What His Story Reveals About What Healthcare Gets Wrong
“Physicians decide what's best for their patients — not
insurance companies.” — American Medical Association, National
Advocacy Update, August 14, 2026
The most important healthcare story this week may have
nothing to do with AI, drugs, billion-dollar deals, or the latest medical
breakthrough.
It may be about a man named Mike Salmon.
And a blueberry-cinnamon pie.
In January 2026, 73-year-old Mike Salmon had already endured
an extraordinary medical ordeal.
Three operations related to aortic aneurysms.
Sepsis.
ICU delirium.
Another dangerous aneurysm.
His wife, Kim Clark, watched as doctors confronted
Mike with an agonizing choice.
More surgery.
Or hospice.
The additional operations would be risky.
Mike decided he did not want them.
One doctor told him that without the procedures, he might
have only weeks to live.
So Mike went home with hospice.
And then something happened that medicine does not always
know what to do with.
Mike started getting better.
He slept.
He ate.
He walked.
He regained strength.
He gardened.
He played bridge.
He returned to making his blueberry-cinnamon lattice-topped
pies.
By May, Mike had improved enough that he was discharged from
hospice because he no longer met the eligibility criteria.
His wife, Kim Clark, had watched something profoundly
uncomfortable happen:
The patient's future turned out to be less predictable
than the system's prediction.
Mike called the time that followed his expected death his “bonus
days.”
Then he said:
“This is one of life’s sweet spots.”
That sentence deserves to stay with us.
Because it raises a question far bigger than hospice.
What happens when healthcare becomes so focused on
predicting, measuring, coding, categorizing and processing human beings that it
forgets how unpredictable human beings actually are?
That is where this story gets interesting.
And where I believe it has something important to teach
every physician, clinic owner and healthcare technology founder.
Healthcare's biggest problem may not be that we lack
information. It may be that we have built too many systems around the
information.
Think about what happens to a patient.
A person walks into a clinic.
They have symptoms.
A history.
A family.
Fears.
Preferences.
A job.
Children.
A mortgage.
A religious or cultural background.
A life outside the clinic.
The physician sees all of that.
Then the healthcare machine starts translating the person.
Symptoms become documentation.
Documentation becomes diagnoses.
Diagnoses become codes.
Services become procedure codes.
Clinical decisions become claim data.
Claims become transactions.
Transactions become payment.
Eventually, the person has been reduced to a series of
fields in a database.
Necessary?
Yes.
Sufficient?
Absolutely not.
The human being is larger than the data generated about
them.
Mike Salmon's story makes that painfully obvious.
A prognosis said one thing.
Life did something else.
The Healthcare System Has a Prediction Problem
Medicine loves prediction.
Risk scores.
Algorithms.
Clinical pathways.
Expected length of stay.
Mortality estimates.
Readmission probabilities.
Disease trajectories.
Payment models.
Utilization forecasts.
Prediction is useful.
It saves lives.
But prediction has a hidden danger:
We can begin treating the prediction as if it were the
patient.
Mike Salmon is a reminder that a prognosis is not a destiny.
A risk score is not a life.
A probability is not a person.
A diagnosis is not an identity.
And a claim is certainly not a human being.
This does not mean physicians should ignore evidence.
Quite the opposite.
Good medicine depends on evidence.
But good medicine also requires humility about what the
evidence cannot know.
The Same Problem Exists in Medical Billing
Now let me make the uncomfortable leap.
What does Mike Salmon have to do with medical billing?
More than it might seem.
The same healthcare system that tries to predict what will
happen to a patient also tries to translate what happened to that patient into
a standardized financial transaction.
That transaction is the claim.
And here is the problem:
Claims are compressed versions of reality.
The patient is complicated.
The claim is structured.
The encounter is nuanced.
The code is standardized.
The physician's reasoning may take paragraphs.
The claim may reduce it to a handful of codes.
This compression is necessary.
But every compression creates risk.
Something can get lost.
A modifier.
A diagnosis.
A clinical detail.
A relationship between conditions.
A documentation element.
