Why Clinical Validation Requires More Than Finding the Right Words in the Chart
There it is.
Right there in the progress note.
Acute respiratory failure.
The diagnosis has been documented by the physician. You know the ICD-10-CM code. You know the condition may significantly affect the final coded record.
It would be easy to keep moving.
But something makes you stop.
You go back through the emergency department note. You look at the patient’s oxygen requirements. You review the respiratory assessments. You examine the laboratory findings. You read the consultant’s note. Then you look at what happened over the next 24 hours.
And now you have a question.
Not because you are trying to diagnose the patient yourself.
Not because you believe your clinical judgment should replace the physician’s.
And certainly not because you are looking for a reason to remove a diagnosis.
You have a question because the words in the chart and the story in the chart do not appear to be saying the same thing.
That is where clinical validation begins.
Documentation Is the Starting Point, Not the End of Thought
One of the first lessons people learn in coding is that diagnoses are coded from provider documentation.
That principle remains fundamental.
In fact, ICD-10-CM guidance specifically addresses the relationship between provider documentation and clinical criteria. Code assignment is based on the provider’s diagnostic statement that a condition exists; the coding professional does not independently apply clinical diagnostic criteria to establish or reject the diagnosis. If documentation conflicts within the record, clarification should be obtained.
That last distinction matters.
It means a coder should not look at a laboratory value, decide that a physician’s diagnosis is wrong, and quietly remove the diagnosis.
It also means a CDI professional should not look at a set of clinical criteria and declare a diagnosis on behalf of the provider.
But it does not mean we stop thinking.
Clinical documentation integrity requires us to recognize when something in the record deserves another look.
AHIMA’s clinical-validation guidance specifically addresses diagnoses that may lack sufficient clinical evidence and encourages organizations to establish processes for identifying and resolving those situations.
That is the space between blindly accepting everything we read and attempting to practice medicine ourselves.
And that space requires judgment.
What Does “Clinically Supported” Actually Mean?
This is where the conversation becomes more complicated.
Healthcare rarely operates like an equation:
Finding A + Finding B = Diagnosis C.
Patients do not always present according to textbooks.
Laboratory values change.
Treatment begins before every test result returns.
Chronic disease alters what “normal” looks like.
Medications affect findings.
Patients arrive at different stages of illness.
Physicians sometimes make diagnoses using information that is not immediately obvious to someone reviewing the chart retrospectively.
And legitimate clinical criteria can evolve over time.
That is why clinical validation cannot simply mean:
“The patient didn’t meet my checklist, so the diagnosis is wrong.”
That is not clinical validation.
That is replacing the provider’s judgment with our own.
The better question is:
Does the totality of the documentation provide a reasonable clinical story supporting the diagnosis—and, when it does not appear to, is clarification necessary?
The distinction is subtle, but it is enormously important.
Think About Sepsis
Few diagnoses illustrate this challenge better than sepsis.
Imagine opening a record and finding:
Assessment: Sepsis secondary to urinary tract infection.
You could stop there.
Or you could continue reading.
What brought the patient to the hospital?
What signs of infection were present?
What did the physician identify as the source?
Were cultures obtained?
What treatment was provided?
Were there signs of organ dysfunction?
How did the patient respond?
Did subsequent physicians continue documenting sepsis?
Did a consultant disagree?
Was the diagnosis carried into the discharge summary?
Did the clinical picture change as additional information became available?
None of those questions gives the coder permission to independently determine whether the patient “really had” sepsis.
They do something different.
They help the reviewer understand whether the documentation is internally coherent.
That is an important distinction.
A patient may not look critically ill by the time you review a progress note because treatment has already begun.
A laboratory abnormality may have corrected.
Blood cultures may be negative even though an infection was clinically diagnosed.
One individual finding rarely tells the entire story.
Clinical validation therefore requires context—not simply thresholds.
Or Consider Acute Respiratory Failure
Respiratory failure is another diagnosis that can create difficult reviews.
Suppose the physician documents acute hypoxic respiratory failure.
The reviewer should not ask:
“Can I prove respiratory failure using one number?”
A better review examines the course of the patient.
What was the patient’s baseline respiratory status?
Were they normally on oxygen?
What changed?
How much respiratory support was required?
Was there increased work of breathing?
What were the oxygenation findings?
What did respiratory therapy document?
Was there an underlying pneumonia, COPD exacerbation, heart failure, pulmonary embolism, or other condition?
How quickly did the patient improve?
What did the physician believe was happening?
