrelated and they turn on the same idea.

Q3's answer is no because the practice had no specific, individualized expectation of denial — and an ABN without one is improper. Q5's answer is that M25.561 was correct because no definitive diagnosis had been established on that date.

Both are answers about what was true at the time, and in both cases the temptation is to reason backward from what happened later — to collect a notice because a denial eventually arrived, or to re-code March because April produced a diagnosis.

Neither is legitimate, and the principle is the same one Chapter 21 §21.8 stated about documentation: the record is evidence of what was known when it was written. A claim is coded from that record. A later fact is a later fact.

It is also, quietly, the answer to a question students ask constantly — how do I code when I do not know yet? You code what is established, you say so, and you let the next encounter carry the next thing. The March 14 note does this explicitly: "no definitive diagnosis established today." That clause is the physician doing exactly the right thing, and it is the reason a coder six weeks later can tell the difference between a code that was wrong and a code that was current.

Summary

Medical necessity is a COVERAGE word. A clinician's judgment that a patient needs a service and a payer's determination that it will pay are different questions with different criteria. A service can be necessary and not covered, or covered and not necessary for this patient, and the physician's opinion is evidence rather than the standard.

"But the doctor ordered it" does nothing. The payer knows. Show that the record contains what the policy requires.

"Reasonable and necessary" is statutory — Social Security Act § 1862(a)(1)(A) — written as an exclusion, limited to diagnosis or treatment of illness or injury, and undefined by the statute itself. The content lives in coverage determinations.

NCDs are nationwide, binding, relatively few, frequently "covered with conditions" — and some are covered only under coverage with evidence development, which imposes registry and reporting requirements a coder would not otherwise expect.

LCDs are jurisdictional, far more numerous, cannot contradict an NCD, and are developed through a public process with a comment period, a response to comments (the best preparation available for an appeal), a notice period, and a reconsideration process anyone may use. That is the only door in this book through which the person doing the work can change the rule rather than comply with it.

The BILLING AND CODING ARTICLE is the document coders actually need — the CPT/HCPCS codes, the ICD-10-CM codes that support medical necessity, documentation requirements, and utilization expectations. Free, and revised more often than the LCD.

Finding the policy: payer → NCD → LCD in your jurisdiction → the article → and if nothing exists, that is the normal case, and the general standard applies.

Linkage fails two ways. The diagnosis is wrong — a coding error, fixable. The diagnosis is right and does not support the service — not a coding error and not fixable.

The tell is direction. Reading the record and then finding the code is coding. Reading the covered-diagnosis list and then going looking in the record is something else.

"Once per calendar year" and "once every 12 months" are different rules. The clock runs from the last date of service, which your system may not know about.

An ABN must be in advance, specific as to service and reason, include a cost estimate, and offer a choice. A ROUTINE or BLANKET ABN is improper. Statutorily excluded services need no ABN (GY, plus GX for a voluntary notice). GA requires a VALID ABN. GZ is an admission and it is better than the alternative.

The three options are the beneficiary's to select, not the practice's: Option 1 — furnish, bill Medicare, preserves appeal rights; Option 2 — furnish, do not bill, destroys appeal rights; Option 3 — do not furnish. Option 1 is almost always better for the patient.

A DEFECTIVE ABN IS WORSE THAN NO ABN. With none, the answer is GZ and the patient is not billed. With a defective one, the practice reports GA — asserting a valid notice that is not valid — and bills a patient on paper that shifts nothing.

Commercial plans use medical policies and prior authorization. An authorization is not a payment guarantee, it is tied to specifics that a drifting service can violate, and a denial at prior authorization is the most useful moment to learn which criterion is unmet. Peer-to-peer review frequently resolves what written appeals do not, because a conversation surfaces facts a form did not ask for — and practices lose cases every week because nobody requested one inside the window.

CODED WRONG vs. NEVER GOING TO BE COVERED

Account 10-5502's MRI is the second. Four weeks is not six, and a patient's report is not documentation. No appeal fixes it. A phone call, four more documented weeks, and a resubmission do.

