What a nurse, a husband, five daughters—and two heartbreaking healthcare stories—can teach us about seeing problems before they become expensive
“AI [is] a helpful second opinion, not a replacement, for
doctors.” — Atul
Gawande, MD, MPH
From Dr.
Atul Gawande’s September 14, 2026 conversation on “10% Happier with Dan
Harris,” discussing medicine, aging, difficult diagnoses and the role of AI in
healthcare.
There is something strange about healthcare.
We have more data than any civilization in history.
More dashboards.
More alerts.
More analytics.
More artificial intelligence.
More predictive models.
More software.
And somehow, we still miss things.
Sometimes spectacularly.
Sometimes expensively.
Sometimes tragically.
And sometimes the problem isn't that nobody had the
information.
The problem is that nobody recognized what the
information meant soon enough.
That distinction matters.
A lot.
And it is why the story of Jenna VanderWeide caught
my attention.
Because Jenna is a nurse.
She understands healthcare.
She understands warning signs.
She understands what can happen when a patient's condition
changes.
Then one morning, she became the patient.
And suddenly, knowing medicine wasn't enough.
The 50 Feet That Nearly Changed Everything
The morning began normally.
Jenna VanderWeide changed the dressing on a wound on her
leg.
She grabbed a book.
She went outside.
Nothing about the morning announced:
“Today is going to be the day your life changes.”
Then she became lightheaded.
She tried to walk the roughly 50 feet from her back porch to
her bedroom.
She couldn't.
She walked a few steps.
Sat down.
Tried again.
Sat down again.
As a nurse, Jenna did what many healthcare professionals
would do.
She checked her blood pressure.
It was dangerously low.
When her husband, Ryan VanderWeide, arrived, Jenna
told him she needed to go to the hospital.
She was suffering from sepsis, a life-threatening
response to infection that can progress rapidly and damage organs.
Within days, the situation became terrifying.
On February 14, Jenna's condition had deteriorated so
severely that Ryan and their five daughters were brought to the
hospital.
Their oldest daughter was 15.
Their youngest was 11.
They were there because there was a real possibility they
were about to say goodbye to their mother.
Ryan remembered the situation bluntly:
“It was just not looking good.”
Jenna doesn't remember much of that day.
In fact, she remembers the morning of February 10.
Then she says she essentially woke up months later.
About three months of her life had disappeared from her
memory.
She survived.
But survival was only the beginning.
She eventually required a tracheostomy and feeding tube and
underwent prolonged recovery and rehabilitation.
Her objective was remarkably simple:
Get home to her five daughters.
Then Something Beautiful Happened
Months later, Jenna returned to AdventHealth Winter Garden.
Something unexpected happened.
People started coming into her room.
They knew her.
She didn't know them.
The nurses remembered caring for her during the period she
could barely remember herself.
The people who had been part of her worst days remembered
her when she couldn't remember them.
And Jenna realized something profound.
She hadn't been merely a patient.
She hadn't been merely a bed number.
She hadn't been merely a chart.
She had been a person.
A mother.
A wife.
A nurse.
A daughter.
A member of a community.
She later said:
“I wasn’t just one person or one number or someone in the
bed. The whole person was being cared for.”
That sentence deserves to travel far beyond the hospital.
Because it raises a much bigger question:
What happens when healthcare systems begin treating
signals as numbers instead of context?
And that brings us somewhere unexpected.
Medical billing.
Wait. What Does Sepsis Have to Do With Billing?
Quite a lot.
And almost nothing.
Let me explain.
A sepsis crisis and a denied claim obviously are not
equivalent.
One can kill someone.
The other can annoy a biller.
Nobody should confuse the two.
The connection is much more specific:
Both illustrate what happens when meaningful signals are
recognized too late.
Jenna's body was producing signals.
Her ability to walk changed.
She became lightheaded.
Her blood pressure dropped.
Her condition deteriorated.
The information mattered.
But information only becomes useful when someone recognizes
its significance and acts.
The same principle exists in healthcare operations.
A claim contains information.
A denial contains information.
A payer response contains information.
A documentation pattern contains information.
A recurring rework problem contains information.
But if we only notice the pattern after the claim fails,
we're looking backward.
And that leads to my contrarian proposition:
Maybe medical billing has been looking in the wrong
direction.
The Claim May Not Be the Problem
Healthcare has become remarkably good at dealing with things
after they go wrong.
Claim denied?
Work the denial.
Claim rejected?
