The ICD-10 emperor has no clothes

ICD-10 showed up in American healthcare like a glossy brochure for a luxury car: sharper detail, better performance, improved handling, andsomehowlower
costs. We were promised cleaner data, smarter reimbursement, and public-health analytics so powerful they’d practically predict the next flu season while
making coffee.

And then we used it.

In practice, ICD-10 often feels less like a medical classification system and more like an elaborate scavenger hunt where the prize is… getting paid for
the care you already delivered. It’s not that ICD-10 is “bad.” It’s that the way we’ve asked it to serve as the backbone for billing, measurement, risk
adjustment, and research has exposed an awkward truth: ICD-10 is a suit tailored for statistics and reimbursement, not the lived reality of clinical care.
The emperor isn’t naked because the fabric doesn’t exist; the emperor is naked because we keep pretending one outfit fits every occasion.

ICD-10 101: what it is (and what it absolutely isn’t)

ICD stands for International Classification of Diseases. In the U.S., when people say “ICD-10,” they usually mean two distinct code sets used in
administrative transactions:

ICD-10-CM: diagnoses

ICD-10-CM (Clinical Modification) is used to code and classify diagnoseswhat’s wrong with the patient, clinically speaking. It powers morbidity data
across settings: inpatient, outpatient, emergency, and professional claims.

ICD-10-PCS: inpatient procedures

ICD-10-PCS (Procedure Coding System) is used for inpatient hospital procedures. It’s not used by physicians for professional claims in
the same way; it lives primarily in facility billing and inpatient coding workflows.

Here’s the key: ICD-10 is not a clinician’s natural language. Most clinicians think in problems, findings, and plansoften documented in narrative form or
recorded with clinical terminologies (like SNOMED CT) inside the EHR. ICD-10 is a classification, optimized for grouping and reporting. It
was never meant to be the one true vocabulary for everything we do in healthcare.

The promise: specificity, smarter data, and fewer “miscellaneous” buckets

On paper, the rationale for ICD-10 is easy to love:

• More specificity to describe diagnoses and procedures with greater precision.
• Better analytics for population health, quality measurement, and research.
• Improved surveillance for injuries, external causes, and emerging conditions.
• Modernization from an aging code set that struggled to keep up with medical advances.

And yessome of this is real. ICD-10 expanded the code universe dramatically (especially in injury and external cause categories). It introduced laterality
(right vs. left), more detailed complication categories, and more explicit links between conditions and manifestations.

The sales pitch is essentially: “If we name things more precisely, we can understand them more precisely.” Which is true… right up until you remember that
coding depends on documentation, workflows, incentives, and the fact that humans are doing this while also trying to keep people alive.

Reality check: specificity is not the same as truth

ICD-10’s expanded detail is frequently interpreted as “more accurate.” But detail only helps when it reflects reality in the chartand when the chart
reflects reality in the patient.

If a clinician documents “pneumonia,” the coder can’t magically infer aspiration risk factors, organism, severity, or whether it’s ventilator-associated. If
the note says “diabetes,” you don’t automatically get the complication hierarchy, current status, or causal relationships that certain codes require. ICD-10
can encode nuance, but it can’t conjure nuance out of thin air.

This creates a predictable pattern:

• Coder queries increase (because missing details block code selection).
• Clinician burden rises (because documentation becomes a billing instrument, not a clinical tool).
• “Unspecified” codes persist (because the system still needs a code even when reality is… unspecified).

The emperor’s first missing garment is the assumption that “more options” leads to “better data.” In many organizations, more options leads to “more
friction.”

The code explosion: when a classification system becomes an obstacle course

ICD-10 isn’t just a modest update from ICD-9. It’s a redesign with a much larger code set. That expansion isn’t evenly distributedinjuries and external
causes ballooned, and inpatient procedures (PCS) became a high-dimensional grid of attributes.

In practical terms, this means documentation has to support finer distinctions:

• Laterality (right/left/bilateral)
• Encounter type (initial/subsequent/sequela)
• Specific anatomical site and nuance (not just “arm,” but which part of which bone)
• External cause details (how it happened, where, and sometimes what the patient was doing)

When the clinical story is crystal clear, this can feel reasonable. When the clinical story is messyas it often isICD-10 can feel like forcing a
watercolor into a spreadsheet.

“Struck by duck” and other jokes that are actually a design clue

The internet loves weird ICD-10 codes because they’re objectively funny, like a medical version of trivia night. “Struck by duck” gets shared because it
sounds like a cartoon headline. “Burn due to water-skis on fire” sounds like a dare gone wrong.

