Why knowing what not to capture may be one of the most important skills in coding and clinical documentation integrity
You see it.
An abnormal laboratory result.
A diagnosis buried in the problem list.
A condition mentioned in a specialist’s note.
An old disease carried forward from a previous admission.
An imaging finding that sounds significant.
A medication that seems to point toward a diagnosis.
For someone trained to look for documentation opportunities, the instinct can be immediate:
Can I code this?
That is a reasonable question.
But it may not be the best first question.
Sometimes the more important question is:
Should I?
There is an enormous difference between those two questions.
One asks whether a code can technically be assigned.
The other asks whether assigning it would accurately represent this particular encounter.
That second question requires judgment.
And judgment is one of the hardest skills to teach.
More Codes Do Not Necessarily Mean Better Coding
Early in a coding career, there can be a natural tendency to believe that finding more diagnoses means performing a more thorough review.
After all, coders are trained not to miss reportable conditions.
CDI specialists are trained to recognize gaps in documentation.
Auditors are trained to identify discrepancies.
Everyone is looking for something.
That can quietly create a mindset in which a chart with twelve diagnoses somehow feels more complete than a chart with six.
But the goal of coding is not to capture the greatest number of diagnoses possible.
The goal is to create the most accurate representation of the encounter possible.
Sometimes accuracy means adding a diagnosis that was overlooked.
Sometimes it means sending a query.
Sometimes it means obtaining greater specificity.
And sometimes accuracy means deliberately leaving something alone.
That is not incomplete coding.
That is disciplined coding.
The Medical Record Contains More Information Than the Claim Should
A hospital record can contain thousands of individual pieces of information.
Laboratory values.
Radiology findings.
Nursing observations.
Prior diagnoses.
Medication histories.
Problem lists.
Consultation notes.
Previous surgeries.
Family history.
Symptoms.
Conditions being considered.
Conditions that were ruled out.
Conditions that were treated.
Conditions that were merely mentioned.
Conditions that have followed the patient for years but did not affect the current encounter.
All of that information may belong in the medical record.
It does not necessarily all belong in the coded data for the current encounter.
For inpatient reporting, the ICD-10-CM Official Guidelines define additional diagnoses as clinically significant conditions that affect patient care by requiring clinical evaluation, therapeutic treatment, diagnostic procedures, extended length of stay, or increased nursing care or monitoring. The same guidance states that diagnoses from earlier episodes that have no bearing on the current hospital stay are excluded from reporting.
That distinction is enormously important.
A condition can be true and still not necessarily be reportable for a particular hospitalization.
Consider the Problem List
The problem list is useful.
It is also dangerous when viewed without context.
Imagine a patient admitted for community-acquired pneumonia.
Their electronic problem list includes:
Hypertension.
History of kidney stones.
Seasonal allergies.
Anxiety.
Hyperlipidemia.
A remote ankle fracture.
Prediabetes.
A history of migraine headaches.
Perhaps all of those conditions are legitimate.
But the existence of a condition somewhere in the electronic record does not automatically establish that it affected this hospitalization in a reportable way.
The question becomes:
What role did this condition actually play in the current encounter?
Was it evaluated?
Was it treated?
Did it alter the treatment plan?
Did it require monitoring?
Did it affect medication management?
Did it contribute to the reason the patient remained hospitalized?
Or was it simply present in the patient’s historical record?
Those are different situations.
A Diagnosis Can Be Real Without Being Relevant
This is one of the concepts I wish were emphasized earlier in coding education.
There is sometimes an assumption that saying a diagnosis should not be coded means saying the patient does not have the condition.
That is not necessarily true.
A patient may absolutely have chronic migraine.
But if the patient is hospitalized for an unrelated surgical procedure, experiences no migraines, receives no migraine-specific treatment, requires no evaluation related to migraine, and the condition has no meaningful effect on the encounter, its existence in the patient’s history does not automatically make it a reportable inpatient diagnosis.
Likewise, a patient may have had pneumonia six months ago.
That history may be completely accurate.
But accuracy of the historical fact does not make pneumonia a diagnosis for today’s admission.
The medical record answers:
What has happened to this patient?