Correct it.
Authorization missing?
Fix it.
Documentation insufficient?
Chase it.
Coding issue?
Rework it.
Payment delayed?
Investigate it.
Staff overloaded?
Buy another dashboard.
Another dashboard?
Buy another dashboard to explain the first dashboard.
At some point, we should probably stop.
The question shouldn't always be:
“How do we fix the problem faster?”
It should be:
“Why did we keep allowing the same problem to happen?”
That's a different question.
And it changes the economics.
The Revenue Cycle Has a Dirty Little Secret
The claim is not where the story begins.
It is where the story becomes visible.
Before the claim existed, a lot already happened.
A patient was scheduled.
Eligibility was checked.
An appointment occurred.
A physician evaluated the patient.
A service was performed.
Documentation was created.
A diagnosis was recorded.
A procedure was captured.
An authorization may have been required.
A payer relationship already existed.
The practice had historical experience.
Then someone assembled the claim.
So when the claim fails, we often ask:
“What's wrong with this claim?”
Maybe we should ask:
“What happened before this claim was created?”
That is the upstream question.
Think About a Restaurant
Imagine going to a restaurant.
You order dinner.
The kitchen prepares it.
You eat.
Then you receive a bill.
Now imagine something is wrong with the bill.
Would you investigate only the receipt?
Probably not.
You might ask:
Did we order that?
Was the price correct?
Was there a mistake?
Was something added?
The receipt is evidence of what happened earlier.
The claim is similar.
It is not the entire billing process.
It is the financial representation of decisions that
happened upstream.
So perhaps we have spent too much time perfecting the
receipt printer.
And not enough time asking why the order was entered
incorrectly.
My Hot Take
Here is the provocative part.
Medical billing doesn't necessarily need more
intelligence at the back end.
It needs more intelligence before the back end becomes
necessary.
We have spent years building increasingly sophisticated ways
to manage failure.
Denial management.
Appeals.
Work queues.
Claims scrubbing.
Payment posting.
AR analytics.
Collections.
All useful.
But there is a paradox:
The better you become at fixing a recurring problem, the
easier it can become to tolerate the problem.
That is dangerous.
A denial team that gets incredibly efficient at processing
10,000 recurring denials may accidentally become an argument for keeping the
upstream process exactly as it is.
The machine gets better.
The problem remains.
Congratulations.
You've optimized the fire department.
You haven't necessarily reduced the number of fires.
Downstream Is Comfortable
Downstream work feels productive.
There is a queue.
There are numbers.
There are tasks.
There are completed items.
Someone can say:
“We processed 4,800 denials this month.”
Impressive.
But here's the uncomfortable question:
How many of those 4,800 denials should never have
existed?
That number is much harder to measure.
And probably much more interesting.
What Is Upstream Billing Intelligence?
I use Upstream Billing Intelligence to describe an
operating model built around earlier recognition.
Instead of waiting for the claim to tell us something went
wrong, the practice asks:
What do we already know?
What happened in similar encounters?
What does this payer typically require?
What patterns have we seen?
Is this encounter unusual?
Is the documentation likely to create friction?
Is there an operational variable worth reviewing?
Does the situation deserve human attention before
submission?
The goal isn't perfect prediction.
The goal is earlier awareness.
Earlier awareness creates options.
Options create choices.
Choices create better decisions.
And No, I Don't Think AI Is the Answer to Everything
This is where I may disappoint the AI crowd.
AI is not a strategy.
AI is a capability.
Putting AI into a broken process doesn't magically make the
process intelligent.
Sometimes it just makes the broken process move faster.
Which is a fascinating way to arrive at failure more
efficiently.
Healthcare technology should start with:
What decision are we trying to improve?
Then:
What information is needed?
Then:
What should the human see?
Then:
Can software help?
Then:
Can AI help?
The order matters.
Atul Gawande's Point Is Bigger Than AI
That is why the recent conversation with Atul Gawande,
MD, MPH, caught my attention.
In his September 14 conversation with Dan Harris, Gawande
discussed AI as a “helpful second opinion” rather than a replacement for
doctors.
That idea applies far beyond clinical medicine.
A useful AI system shouldn't necessarily say:
“Here is the answer.”
It can instead say:
“Here is what I see. Here is why I see it. Here are the
relevant possibilities. You decide.”
That is much closer to how complex healthcare actually
works.
The Future Is Human-Augmented Intelligence
I call this human-augmented intelligence.
The machine does what machines are good at.