A payer-specific requirement.
And when something gets lost, the financial system may
interpret the encounter differently from the way the physician experienced it.
That is where denials begin.
The Industry's Favorite Question Is the Wrong Question
The healthcare revenue-cycle industry loves to ask:
“How do we recover more denied claims?”
I think that is the wrong starting point.
The better question is:
“Why did the claim become deniable in the first place?”
That is a very different question.
One is reactive.
The other is preventive.
One creates more work.
The other attempts to eliminate work.
One asks how to become better at fixing mistakes.
The other asks how to stop creating so many mistakes.
This distinction is enormously important for independent
practices.
The Denial Is Usually the Crime Scene
Think about a denied claim as a crime scene.
The denial is not necessarily where the problem began.
It may simply be where the problem became visible.
The real problem could have started:
At registration.
During scheduling.
During eligibility verification.
During authorization.
During documentation.
During coding.
During charge capture.
During claim construction.
Or during submission.
By the time the denial arrives, the original mistake may be
weeks old.
Yet many practices attack the final symptom.
They work harder.
They hire more people.
They add more spreadsheets.
They add another software platform.
They create another queue.
Then they wonder why the system keeps producing the same
problem.
More people processing bad information does not
necessarily create better information.
Sometimes it simply creates more expensive bad information.
This Is Why I Believe Medical Billing Is a Data Problem
I have become increasingly convinced of something:
Healthcare billing is not fundamentally a billing
problem. It is a data-quality problem with financial consequences.
That distinction changes the strategy.
If you think billing is a financial problem, you hire more
billers.
If you think billing is a workflow problem, you redesign the
workflow.
If you think billing is a technology problem, you buy
software.
But if you recognize that billing is fundamentally a data
integrity problem, you start asking different questions.
Where was the information created?
Was it complete?
Was it consistent?
Was it interpreted correctly?
Was it translated correctly?
Was the relevant payer rule applied?
Was the claim validated before submission?
Could the error have been detected earlier?
That is the conversation I believe healthcare needs to have.
And This Is Where AI Gets Interesting
Everyone is talking about AI in healthcare.
AI will code.
AI will document.
AI will predict.
AI will automate.
AI will optimize.
AI will transform revenue cycle management.
Maybe.
But here is the contrarian question:
What if we are using AI to automate the wrong layer of
healthcare?
If the underlying data is poor, AI can process poor data
faster.
If the workflow is broken, AI can accelerate the broken
workflow.
If the rule is wrong, AI can apply the wrong rule at scale.
Automation does not automatically create intelligence.
Sometimes it simply creates high-speed consistency around
a bad process.
That is not innovation.
That is industrialized error.
The AI Question Nobody Wants to Ask
Don't ask:
“Does your billing platform use AI?”
Ask:
“What measurable problem does the AI prevent?”
Does it reduce preventable denials?
Does it identify documentation gaps before submission?
Does it detect inconsistent information?
Does it recognize payer-specific risk?
Does it reduce staff hours?
Does it identify underpayments?
Does it shorten accounts receivable?
Does it reduce physician interruptions?
If the answer is simply:
“It has generative AI.”
That is not an answer.
That is marketing.
Mike Salmon's Story Has Another Lesson
There is another part of Mike's story that deserves
attention.
He did not recover because someone discovered a magical new
technology.
The story describes something much less glamorous.
He went home.
He slept.
He ate.
He moved.
He spent time with people.
He returned to familiar routines.
He began doing ordinary things.
That is important.
Because healthcare sometimes has a strange bias toward the
extraordinary.
The newest machine.
The newest drug.
The newest algorithm.
The newest platform.
But human beings often recover in very ordinary
environments.
Sometimes the innovation is not adding something. It is
removing something.
Remove noise.
Remove unnecessary intervention.
Remove administrative friction.
Remove duplicated work.
Remove unnecessary handoffs.
Remove confusion.
Remove delay.
Remove the things that prevent people from doing what they
already know how to do.
What If Administrative Burden Is a Clinical Problem?
Physicians know the feeling.