The coded encounter answers a narrower question:
What conditions are appropriately reportable for this episode of care?
Experienced professionals understand the difference.
The Abnormal Lab Trap
This is one of the easiest places to overreach.
A sodium level is low.
A potassium level is elevated.
Creatinine changes.
Hemoglobin drops.
Lactate rises.
Platelets fall.
Glucose is outside the reference range.
It is tempting to translate the abnormal number directly into a diagnosis.
But a laboratory result is not automatically a diagnosis.
Current ICD-10-CM guidance specifically states that abnormal laboratory, imaging, pathologic, and other diagnostic findings generally are not coded unless the provider indicates their clinical significance. When an abnormal finding is being evaluated or treated, clarification may be appropriate.
That means:
Na 132 does not automatically equal hyponatremia for coding purposes.
An elevated creatinine does not give the coder authority to diagnose acute kidney injury.
A low hemoglobin does not automatically establish acute blood loss anemia.
An elevated lactate does not independently establish sepsis.
The finding may be a clue.
It may be important.
It may support a query.
But the laboratory value itself is not permission for us to diagnose the patient.
Sometimes the Correct Action Is a Query.
And Sometimes It Isn’t.
This is where CDI judgment becomes particularly important.
Suppose you notice an abnormal laboratory finding.
The provider has not addressed it.
Do you query?
Not automatically.
First understand the case.
Was the abnormality clinically meaningful?
Was it transient?
Was there active treatment?
Did the physician evaluate it?
Was additional monitoring performed specifically because of it?
Was it part of a known condition already documented?
Would clarification improve the accuracy of the patient’s record?
There should be a defensible clinical reason for asking the provider to clarify.
A query should not exist simply because an abnormal number gives us an opportunity to ask one.
AHIMA’s CDI toolkit emphasizes that queries should be supported by relevant clinical evidence and should never lead a provider toward a particular response. It also states that queries should not communicate potential reimbursement or quality-reporting impact.
That last point deserves attention.
A query should answer a documentation need.
It should not be a mechanism for pursuing a more valuable code.
The Incidental Imaging Finding
Now imagine another case.
A patient undergoes abdominal imaging because of severe abdominal pain.
The radiology report includes several findings.
The first explains the patient’s symptoms.
Another finding, farther down the report, is described as a small renal cyst.
Perhaps there is also mild degenerative change.
Maybe there is a small hiatal hernia.
Maybe there is an incidental anatomic finding the patient never knew existed.
The temptation is to think:
It’s documented. Why wouldn’t we code it?
Because the radiology report and the final coded encounter serve different purposes.
The radiologist is responsible for describing what is seen.
The coder is responsible for assigning diagnoses according to coding and reporting requirements.
An incidental finding may be real while still having no meaningful impact on the encounter.
If it was not clinically addressed, did not affect treatment, did not require additional evaluation, and was not established as a reportable diagnosis by the appropriate provider, automatically coding it can actually make the coded story less accurate, not more.
More information is not always better information.
The Medication Trap
Medications create another subtle risk.
A patient receives furosemide.
Does that prove heart failure?
No.
A patient receives insulin.
Does that automatically tell us the type of diabetes or establish diabetes at all?
No.
A patient receives antibiotics.
That does not independently tell us whether the patient has sepsis, pneumonia, a urinary tract infection, prophylaxis for surgery, or something else entirely.
A patient receives an anticonvulsant.
That does not automatically establish epilepsy.
Medication use can be an important clinical indicator.
It can help the reviewer understand the patient’s treatment.
It can support other evidence.
But it should not be treated as a substitute for diagnosis documentation.
The same principle applies to procedures.
What clinicians do can help us understand what they believe is happening.
It does not always give us the authority to name the diagnosis ourselves.
Then There Is the Problem of Copy-Forward Documentation
Electronic health records are designed to preserve information.
That is useful.
But preservation can turn into propagation.
A diagnosis documented several admissions ago may continue appearing in the problem list.
Yesterday’s assessment may appear in today’s note.
A provisional diagnosis may be copied into subsequent documentation even after the clinical picture changes.
A physician may continue carrying a condition forward because it has become part of the note template.
This creates a dangerous illusion:
repetition can look like confirmation.