It can:
- organize
information
- compare
patterns
- search
historical data
- identify
anomalies
- summarize
context
- surface
potential risks
- present
possible options
The human does what humans should continue doing.
- judgment
- context
- accountability
- ethics
- exceptions
- communication
Then something important happens.
The human decision becomes data.
The outcome becomes feedback.
The feedback improves the next recommendation.
The loop becomes:
Data → Context → Recommendation → Human Decision →
Outcome → Learning
That is far more interesting than:
AI → Magic → Profit
Physicians Don't Need Another Robot Boss
Let's be honest.
Physicians already have enough alerts.
Enough inboxes.
Enough notifications.
Enough “action required” messages.
Enough systems that supposedly save time while requiring six
hours of training.
Healthcare software has occasionally achieved the remarkable
feat of turning a five-minute task into a 14-step workflow.
If the solution requires a 72-page manual and three
super-users, perhaps the problem is not the user.
Perhaps the software needs therapy.
Practice-Specific Intelligence Changes the Equation
One of the biggest problems with generic billing
intelligence is that practices are not generic.
A dermatology practice is not a cardiology practice.
A rural family practice is not a metropolitan multispecialty
group.
A physician-owned clinic is not a large health system.
Payer mix differs.
Patient populations differ.
Geography differs.
Contracts differ.
Documentation differs.
Staffing differs.
Workflows differ.
Historical performance differs.
So why should every practice receive the same supposedly
“best” answer?
Maybe there isn't one universal answer.
Maybe there is:
the best-informed answer for this practice, in this
situation, at this moment.
That is a much more interesting problem.
The Practice Should Have a Memory
Imagine a billing system that actually remembers what the
practice has learned.
Not merely:
“What happened?”
But:
“What did we learn?”
Suppose a payer repeatedly creates a particular problem.
The practice shouldn't have to rediscover it every month.
Suppose one documentation pattern repeatedly creates rework.
The system should recognize it.
Suppose one workflow consistently performs better.
The system should remember.
Over time, the practice develops something valuable:
its own operational intelligence.
Not generic intelligence.
Not somebody else's benchmark.
Its intelligence.
The Best Answer May Not Be One Answer
Here's another contrarian idea.
Healthcare technology loves the word optimal.
Optimal workflow.
Optimal code.
Optimal reimbursement.
Optimal outcome.
Reality is usually less cooperative.
Sometimes there are several reasonable choices.
One approach may minimize administrative risk.
Another may reduce staff burden.
Another may require more documentation.
Another may be operationally faster.
The appropriate choice depends on context.
So why force the software to pretend there is only one
answer?
A better system could present a spectrum of approaches.
For example:
Lower operational risk
Balanced approach
Higher complexity / greater administrative burden
Explain the trade-offs.
Let the human decide.
Record the decision.
Measure the result.
Learn.
That is decision support.
Not decision replacement.
The Human Decision Should Matter
Suppose the system recommends Option A.
The biller chooses Option B.
The claim is paid.
What happened?
The system learned.
Suppose the system recommends Option A.
The biller chooses Option B.
The claim is denied.
The system also learned.
Suppose the system recommends nothing.
The claim is denied.
Again:
information.
This is why I don't believe the objective should be an AI
system that is never wrong.
That standard is unrealistic.
The better objective is:
A system that becomes more useful because the
organization learns from every important decision.
Failure Is Data
This may be one of the most important ideas in modern
healthcare technology.
Failure shouldn't automatically disappear into a denial
queue.
It should become structured learning.
What happened?
Why?
Was the system aware?
Was the human aware?
Was the recommendation ignored?
Was the recommendation wrong?
Was the underlying information incomplete?
Was the payer behavior unusual?
Did the workflow itself create the problem?
Failure is not just an expense.
Failure is information about the system.
The Audit Trail Becomes More Important
If software influences an important billing decision, the
organization should be able to reconstruct the decision later.
What did the system see?
What did it recommend?
What did the human see?
What changed?
What did the human reject?
Why?
What happened afterward?
This isn't about turning every billing interaction into a
legal deposition.
It is about accountability.
If an algorithm affects a decision, the practice should not
have to say:
“We don't know. The software did it.”
That's not governance.
That's outsourcing responsibility.
The Legal and Compliance Line
There is another important distinction.
The objective of billing technology should never be:
“How do we get paid more?”
It should be:
“How do we represent the care accurately, support
appropriate reimbursement and reduce avoidable administrative friction?”