And perhaps most importantly:
Does the rest of the record support, contradict, or leave unanswered questions about the diagnosis?
That is a very different exercise from simply checking whether a single laboratory value crosses a predetermined line.
A Diagnosis Can Be Clinically Reasonable Even When the Patient Does Not Fit a Perfect Template
This is one of the most important concepts for developing CDI professionals to understand.
Clinical criteria are extremely useful.
They help establish consistency.
They help organizations develop internal standards.
They support education.
They help reviewers recognize cases that deserve closer examination.
But criteria should support critical thinking—not replace it.
AHIMA’s clinical-validation guidance recommends that organizations develop internal criteria and policies to promote consistent review rather than leaving every reviewer to create an individual standard for what constitutes a valid diagnosis.
That matters because inconsistency creates its own integrity problem.
Imagine three CDI specialists reviewing the same diagnosis.
One uses one set of criteria.
Another uses a different definition she learned at a previous organization.
A third relies primarily on personal experience.
Now the outcome of a clinical validation review may depend more on who opened the chart than on what the record contains.
That is not a strong documentation-integrity process.
Good clinical validation needs structure.
But structure should still leave room for clinical complexity.
Now Consider the Opposite Problem
Sometimes the documentation contains very few words—but the clinical picture contains a great deal of information.
A patient arrives confused.
Nursing repeatedly documents disorientation.
The physician orders frequent neurological assessments.
Laboratory abnormalities are treated.
Potential medication causes are evaluated.
Imaging is performed.
The patient’s mental status improves as the underlying condition is corrected.
Yet the documentation never progresses beyond:
Altered mental status.
Now the question changes.
Instead of:
“Is the documented diagnosis supported?”
we may need to ask:
“Does the record support greater diagnostic specificity that the provider has not documented?”
Again, the reviewer does not decide that the patient has metabolic encephalopathy.
The reviewer recognizes an opportunity for clarification.
That difference is the heart of CDI.
We are not manufacturing diagnoses.
We are identifying gaps between the clinical story and the documented story.
Clinical Indicators Are Clues, Not Codes
This deserves repeating because it is one of the easiest mistakes to make when moving from traditional coding into CDI.
A laboratory value is not a diagnosis.
A medication is not a diagnosis.
An imaging finding is not automatically a diagnosis.
An oxygen requirement is not a diagnosis.
A symptom is not necessarily the final diagnosis.
And a treatment does not automatically prove why it was given.
These are clinical indicators.
They help us understand what may be occurring.
They may support existing documentation.
They may contradict existing documentation.
They may reveal missing specificity.
They may support a query.
Or they may simply be part of the patient’s clinical picture without generating any additional reportable diagnosis at all.
The skill is knowing the difference.
This Is Where Less Experienced Reviewers Can Get Into Trouble
When someone first learns CDI, it is tempting to search charts for certain patterns.
Low sodium?
Look for hyponatremia.
Elevated creatinine?
Look for acute kidney injury.
Confusion?
Look for encephalopathy.
Low hemoglobin?
Look for anemia.
High lactate?
Look for sepsis.
That approach can be useful as an early learning tool—but it is not where CDI thinking should end.
Otherwise, chart review becomes a scavenger hunt.
Abnormal value → diagnosis opportunity → query.
Real patients are more complicated than that.
The better reviewer begins asking:
What was the patient’s baseline?
Is this finding clinically significant?
Did the provider evaluate it?
Was treatment directed toward it?
Did it affect the patient’s care?
Is there another explanation?
Does it persist?
Is the condition already documented?
Would clarification materially improve the accuracy of the record?
And most importantly:
Am I trying to understand the patient’s story, or am I searching for a code?
That one question can change the quality of an entire CDI review.
Not Every Abnormality Needs a Diagnosis
The human body produces an enormous amount of measurable information.
Hospitalized patients may have dozens—sometimes hundreds—of abnormal results during an encounter.
If every abnormal laboratory value became a diagnosis, medical records would be filled with conditions that had little meaningful relationship to the patient’s care.
The objective is not maximum coding.
The objective is accurate coding.
That distinction should influence everything from chart review to provider queries to education.
AHIMA’s CDI materials describe the purpose of CDI as helping ensure that clinical documentation accurately represents the patient’s clinical status.
Accuracy sometimes means capturing a diagnosis that is missing.
Sometimes it means obtaining greater specificity.
Sometimes it means clarifying conflicting documentation.
Sometimes it means validating a diagnosis already documented.