But a diagnosis appearing seven times does not necessarily mean seven clinicians independently evaluated and confirmed it.
Sometimes it means the same phrase traveled through seven notes.
This is why the skilled reviewer asks:
Where did this diagnosis originate?
Was it clinically addressed during this encounter?
Has later documentation changed the picture?
Is this active, historical, resolved, or simply carried forward?
Does the discharge documentation clarify the final conclusion?
The person who merely counts appearances may miss what the person following the chronology will see immediately.
“History Of” Is Not a Minor Phrase
Words matter.
Has and had do not mean the same thing.
Active and history of do not mean the same thing.
Resolved does not mean the same thing as current.
Status post does not automatically mean the previous condition remains active.
These distinctions are especially important when patients have complicated medical histories.
Imagine a patient with a history of cancer.
They may still be undergoing active treatment.
They may have completed treatment and have no evidence of disease.
They may have metastatic disease.
They may have a history of one cancer and an entirely new primary malignancy.
Simply finding the word “cancer” somewhere in the chart is not enough.
The reviewer has to understand the present clinical state.
The same principle applies to previous strokes, infections, kidney injuries, respiratory failure, fractures, thromboses, and countless other conditions.
Past disease can matter.
But historical documentation must be interpreted as historical documentation unless the current record establishes otherwise.
What About Conditions That Are Stable?
This is where things become more nuanced.
A condition does not have to be dramatic to be reportable.
A patient does not have to deteriorate.
The diagnosis does not have to prolong the hospitalization.
And there is no rule saying a chronic condition must be unstable before it can qualify as an additional diagnosis.
A stable chronic disease may still affect the patient’s treatment or require monitoring.
For example, a chronic condition may influence medication selection, perioperative management, fluid decisions, glucose monitoring, anticoagulation decisions, renal dosing, respiratory management, or nursing care.
So the question is not:
“Was the condition severe?”
The question is:
“Did this condition affect the patient’s care during this encounter in a way that makes it reportable?”
That requires understanding the chart—not merely counting diagnoses.
A Condition Does Not Need an MCC to Matter
There is another habit worth examining.
Some diagnoses seem to receive more attention because they affect reimbursement.
Acute respiratory failure.
Sepsis.
Acute kidney injury.
Encephalopathy.
Severe malnutrition.
Certain complications.
Because these conditions may influence severity, risk adjustment, quality measures, or MS-DRG assignment, reviewers naturally pay close attention to them.
There is nothing wrong with understanding reimbursement.
Revenue-cycle professionals should understand the financial consequences of accurate documentation.
The danger appears when reimbursement becomes the reason a diagnosis is pursued.
A condition should not become more clinically real because it changes the DRG.
And a diagnosis should not become less important simply because it does not.
Clinical documentation integrity has to work in both directions.
Capture what should be captured.
Clarify what needs clarification.
Challenge what requires validation.
Leave alone what does not belong.
The principle should remain the same regardless of whether the diagnosis changes reimbursement by zero dollars or thousands.
“But the Physician Documented It” Is Only Part of the Conversation
This is an area where precision matters.
Coding professionals generally assign diagnosis codes based on provider documentation. They do not independently apply clinical diagnostic criteria to create diagnoses.
At the same time, CDI and clinical validation processes exist because documentation integrity involves more than locating a diagnostic statement.
AHIMA’s clinical-validation guidance specifically encourages organizations to identify diagnoses that may lack sufficient clinical evidence and use established validation processes and queries to reduce inappropriate reporting.
Those concepts are not contradictory.
They define professional boundaries.
The coder should not silently override the physician’s diagnosis based on personal clinical judgment.
The CDI specialist should not independently diagnose the patient.
But when the documentation and clinical picture appear inconsistent, the answer is not always to ignore the inconsistency.
Sometimes the answer is clarification.
That is a very different action from deciding the diagnosis yourself.
Knowing When Not to Query Is Just as Important
There is another side to restraint.
Not every possible documentation improvement deserves a query.
Read that again.
A reviewer can technically identify a potential clarification opportunity without that opportunity necessarily warranting provider interruption.
Physicians already work in environments saturated with electronic notifications, alerts, messages, documentation requirements, inbox tasks, and administrative demands.