You finish seeing patients.
But you're not finished.
Charts remain.
Messages remain.
Prior authorizations remain.
Claims remain.
Documentation remains.
Inbox remains.
Phone calls remain.
The administrative day begins after the clinical day.
Except it isn't really after.
It overlaps.
And over time, the boundaries disappear.
The physician becomes part clinician, part administrator.
That creates a hidden cost.
Administrative work consumes physician attention.
Attention is finite.
A physician who spends an hour fighting a preventable
billing problem has lost an hour that could have gone somewhere else.
Maybe another patient.
Maybe family.
Maybe education.
Maybe rest.
Maybe thinking.
We rarely put a dollar value on that lost attention.
We should.
The Real Cost of a Denial
Suppose a claim worth $500 is denied.
The obvious problem is $500.
But that is not the whole cost.
Someone must identify the denial.
Someone must determine why it happened.
Someone must find the documentation.
Someone must contact someone.
Someone must correct the claim.
Someone must resubmit it.
Someone must monitor it.
Someone must reconcile the payment.
Maybe someone must appeal it.
And during all of that:
Cash flow is delayed.
Staff time is consumed.
Physician time may be consumed.
Stress increases.
The actual cost may be much higher than the number on the
denial report.
The cheapest denial is the denial that never happens.
That sounds obvious.
Yet the industry has built enormous infrastructure around
recovering from preventable mistakes.
We should spend more energy preventing them.
Three Experts, Three Ideas We Should Not Ignore
Atul Gawande: Ask what the patient actually wants
Atul Gawande's work on serious illness and end-of-life care
has repeatedly pushed healthcare toward a deceptively simple question:
What matters to the patient?
Not merely:
What can we do?
But:
What should we do?
That distinction matters in billing and healthcare
technology too.
We can automate almost anything.
But should we?
We can collect almost every piece of data.
But do we need it?
We can create another workflow.
But does anyone need it?
Capability is not the same as value.
Healthcare technology needs more restraint.
Diane Meier: More treatment is not always better care
Dr. Diane Meier's work in palliative care has helped
redefine what quality care means for seriously ill patients.
Her work reinforces a crucial idea:
Care should be aligned with the patient's needs, goals
and quality of life.
That principle should extend to healthcare operations.
More software does not necessarily mean better operations.
More automation does not necessarily mean better care.
More metrics do not necessarily mean better decisions.
The objective should be meaningful improvement, not
technological accumulation.
Don Berwick: Design healthcare around people
Dr. Don Berwick has spent decades emphasizing
patient-centered care and healthcare improvement.
His work points toward a principle that healthcare leaders
sometimes forget:
The system should be designed around the people using it.
Not the other way around.
That means asking physicians where workflows break.
Asking nurses where handoffs fail.
Asking billers which errors repeat.
Asking patients where the system becomes confusing.
And then listening.
Healthcare technology companies often spend too much time
designing from the boardroom.
The workflow is on the floor.
Go watch it.
Three Expert Lessons for Physician Owners
Gawande: Start with what matters.
Meier: Don't confuse more intervention with better
care.
Berwick: Design the system around people.
Put those three ideas together and you get a powerful
operating principle:
Build healthcare infrastructure that removes unnecessary
complexity from human beings.
That includes patients.
And physicians.
And staff.
The Statistics Behind the Story
CMS reported approximately 1.92 million unique Medicare
fee-for-service beneficiaries used hospice in FY2025.
That is a massive number of people entering one of
healthcare's most human forms of care.
It also demonstrates how important hospice has become within
American healthcare.
But the broader lesson is about scale.
Healthcare is increasingly dependent on information.
More patients.
More data.
More payers.
More regulations.
More reporting.
More measurements.
More transactions.
That means data quality is no longer a back-office issue.
It is infrastructure.
And infrastructure determines performance.
The Most Dangerous Word in Healthcare
I would argue that word is:
“Routine.”
Routine billing.
Routine documentation.
Routine claims.
Routine authorizations.
Routine follow-up.
Routine reconciliation.