Q3 is CLOSED: no ABN, for two independent reasons — the patient is commercial, and even for a Medicare patient there was no specific individualized expectation of denial, so a routine ABN would have been improper. The day-17 denial was an edit denial, not a necessity denial.

Q5 is CLOSED: the knee is M17.11 — and M25.561 was CORRECT for March 14. A diagnosis established later does not reach backward.

Key Terms

Medical necessity — a payer's determination, under published criteria, that it will pay for a service. Distinct from a clinician's judgment that a patient needs it. (Ch.22)

Reasonable and necessary — the statutory standard at Social Security Act § 1862(a)(1)(A), written as an exclusion and given content by coverage determinations. (Ch.22)

National coverage determination (NCD) — a nationwide, binding CMS determination whether Medicare covers an item or service. (Ch.22)

Local coverage determination (LCD) — a Medicare Administrative Contractor's determination, within its jurisdiction, whether an item or service is reasonable and necessary. (Ch.22)

Billing and coding article — the document accompanying an LCD containing the applicable CPT and HCPCS codes, the ICD-10-CM codes that support medical necessity, documentation requirements, and utilization expectations. (Ch.22)

Covered indication — a diagnosis or clinical circumstance under which a policy covers a service. (Ch.22)

Diagnosis linkage — the relationship, carried on the claim by diagnosis pointers, between a service and the diagnoses that justify it. (Ch.22)

Frequency limitation — a coverage criterion restricting how often a service is payable, expressed per calendar year, per rolling period, per number of years, per lifetime, or per episode. (Ch.22)

Advance Beneficiary Notice (ABN) — a standardized written notice given to a Medicare beneficiary before a service, when denial is expected as not reasonable and necessary, to permit an informed choice. (Ch.22)

Routine ABN — an ABN given to every patient for every service as a matter of policy. Improper. (Ch.22)

Blanket ABN — an ABN given without a specific, individualized reason to expect denial. Improper. (Ch.22)

Statutorily excluded — not a Medicare benefit at all; no ABN is required and the patient may be billed. (Ch.22)

Prior authorization — a coverage determination made before a service, on submitted clinical information. Not a payment guarantee. (Ch.22)

Medical policy — a commercial payer's published coverage criteria, performing the function of an LCD and its article. (Ch.22)

Coverage with evidence development — an NCD mechanism covering an item or service only where the beneficiary participates in an approved clinical study or registry. (Ch.22)

Response to comments — the document published with a final LCD stating what objections were raised and why the policy did or did not change. (Ch.22)

LCD reconsideration — the public process by which anyone may request that an existing LCD be revised, with supporting evidence. (Ch.22)

Peer-to-peer review — a conversation between the treating physician and a payer's physician reviewer about a specific case, generally available after an adverse determination and within a window. (Ch.22)

Spaced Review

1. State the difference between a clinical determination of need and a coverage determination. Give an example running each way.

2. Where does "reasonable and necessary" come from, and why does it matter that it is written as an exclusion?

3. Distinguish an NCD, an LCD, and a billing and coding article. Which one contains the list of diagnoses that support a service?

4. (Chapter 20) Which of the four reasons a valid code does not pay is this chapter about? Why is it the only one where the record can change the answer?

5. A service denies for medical necessity. Name the two ways linkage can have failed and say which one an appeal can fix.

6. "Once per calendar year" versus "once every 12 months." Give two dates that are payable under one and not the other.

7. Name the four requirements of a valid ABN. Then name two circumstances in which giving one is improper.

8. (Chapter 14) A statutorily excluded service is furnished and a voluntary notice was given. Which two modifiers?

9. Account 10-5502's MRI. Was the claim coded wrong or was it never going to be covered? What fixes it, and what does not?

10. M17.11 is established in April. Was M25.561 wrong on March 14? State the rule and the general principle about later diagnoses.

11. Name the three options on an ABN and say who selects one. Which option destroys the beneficiary's appeal rights, and which is almost always better for them?

12. Why is a defective ABN worse than no ABN at all? Name three things that make one defective.

13. An LCD is proposed that affects a service you bill. Name three things the public process gives you, and say which one is the best preparation for a future appeal.

14. What is a peer-to-peer review, who conducts it, and what is the coder's role?