And sometimes it means recognizing that no query is necessary at all.
That last skill receives far less attention than it should.
The Most Valuable Query May Be the One You Decide Not to Send
There is a tendency to measure CDI productivity by activity.
How many charts were reviewed?
How many queries were sent?
How many diagnoses were added?
How much financial impact was generated?
Those metrics have operational value.
But they do not tell the entire story.
A seasoned CDI professional can review a case with several abnormal findings and conclude:
The documentation already accurately reflects what happened.
No query.
No additional diagnosis.
No intervention.
That is still valuable work.
In fact, knowing when not to intervene is one of the clearest signs that someone understands documentation integrity rather than simply documentation capture.
A query should exist because clarification is needed—not because the reviewer feels obligated to produce one.
AHIMA and ACDIS guidance emphasizes that compliant queries should be supported by patient-specific clinical indicators and should seek clarification rather than steer the provider toward a predetermined outcome.
The purpose is not:
“How do I get this diagnosis?”
The purpose is:
“What does the record need in order to accurately tell this patient’s story?”
Those are two very different mindsets.
Clinical Validation Is Also About Conflicting Stories
Sometimes the problem is not that a diagnosis lacks support.
Sometimes the problem is that the record contains two different stories.
The emergency physician documents one diagnosis.
The hospitalist documents another.
The consultant disagrees.
The pathology result introduces new information.
The discharge summary carries forward an earlier diagnosis even though later documentation suggests something changed.
Or a problem list continues displaying a diagnosis that may no longer represent the current encounter.
This is why simply searching the electronic health record for a word can be dangerous.
The word may be there.
But when was it written?
By whom?
Under what circumstances?
Was it copied forward?
Was it later revised?
Was it an active diagnosis or historical information?
Did another provider clarify it?
Did subsequent testing change the clinical impression?
The electronic record contains information.
The reviewer has to determine how that information relates to the encounter.
Copy-and-Paste Makes This Even More Important
Modern electronic records make it remarkably easy for documentation to persist.
A diagnosis can travel from one progress note to the next.
A problem list can follow a patient for years.
Elements of yesterday’s assessment can appear in today’s note.
That convenience saves time for clinicians, but it can also create ambiguity for anyone responsible for translating documentation into coded data.
A diagnosis appearing repeatedly in a record can feel more convincing simply because we have seen it five times.
But repetition is not necessarily validation.
Sometimes it is simply repetition.
This is where experienced reviewers learn to distinguish between documentation volume and documentation meaning.
Five copied statements may tell us less than one carefully written assessment explaining what the physician actually believes is happening.
Clinical Validation Protects More Than the DRG
Clinical validation is often discussed in the context of reimbursement.
That makes sense.
Certain diagnoses can change an MS-DRG.
They can affect severity.
They can influence risk-adjusted outcomes.
And unsupported diagnoses can become targets during payer audits.
But reducing clinical validation to reimbursement misses the bigger issue.
Codes become data.
And healthcare data travels.
It may be used in:
research,
quality reporting,
risk adjustment,
population-health analysis,
benchmarking,
resource planning,
provider performance measurement,
predictive models,
and increasingly, artificial-intelligence systems.
A diagnosis that enters the coded record may ultimately influence decisions far removed from the original hospital stay.
That raises the stakes.
The goal should not simply be to create a claim that gets paid.
The goal should be to create a coded representation of the encounter that can withstand the question:
“Why is this diagnosis here?”
Imagine Someone Reviews the Chart Two Years From Now
This is one of the exercises I believe every CDI and coding professional should occasionally use.
Forget the current bill.
Forget the current DRG.
Imagine that two years from now, someone who has never seen this patient opens the record.
Maybe it is:
a payer auditor,
a physician researcher,
a quality analyst,
a compliance professional,
a data scientist,
another healthcare organization,
or an AI system analyzing historical patient data.
They see the diagnosis you reported.
Would the record allow them to understand why it was there?
Would the physician’s documentation make sense alongside the clinical course?
Would contradictory information have been resolved?
Would important specificity have been captured?
Would the diagnosis accurately represent the condition being treated?
That is data integrity.
And that is why CDI is much larger than reimbursement optimization.
AI Will Make This Question More Important, Not Less
Artificial intelligence is becoming increasingly capable of identifying potential diagnoses in medical records.
That is impressive.
It is also relatively predictable.
Machines are exceptionally good at recognizing patterns.