But those codes aren’t proof that ICD-10 is silly. They’re proof that ICD-10 is trying to do multiple jobs at once:

• Clinical categorization (what condition is being treated)
• Public health and surveillance (how injuries occur, trends, prevention opportunities)
• Administrative and financial logic (grouping for payment models, risk adjustment, denials)

The “absurd code” phenomenon is basically ICD-10 admittingquietlythat it is not purely clinical. It’s administrative infrastructure. And administrative
infrastructure is allowed to be weird as long as it’s consistent.

Here’s what makes the joke codes useful as an analogy: ICD-10 can represent edge cases with stunning granularity, but that granularity doesn’t automatically
improve everyday care. If your clinic can’t reliably capture whether a condition is acute or chronic, “activity, computer keyboarding” isn’t going to save
your analytics strategy.

Money talks: ICD-10 as the language of reimbursement

In the U.S., ICD-10 doesn’t just describe realityit influences incentives. Payment models, risk adjustment, quality programs, and utilization management all
depend on coded data. That means codes can shape:

• What gets paid (and how quickly)
• What gets denied (and how often you’ll appeal)
• How “sick” your population appears (risk scores and benchmarking)
• Which diagnoses become documentation priorities (hello, problem list politics)

This isn’t a moral failure; it’s a system design outcome. If ICD-10 codes drive revenue, organizations will invest in coding optimizationsometimes more
aggressively than they invest in clinical clarity.

Clinical documentation improvement (CDI): the coping mechanism that became a specialty

CDI programs exist because ICD-10’s specificity requirements collide with real-world documentation habits. CDI is often framed as “improving documentation,”
but the unspoken job is “translating clinical reality into the code structure payers want to see.”

Done well, CDI improves clarity and communication. Done poorly, it can devolve into query spam and note-bloating: clinically irrelevant sentences added
solely to satisfy code selection or payer policies. That’s how you get notes that read like: “Patient denies being struck by duck.” (Not helpful, but
technically very specific.)

Data quality: ICD-10 can’t fix what the workflow breaks

A big selling point for ICD-10 was better data. But coded data quality depends on three things that ICD-10 doesn’t control:

1. Clinician documentation quality (clarity, completeness, and causal links)
2. Coder interpretation and training (consistency across humans and organizations)
3. Incentives (what details are rewarded, and what details are ignored)

If any of those are shaky, “more codes” can become “more noise.”

The “unspecified” paradox

ICD-10 has many unspecified options, and they’re not a scandalthey’re a necessity. Medicine is uncertain. Patients arrive mid-story. Diagnostic workups take
time. The chart may not contain the detail needed for perfect specificity. So unspecified codes persist, which means the promised data precision is often
theoretical.

The emperor’s second missing garment is the belief that we can force precision into places where the underlying clinical truth isn’t yet precise.

Mapping: when nuance gets lost on the way to billing

Many EHRs store clinical concepts in richer terminologies than ICD-10, then map them for claims. That mapping step can be lossy. A clinician may document a
condition with nuance (severity, suspected cause, timeline), but the claim ends up with a code that’s “close enough” for classification. The result:
administrative data that looks authoritative but may be missing clinically meaningful dimensions.

The productivity hit: why October 1 came and went… and work got harder anyway

The U.S. transition date to ICD-10 is in the rearview mirror, but the operational impact continues. ICD-10 required:

• Software changes (EHRs, billing systems, clearinghouses)
• Training (coders, clinicians, billers, auditors)
• Testing with payers
• Workflow redesign (especially for high-volume specialties)

Even when the “go-live” was stable, coder productivity often dipped because the search space exploded and documentation requirements became more sensitive.
Organizations compensated with overtime, temporary coders, coding backlogs, and more CDI outreach. The emperor’s third missing garment: the assumption that a
stable implementation date equals a stable workflow.

So… is ICD-10 useless? Not even close.

Here’s the fair, grown-up take: ICD-10 is valuable for what it’s built to do.

• Population-level classification for trends, reporting, and surveillance
• Standardized transactions across payers and providers
• Public health analytics that need consistent groupings
• Payment and policy infrastructure (even if we wish it didn’t require so many gymnastics)

The problem is not the existence of ICD-10. The problem is treating ICD-10 as if it were a faithful mirror of clinical reality, rather than a structured
approximation shaped by documentation, rules, and incentives.

If the suit doesn’t fit, tailor the systemnot the patient story

If you’re in healthcare operations, revenue cycle, or clinical leadership, the practical question is: what do we do with the system we actually have?
Here are strategies that work better than pretending ICD-10 will magically become intuitive.

1) Separate clinical truth from billing truth (without starting a turf war)

Encourage clinicians to document for care first: assessment, differential, and plan. Then use structured support (problem list hygiene, templates used
sparingly, and CDI education) to bridge to coding needs. The goal is clarity, not verbosity.