The moment something becomes routine, people stop
questioning it.
That is where waste hides.
A practice may have been doing something the same way for
ten years.
That does not mean it is efficient.
It may simply mean nobody has challenged it.
Question the Best Practice
Healthcare loves the phrase best practice.
I am skeptical.
The best practice for a 500-bed academic medical center may
be absurd for a five-physician clinic.
A workflow designed for a national health system may
overwhelm an independent practice.
A billing process that makes sense at one payer may fail at
another.
There is no universal administrative workflow.
There are principles.
Accuracy.
Transparency.
Compliance.
Accountability.
Patient-centeredness.
Efficiency.
But the implementation should fit the organization.
The better question is:
What is the simplest reliable workflow for this practice?
Not:
What does everyone else do?
The Independent Practice Is the Ultimate Stress Test
Small and medium-sized clinics are where healthcare
infrastructure gets tested most honestly.
Why?
Because they do not have unlimited resources.
They cannot afford five departments to fix one workflow.
They cannot tolerate endless administrative duplication.
They cannot absorb every payer mistake.
And physicians cannot spend half their week functioning as
unpaid revenue-cycle managers.
Independent practices need leverage.
Not complexity.
That is where technology can help.
What OnnX Is Trying to Solve
This is the thinking behind OnnX.
I am not interested in building another piece of software
that gives practice owners another screen to monitor.
The goal is more fundamental:
Make the revenue cycle less reactive.
Instead of waiting for the claim to fail:
Identify risk earlier.
Instead of asking staff to find the error manually:
Surface the problem earlier.
Instead of burying the reason inside a workflow:
Make the reason understandable.
Instead of adding another middleman:
Give physician-owned practices more direct control.
That is the thesis.
Not AI for the sake of AI.
Not automation for the sake of automation.
Infrastructure that protects human attention.
Five Things I Would Change in a Practice This Week
1. Stop starting with denials
Start with the encounter.
Ask where the information becomes unreliable.
2. Study your last 100 denials
Do not read them randomly.
Categorize them.
Then look for repetition.
Patterns are more valuable than anecdotes.
3. Find the three most expensive recurring errors
Not the three most annoying.
The three most expensive.
Then calculate:
Revenue lost + staff time + physician time + delay.
That is the real cost.
4. Move validation upstream
If an error can be identified before claim submission,
identify it there.
Do not wait for the payer to teach you what went wrong.
5. Measure physician administrative time
This may be the metric your practice is ignoring.
Ask:
How many hours of physician attention are consumed every
month by problems that should be preventable?
Then try to reduce that number.
The Metrics I Would Watch
Clean claim rate
Useful.
But insufficient.
Denial rate
Useful.
But insufficient.
Preventable denial rate
Much more interesting.
Days in A/R
Important.
Net collection rate
Important.
Underpayment rate
Often neglected.
Appeal recovery
Important.
Administrative labor per claim
Very useful.
But I would add one more:
Physician administrative hours per 100 encounters.
Because if your revenue cycle improves financially while
physicians become more buried, you may have optimized the wrong thing.
The Biggest Billing Myths
Myth: More billers solve billing problems.
Sometimes.
But if the workflow is producing the same errors, you may
simply be hiring more people to repair the same broken process.
Myth: AI eliminates billing errors.
No.
AI can reduce some errors.
It can also amplify bad assumptions.
AI scales whatever you give it.
That is why data quality and governance matter.
Myth: Every denial is a billing department failure.
No.
The error may have started upstream.
The billing department may simply be where it became
visible.
Myth: Automation means fewer people.
Not necessarily.
The better goal is fewer low-value tasks.
Humans should handle exceptions, judgment and relationships.
Machines should handle repetition.
Myth: The highest-volume denial is always the biggest
problem.
No.
A low-volume denial involving a high-value procedure may
cost more.
Always calculate financial impact.
The Ethical Question
Here is the ethical question I wish more healthcare
technology founders asked:
What human capacity does this technology return?
Does it give physicians time?
Does it give nurses time?
Does it give billers time?