A thoughtful CDI program should therefore care about the quality of queries, not simply quantity.
AHIMA and ACDIS have long emphasized compliant, clinically supported query practices, with the underlying goal of ensuring that documentation accurately reflects the patient’s clinical status. Their 2026 work on updated query guidance continued to center that goal.
Before sending a query, we should be able to explain why the clarification matters to the integrity of the record.
Not merely:
“I found something I could ask about.”
But:
“The record is incomplete, unclear, inconsistent, conflicting, or clinically ambiguous in a way that warrants provider clarification.”
Those are different standards.
Think About the Difference Between Finding and Chasing
Finding is part of good chart review.
Chasing is different.
Finding means noticing that the patient’s creatinine rose significantly from baseline and reviewing the record to understand what happened.
Chasing means deciding that because acute kidney injury could be reportable, every creatinine fluctuation needs to become an AKI query.
Finding means recognizing altered mental status and reviewing the clinical picture.
Chasing means assuming every episode of confusion should become encephalopathy.
Finding means seeing oxygen requirements increase and examining respiratory documentation.
Chasing means approaching every oxygen-dependent patient as a potential respiratory-failure opportunity.
Finding requires curiosity.
Chasing begins with the desired answer.
That difference may be invisible in a productivity report.
But it is enormous from an integrity standpoint.
The Strong Reviewer Is Comfortable With “Nothing to Add”
Imagine reviewing a complicated admission for twenty minutes.
You check the emergency department record.
You follow the labs.
You examine the clinical course.
You review consultations.
You evaluate documentation consistency.
You look for conditions that may require clarification.
And after all of that work, you conclude:
The documentation accurately reflects the patient’s condition.
No query.
No additional diagnosis.
Nothing to change.
Was that review unsuccessful?
Absolutely not.
That may have been excellent CDI work.
You validated the integrity of the existing documentation.
There is a tendency in productivity-driven environments to value visible activity.
Queries are visible.
DRG changes are visible.
Added diagnoses are visible.
A decision not to intervene is much harder to measure.
But invisible judgment still has value.
Sometimes the best thing a knowledgeable professional does is prevent unnecessary intervention.
Overcoding Is Also a Data-Integrity Problem
Undercoding receives significant attention.
And appropriately so.
When legitimate diagnoses are missed, the patient’s severity may not be accurately represented. Quality data may be distorted. Organizations may not receive appropriate reimbursement for resources provided.
But overcoding creates its own problems.
Adding diagnoses that do not accurately represent the encounter can distort:
severity data,
risk adjustment,
quality measurements,
population-health analytics,
utilization patterns,
research datasets,
payer relationships,
and potentially reimbursement.
A record overloaded with irrelevant or unsupported diagnoses is not more complete.
It is noisier.
And noise becomes particularly important when coded data is reused downstream.
Now Add Artificial Intelligence
This issue becomes even more interesting when AI enters the workflow.
AI systems can scan a medical record and identify an enormous number of potential coding and CDI opportunities.
That capability can be valuable.
A machine may notice relationships a person could easily miss.
It may find documentation buried deep inside the record.
It may detect laboratory trends.
It may identify possible diagnoses based on combinations of symptoms, medications, procedures, and clinical findings.
But imagine what happens if we begin treating every AI-generated possibility as something that should be captured.
We could move from under-documentation to over-suggestion.
The future challenge may not be finding enough information.
It may be filtering too much information.
That changes the value of human expertise.
The professional of the future may increasingly have to distinguish:
clinically meaningful from incidental,
active from historical,
supported from questionable,
reportable from merely present,
query-worthy from unnecessary,
and signal from noise.
AI can potentially generate thousands of possibilities.
Someone still needs to determine which possibilities deserve attention.
That is judgment.
The Same Problem Exists in Auditing
Auditors can fall into the same trap from the opposite direction.
A coder may look for diagnoses to add.
A payer auditor may look for diagnoses to remove.
Both approaches can become problematic when the reviewer begins with the desired outcome rather than the clinical record.
A diagnosis should not be accepted simply because it benefits reimbursement.
It should not be rejected simply because removing it reduces reimbursement.