A system may identify that a patient’s laboratory results, medications, respiratory requirements, vital signs, nursing documentation, and physician notes resemble thousands of prior patients who carried a certain diagnosis.
That information could become extraordinarily useful.
But there is still a difference between:
“The data resembles this condition.”
and
“The responsible provider has clinically established and appropriately documented this condition for this patient.”
That difference is where documentation integrity lives.
As AI becomes better at finding possible diagnoses, human reviewers may spend less time asking:
“Did I find something?”
and more time asking:
“Is it meaningful?”
That is a much more sophisticated role.
And it is one of the reasons I believe the future of CDI will require stronger—not weaker—critical-thinking skills.
The Reviewer Who Always Finds Something May Not Be the Best Reviewer
This may be uncomfortable to say.
But it matters.
A CDI professional who generates a query on nearly every chart is not necessarily more skilled than one who queries less frequently.
A coder who produces the highest-severity DRG is not automatically the strongest coder.
An auditor who removes the greatest number of diagnoses is not automatically the strongest auditor.
Volume does not prove judgment.
The strongest reviewer may be the person who can look at a complicated record and explain:
“Yes, this diagnosis is supported.”
Or:
“Something does not align here. We need clarification.”
Or:
“The clinical indicators suggest there may be additional specificity available, but the provider needs to determine that.”
Or simply:
“The documentation is appropriate. No intervention is necessary.”
Every one of those conclusions can represent excellent CDI work.
Before Challenging—or Chasing—a Diagnosis
When something in a chart catches your attention, do not immediately decide whether the diagnosis should stay or go.
Instead, work through the patient.
Start with the reason for admission.
Then understand the baseline.
Follow the timeline.
Look at the relevant clinical indicators.
Look at treatment.
Read what the responsible providers actually documented.
Pay attention to consultants.
Look for contradictory statements.
Look at the patient’s response.
Then look at the final documentation.
Only after you understand the story should you decide whether the documentation requires clarification.
That process feels slower initially.
Eventually, it becomes instinctive.
And that instinct is what separates someone who can code a chart from someone who can truly review a record.
The Goal Is Not to Agree With Every Diagnosis
Nor is the goal to disagree with them.
The goal is documentation integrity.
Sometimes the documentation will support exactly what the provider wrote.
Sometimes the clinical story will raise legitimate questions.
Sometimes greater specificity is available.
Sometimes two providers need to be reconciled.
Sometimes a condition may have been considered early in the hospitalization but no longer represents the final clinical conclusion.
Sometimes a diagnosis that initially looks questionable becomes completely understandable once the entire record is reviewed.
That is why the word validation matters.
Validation does not mean proving the diagnosis wrong.
It does not mean trying to prove it right.
It means asking whether the documentation and the clinical story make sense together—and taking the appropriate compliant step when they do not.
Documentation Integrity Requires Intellectual Humility
Perhaps one of the least discussed CDI skills is knowing the boundary between what we know and what we do not know.
We may have years of clinical experience.
We may understand pathophysiology.
We may know coding guidelines intimately.
We may recognize patterns before they are explicitly documented.
But the ability to recognize a condition is not the authority to diagnose it.
Likewise, the ability to question clinical support is not the authority to independently erase a physician’s diagnosis.
The role requires enough knowledge to recognize that something deserves clarification—and enough discipline to allow the appropriate provider to make the clinical determination.
That balance is not weakness.
It is professional maturity.
So, Is the Diagnosis Supported?
Maybe.
But do not answer after reading one line.
Look at the patient.
Look at the presentation.
Look at the clinical indicators.
Look at what was treated.
Look at what changed.
Look at what the specialists said.
Look at what the provider ultimately concluded.
Look for contradictions.
And when the record still leaves you with a reasonable question, ask it compliantly.
Because the job is not simply to determine whether a diagnosis appears somewhere in the medical record.
The real work is determining whether the medical record tells a clear, complete, consistent, and defensible story about the patient.
That is the difference between finding documentation and understanding documentation.
And that difference is where clinical documentation integrity truly begins.
Zed Perspective
At Zed Health Intelligence, we believe technical knowledge is essential—but it is only the foundation.
The future of health information will belong to professionals who can move beyond finding information to interpreting context, recognizing inconsistencies, asking better questions, and protecting the integrity of the data that ultimately represents the patient.
Because a diagnosis is more than a word.
A code is more than a number.
And behind both of them is a patient whose story deserves to be represented accurately.
Discover more from Zed Health Intelligence
Subscribe to get the latest posts sent to your email.