Those are different philosophies.
A system that identifies a legitimate documentation gap is
useful.
A system that encourages unsupported coding because a
particular code pays more is a problem.
Optimization must remain inside the boundaries of accurate
documentation, coding rules, payer requirements and applicable law.
Revenue should be the result of accurate healthcare—not
the target that distorts it.
The Numbers Are Big Enough to Matter
The CDC's current 2026 data says approximately 1.7
million adults develop sepsis each year in the United States, and at least 350,000
adults who develop sepsis die during hospitalization or are discharged to
hospice.
The point isn't to turn a sepsis story into a billing
analogy.
The consequences are fundamentally different.
The point is that healthcare systems can have meaningful
signals before a crisis becomes obvious.
And when recognition is delayed, options can disappear.
That principle matters clinically.
It also matters operationally.
Then Amanda Ireton Became the Daughter
Jenna's story isn't the only reason this week's sepsis
stories matter.
Amanda Ireton, an infection prevention manager at
AdventHealth Avista in Louisville, Colorado, recently shared another deeply
personal experience.
Two years ago, she lost her mother to sepsis.
Ireton already worked in infection prevention.
She knew sepsis clinically.
But experiencing it as a daughter changed everything.
She described feeling “completely helpless” while
trusting the healthcare team.
She also said the hospital where her mother was seen did not
have a sepsis program or sepsis alert, and she believes earlier suspicion might
have changed the outcome. That is Ireton's own belief, not an independently
established causal finding.
That distinction matters.
Her story is not proof that an alert would have saved her
mother.
It is evidence of something else:
What looks like a process improvement from inside a
healthcare organization can feel like life-or-death uncertainty from the family
side.
And that is exactly why systems matter.
From Jenna to Amanda to Billing
Look at the progression.
Jenna VanderWeide: nurse → patient.
Ryan VanderWeide: husband → terrified caregiver.
Five daughters: children → possible goodbye.
Amanda Ireton: healthcare professional → grieving
daughter.
Healthcare teams: clinicians → trusted strangers.
Then finally:
Billing teams: professionals → people trying to make
sense of fragmented information.
Different circumstances.
Different stakes.
But a shared operational principle:
The information is only valuable if someone recognizes
its meaning at the right time.
Healthcare Is Not a Spreadsheet
This sounds obvious.
But software design sometimes forgets it.
A patient is not a row.
A physician is not a utilization metric.
A family is not a satisfaction score.
A claim is not merely a transaction.
A denial is not merely a code.
Numbers matter.
But numbers need context.
Otherwise, we end up optimizing what is easiest to measure
rather than what is most important.
The Real Cost of a Denial
A denial isn't simply lost revenue.
It creates work.
Someone must find it.
Someone must interpret it.
Someone must review the record.
Someone may contact the payer.
Someone corrects it.
Someone resubmits it.
Someone monitors it.
Someone explains it.
Someone eventually reports it.
Multiply that by hundreds or thousands.
Suddenly the denial has become:
a labor problem.
Then:
a staffing problem.
Then:
a management problem.
Then:
a physician-owner problem.
The dollar amount is only part of the cost.
The other cost is organizational attention.
The Question Physician Owners Should Ask
Don't ask only:
“What's our denial rate?”
Ask:
“Why are we repeatedly learning the same lesson after the
claim has already failed?”
That question changes everything.
If the same denial occurs 50 times, the solution should not
automatically be:
“Work the 50 denials faster.”
Maybe the better question is:
“Why are we allowing number 51 to happen?”
A Simple Example
Imagine a clinic has a recurring documentation-related
denial.
The biller knows about it.
The practice manager knows about it.
Maybe the physician knows about it.
Yet it keeps happening.
Why?
Because the knowledge lives in people's heads.
That's fragile.
When the biller leaves, the knowledge leaves.
When the physician is busy, the knowledge disappears into
the chaos of clinical care.
When a new employee arrives, everyone starts explaining the
same thing again.
This isn't primarily an AI problem.
It's an organizational memory problem.
Technology can help turn individual memory into
institutional memory.
What Should Be Automated?
Automate repetitive work.
Information retrieval.
Data gathering.
Pattern comparison.
Historical searches.
Routine validation.
Prioritization.
Summaries.
But be careful about automating decisions that require
judgment and accountability.
Ambiguous documentation.
Ethical questions.
Clinical context.
Exceptions.
High-risk situations.
The objective is not:
Automate everything.