The standard should remain:
What does the documentation and the clinical record support, and what do applicable coding rules require?
A strong auditor should be just as willing to say:
“This diagnosis is supported and should remain.”
as:
“This diagnosis requires clarification.”
That is what objectivity looks like.
Seven Questions to Ask Before Adding Another Diagnosis
Before assigning an additional diagnosis—or pursuing one through a query—consider this short restraint test:
- Is this actually a diagnosis, or am I looking at a symptom, test result, medication, imaging finding, or clinical indicator?
- Is the condition active or relevant to this encounter, or am I seeing historical or copied-forward information?
- Has the appropriate provider established the diagnosis when provider documentation is required?
- Did the condition affect patient care through evaluation, treatment, diagnostic workup, length of stay, nursing care, monitoring, or another applicable reporting requirement?
- Does the rest of the chart support the same clinical story, or is there conflicting information that needs clarification?
- If I am considering a query, is there a genuine documentation-integrity need—or am I pursuing a particular diagnosis because I recognize its coding or financial impact?
- If this code were reviewed months or years from now, could I clearly explain why it belongs on this encounter?
If you cannot answer those questions comfortably, the chart may need another look.
Restraint Is Not Passivity
This distinction matters.
Knowing when not to code does not mean becoming conservative to the point of missing legitimate diagnoses.
It does not mean avoiding queries.
It does not mean ignoring clinical indicators.
It does not mean accepting vague documentation.
And it certainly does not mean failing to advocate for complete and accurate documentation.
Restraint means something more sophisticated.
It means being intentional.
When you query, you know why.
When you code, you know why.
When you challenge, you know why.
And when you leave something alone, you know why.
That is professional judgment.
There Is Confidence in Knowing When to Stop
New professionals often look for certainty in more information.
Another diagnosis.
Another code.
Another query.
Another piece of documentation.
Experience eventually teaches a different lesson.
Sometimes the chart is complete.
Sometimes the condition is not reportable.
Sometimes the abnormality is not clinically significant.
Sometimes the history is simply history.
Sometimes the imaging finding is incidental.
Sometimes the documentation already tells the story clearly.
Sometimes there is nothing else to capture.
Being able to recognize that moment is a skill.
It requires technical knowledge because you must understand the coding rules.
It requires clinical knowledge because you must understand what you are reading.
It requires CDI knowledge because you must recognize when clarification is appropriate.
It requires ethics because you cannot allow the desired financial outcome to determine what you see.
And it requires confidence because sometimes your most defensible conclusion is:
No additional code is warranted.
The Goal Was Never to Code Everything
The goal is to represent the patient accurately.
That means capturing the conditions that mattered.
Clarifying the conditions that are unclear.
Validating documentation that appears inconsistent.
Respecting the boundaries between clinical indicators and provider diagnoses.
Recognizing historical information for what it is.
And resisting the temptation to turn every abnormal finding into another code.
The best health-information professionals are not simply good at finding diagnoses.
They are good at discerning which diagnoses belong in the story of the encounter.
That distinction becomes increasingly important as healthcare generates more documentation, more data, more automated recommendations, and more AI-generated possibilities than any individual could review manually.
The future will not reward people simply because they can find information.
Technology will become very good at that.
The greater value will belong to the professionals who can look at everything that has been found and determine:
This matters.
This needs clarification.
This is supported.
This does not belong.
And this chart is complete.
That is not doing less.
That is knowing more.
The Zed Perspective
At Zed Health Intelligence, we believe one of the greatest differences between technical competence and professional expertise is judgment.
Anyone can learn that a diagnosis has a code.
With education, professionals can learn sequencing rules, conventions, reimbursement methodologies, and clinical indicators.
But expertise develops when those pieces begin working together.
You stop asking only:
“What can I code?”
And begin asking:
“What most accurately represents this patient?”
Sometimes those questions lead to exactly the same answer.
Sometimes they do not.
And knowing the difference may be one of the most valuable skills in health information today.
Discover more from Zed Health Intelligence
Subscribe to get the latest posts sent to your email.
Hi, this is a comment.
To get started with moderating, editing, and deleting comments, please visit the Comments screen in the dashboard.
Commenter avatars come from Gravatar.