Does it give patients time?
If the answer is no, what exactly are we optimizing?
Healthcare technology should not merely create efficiency.
It should create capacity for care.
The Legal Reality
Automation does not eliminate responsibility.
A practice using software remains responsible for
appropriate billing, documentation, coding and compliance.
Physician owners should understand:
HIPAA requirements
Payer contracts
Medicare and Medicaid rules
Medical necessity
Documentation requirements
Coding compliance
Overpayment obligations
False Claims Act risks
Data security
Vendor accountability
Before adopting an AI billing platform, ask:
Who is responsible when the system is wrong?
Can the decision be audited?
Can staff override it?
Is the reasoning visible?
How are payer rules updated?
How is patient data protected?
What happens when the model is uncertain?
If the vendor cannot answer those questions clearly, stop.
A Seven-Day Billing Reality Check
Day 1
Pull your last 100 denied claims.
Day 2
Group them by root cause.
Day 3
Calculate the financial impact.
Day 4
Identify where each problem began.
Day 5
Create one upstream prevention rule.
Day 6
Measure staff and physician time involved.
Day 7
Ask whether technology can eliminate the repetitive part.
That final question comes last for a reason.
Technology should follow the problem.
Not the other way around.
The Future of Medical Billing Is Not “AI Billing”
That phrase is too small.
The future should be:
Intelligent revenue-cycle infrastructure.
Clinical data should move more cleanly into administrative
workflows.
Errors should be detected earlier.
Payer-specific risks should be visible.
Claims should be validated before submission.
Underpayments should be identified.
Denials should become learning signals.
Staff should work on exceptions instead of repetitive tasks.
Physicians should have visibility without having to become
billers.
And practice owners should understand where money is being
lost.
That is a much bigger opportunity than simply automating
claims.
The Future Is Human
The irony is that the more technology we introduce into
healthcare, the more important human judgment becomes.
Why?
Because technology handles the predictable.
Healthcare is full of the unpredictable.
Mike Salmon is a perfect example.
A prognosis is valuable.
But Mike was not a prognosis.
A hospice eligibility rule is necessary.
But Mike was not a rule.
A medical record is necessary.
But Mike was not a medical record.
A claim is necessary.
But Mike was never a claim.
He was Mike.
A husband.
A gardener.
A bridge player.
A baker.
A man who expected his life to end and then found himself
making blueberry-cinnamon pies.
That is what healthcare is ultimately trying to protect.
What Healthcare Leaders Should Take Away
The lesson is not:
“Technology is bad.”
It is not:
“AI is bad.”
It is not:
“Billing is bad.”
And it is certainly not:
“Medicine cannot predict anything.”
The lesson is more uncomfortable.
We should be careful about confusing the system's
representation of reality with reality itself.
The diagnosis is a representation.
The prognosis is a probability.
The claim is a transaction.
The dashboard is a summary.
The algorithm is a model.
The patient is the reality.
That distinction should guide healthcare leadership.
The Question Behind Everything
Mike Salmon's story begins with a prediction:
Weeks to live.
It ends with something much harder to measure:
Bonus days.
That gap between prediction and lived experience is where
humility belongs.
And perhaps that is the same place where healthcare
innovation should begin.
Not with:
What can we automate?
Not with:
What can we predict?
Not with:
What can we bill?
But with:
What does the human being actually need?
Then work backward.
That is how patient-centered healthcare should be designed.
That is how medical technology should be designed.
And that is how medical billing should be designed.
Final Thoughts: Don't Optimize the Wrong Patient
Mike Salmon's story has stayed with me because it challenges
a deeply embedded instinct in healthcare.
We want certainty.
We want algorithms.
We want pathways.
We want predictions.
We want clean data.
We want clean claims.
We want measurable outcomes.
All of those things have value.
But human beings do not always follow the spreadsheet.
Sometimes a patient who is expected to die goes home.
Sometimes he starts walking.
Sometimes he starts gardening.
Sometimes he makes a pie.
Sometimes he gets more time.
And sometimes that extra time is the most valuable outcome
of all.