2) Treat CDI like product design, not policing

The best CDI programs reduce friction: fewer queries, smarter timing, better alignment with clinician workflow, and education that sticks. “More queries” is
not a KPI anyone should celebrate.

3) Invest in computer-assisted coding and smarter tooling

Modern coding support can help surface likely codes, highlight missing specificity, and reduce manual searching. The payoff isn’t just speed; it’s
consistencyespecially in large systems where variability is expensive.

4) Audit for meaning, not just compliance

If your audit program only checks whether a code is defensible, you’ll miss whether your coded data is actually useful. Add audits that ask: does the coded
data reflect the clinical story in a way that supports quality work, population health, and research?

What about ICD-11will the next emperor be better dressed?

ICD-11 is real, it’s designed for a more digital world, and it’s already in effect globally for member states. But “global availability” and “U.S. adoption
for billing and quality programs” are very different beasts. The U.S. has deep dependencies baked into software, payment models, and reporting pipelines.

ICD-11 could be an improvementbut it won’t automatically remove the underlying tensions. Any classification system used for reimbursement will generate
documentation pressure, because money turns categories into rules. If ICD-11 arrives and we keep the same incentive structure, we’ll just build a newer,
shinier obstacle course.

Conclusion: stop worshipping the code set

The point of calling ICD-10 “the emperor with no clothes” isn’t to dunk on coders, clinicians, or the concept of classification. It’s to stop confusing a
billing-and-statistics framework with clinical reality.

ICD-10 is a powerful tool when you respect its limits. The moment you expect it to be a universal languageclinical nuance, reimbursement logic, research
truth, and performance measurement all at onceyou’re back to admiring an outfit that isn’t there.

The fix is not pretending harder. The fix is building workflows, tooling, and incentives that let clinical documentation stay clinically meaningful while
producing administrative data that is accurate enough, consistent enough, and honest about what it represents.

Real-world experiences: what it feels like when ICD-10 runs the show (about )

Ask anyone who lives near the intersection of care and billingclinicians, coders, CDI specialists, revenue cycle leadersand you’ll hear the same theme:
ICD-10 isn’t a single task. It’s a constant background negotiation.

In a busy primary care clinic, the day might start with a patient who has three chronic conditions and one new complaint. The clinician documents a sensible
plan: adjust meds, order labs, follow up. Later, a coder or biller flags the visit because the documentation didn’t state whether the diabetes is “with” a
particular complication, or whether a condition is “history of” versus “active.” The clinician isn’t wrong clinicallypatients don’t arrive labeled like
grocery itemsbut the claim wants a label, and it wants it now. By afternoon, the clinician is answering queries that feel like pop quizzes written by a
robot who has never met a human body.

In the hospital, the friction is different. A patient is admitted with an infection, but the story evolves: cultures return, the source becomes clearer,
complications emerge. ICD-10-PCS procedure coding adds its own twist, because it demands a structured description of what was doneapproach, device, body
partoften based on operative reports written for surgeons, not for coders. When documentation is excellent, coding can be elegant. When documentation is
incomplete, the coder becomes a detective, and the medical record becomes a mystery novel where the ending is whether the DRG makes sense.

Then there’s the payer side. Many organizations report that claims can be delayed or denied for missing specificity, mismatched diagnosis-procedure logic, or
documentation that doesn’t support a higher-severity code. This creates a loop: denials drive more documentation, more documentation drives longer notes, and
longer notes make it harder to find the actual clinical signal. Everyone swears they want “better data,” but the daily experience is that the system rewards
“better defensibility.”

Coders often describe the emotional whiplash: you’re expected to be fast and perfect, while the code set grows and guidelines update. CDI teams
feel caught between two worlds: clinicians who are exhausted by queries, and payers who are unimpressed by clinical common sense unless it’s spelled out in a
specific way. Revenue cycle leaders feel like they’re managing an ecosystem where a tiny documentation detail can change reimbursement and reporting.

And yetdespite the complaintsmost people also see what ICD-10 can do when the organization is aligned. When problem lists are maintained, when
documentation is clear and concise, when coders have good tools, and when CDI focuses on education instead of volume, the system runs smoother. Claims move.
Data becomes more reliable. Clinicians spend less time clarifying what they already know.

That’s the real takeaway from the “emperor has no clothes” metaphor: the fix isn’t burning the wardrobe. It’s admitting what the outfit is for, and stopping
the performance. ICD-10 will never be the language of healing. But with honest expectationsand better design around the humans using itit can be a
serviceable language of reporting, reimbursement, and population-level insight.