So perhaps healthcare leaders should ask a more
uncomfortable question.
Are we optimizing the healthcare system—or are we
optimizing the human experience of being cared for?
Those are not always the same thing.
And if we are serious about the future of healthcare, they
need to become much closer.
Get Involved: The Conversation Starts With You
Here is my question for physicians and clinic owners:
What is the one administrative problem in your practice
that everyone has accepted as “just the way healthcare works”—even though you
know it should not be?
Tell me in the comments.
Your answer may reveal the next problem healthcare
technology needs to solve.
If this perspective resonates, share or repost this
article with another physician, clinic owner, practice manager or healthcare
leader.
The healthcare system will not become more human simply
because we talk about patient-centered care.
We have to redesign the systems surrounding the patient.
We have to question the workflows we inherited.
We have to challenge “best practices” that no longer make
sense.
And we have to build technology that returns something more
valuable than efficiency:
Human attention.
That is where real healthcare innovation begins.
About the Author
Dr. Daniel Cham is a physician, healthcare consultant
and entrepreneur working at the intersection of medical technology,
healthcare management and medical billing.
He is the founder of OnnX, an AI-powered medical
billing platform focused on helping small and medium-sized physician-owned
practices reduce administrative friction, improve billing accuracy and gain
greater control over their revenue cycle.
His work focuses on a simple question:
How can technology make healthcare work better for the
people actually delivering and receiving care?
Connect with Dr. Cham on LinkedIn to
learn more.
Continue the Conversation
Healthcare is changing faster than most practices can
absorb.
The challenge is not simply understanding the newest
technology.
It is separating real progress from expensive complexity.
Explore more perspectives on healthcare operations, medical
technology, physician entrepreneurship, medical billing and innovation through
Dr. Cham's work.
Personal website:
DrDanielCham.com
Podcast:
The Health
Momentum Podcast on Spotify
YouTube:
Dr. Cham on YouTube
Facebook:
Follow Dr.
Cham on Facebook
Knowledge is only useful when it changes what we do.
Learn something. Question something. Then improve something.
Free Resource
Looking for something practical?
Check the Featured section of my LinkedIn profile for
a free resource.
No signup. No sales pitch. Just something useful you can
take back to your practice.
And if this article made you question how your practice
handles billing, administration or technology:
Repost it.
Someone in your network may be quietly fighting the exact
same problem.
Disclaimer
This article is provided for general educational and
informational purposes only. It does not constitute medical, legal, coding,
compliance, financial or professional advice. Healthcare rules and payer
requirements vary by circumstance and jurisdiction. Physicians, healthcare
organizations and other professionals should obtain appropriate expert guidance
before making decisions based on the issues discussed here.
References
1. KFF Health News — “My Husband Was Kicked Out of
Hospice for Dying Too Slowly.”
Kim Clark's first-person account of her husband Mike Salmon's unexpected
improvement after entering hospice provides the human story that anchors this
article.
Read
the KFF Health News story
2. Centers for Medicare & Medicaid Services — Hospice
Monitoring Report 2026.
CMS's latest hospice monitoring data provide current national context for
hospice utilization, including approximately 1.92 million Medicare
fee-for-service hospice beneficiaries in FY2025.
Read
the CMS Hospice Monitoring Report
3. Centers for Medicare & Medicaid Services — Hospice
Public Reporting.
CMS's hospice reporting resources explain the current quality-measurement and
public-reporting environment surrounding hospice care.
Explore
CMS Hospice Public Reporting
#MedicalBilling #HealthcareRevenueCycle #PhysicianPractice
#PhysicianEntrepreneur #HealthcareInnovation #HealthcareTechnology
#AIinHealthcare #MedicalPracticeManagement #RevenueCycleManagement
#PhysicianBurnout #IndependentPractice #HealthcareLeadership #HealthTech
#ClinicalDocumentation #HealthcareOperations #MedicalCoding
#PatientCenteredCare #HealthcareAI #PhysicianOwnedPractice #OnnX
No comments:
Post a Comment