The objective is:
Automate what machines are good at so humans have more
time for what machines are bad at.
A 30-Day Upstream Billing Experiment
You don't need sophisticated AI to start.
Week 1: Find the recurring pain
Ask your billing team:
“What problem do we keep fixing over and over?”
Not the biggest problem.
The most repetitive one.
Week 2: Walk backward
Find where the problem began.
Scheduling?
Eligibility?
Authorization?
Documentation?
Coding?
Charge capture?
Claim construction?
Submission?
Week 3: Identify the earliest signal
What information existed before the failure?
Could someone reasonably have recognized it?
Week 4: Add one intervention
One rule.
One checkpoint.
One structured data element.
One human review.
One warning.
Then measure.
Did rework fall?
Did staff time fall?
Did clean claims increase?
Did the intervention create new work?
Did staff trust it?
Did humans override it?
If yes, why?
Now you have something valuable.
Evidence.
The Most Dangerous Word in Healthcare Technology
That word is:
“Automatic.”
Automatic coding.
Automatic authorization.
Automatic billing.
Automatic decision-making.
Automatic everything.
Automatic sounds wonderful in a sales presentation.
Until it makes the wrong decision.
The better word is:
appropriate.
Appropriate automation.
Appropriate review.
Appropriate escalation.
Appropriate human intervention.
Healthcare needs judgment about where automation belongs.
The Second Most Dangerous Word
“Optimal.”
Optimal for whom?
Optimal according to what?
Optimal under which assumptions?
Optimal based on what historical data?
Optimal for the payer?
The practice?
The patient?
The biller?
The physician?
The software vendor?
There may not be one optimal answer.
There may be a set of reasonable answers with different
trade-offs.
That is why transparent decision support matters.
What OnnX Is Trying to Build
This is the philosophy behind OnnX.
Not another generic “AI in healthcare” product.
Not another dashboard that gives physicians 47 colors and
calls it intelligence.
Not another system that waits for the denial and then
congratulates itself for finding it.
The idea is simpler:
Move billing intelligence upstream.
Understand the practice.
Understand its historical patterns.
Understand relevant payer and operational context.
Surface potential issues earlier.
Present the human with understandable options.
Allow the human to decide.
Preserve the decision.
Measure the outcome.
Learn.
Repeat.
The technology should become smarter because the practice
becomes more knowledgeable.
The Practice Becomes the Algorithm
Here's where I think things get particularly interesting.
Imagine two clinics using the same platform.
Over time, should their systems behave identically?
Probably not.
Clinic A has one payer mix.
Clinic B has another.
Clinic A has one documentation culture.
Clinic B has another.
Clinic A has a particular historical pattern.
Clinic B has a completely different one.
The software should eventually understand those differences.
The practice doesn't just use the algorithm.
The practice teaches the algorithm.
That is a fundamentally different model.
The Best AI May Know When Not to Answer
Here's a thought I would like more healthcare AI companies
to embrace:
Sometimes the smartest system is the one that says, “I'm
not sure.”
That's not weakness.
That's calibration.
If the evidence is weak, say so.
If the data is incomplete, say so.
If two approaches are reasonable, show both.
If the human needs to decide, escalate.
Trust isn't created by pretending the machine is omniscient.
Trust is created by knowing when the machine should stop
talking.
What Small Physician Practices Actually Need
Not another technology ecosystem.
Not another 90-minute sales demo.
Not another implementation project that takes longer than
the average television series.
They need tools that answer simple questions:
What should I pay attention to?
Why does it matter?
What are my options?
What happens if I choose each one?
What did we learn last time?
Who made the decision?
Did it work?
That's useful.
The Future of Billing May Look Less Like Billing
The irony is that the best billing technology might
eventually become almost invisible.
No giant dashboard.
No endless queue.
No constant alerts.
No heroic denial department.
The system quietly watches for meaningful patterns.
It surfaces what matters.
It explains why.
It asks for judgment when necessary.
It records what happened.
It learns.
Then it gets out of the way.
That's less glamorous than:
“Our revolutionary AI platform transforms healthcare.”
But it may be considerably more useful.
What I Would Measure
If I were a physician owner evaluating an upstream billing
strategy, I would watch:
Preventable denial rate
Not simply total denials.
Time to recognition
How early does the practice identify potential problems?
Rework hours
How much staff time is spent correcting avoidable issues?
Clean-claim performance
How often does the claim move through without rework?
Payer-specific patterns
Where do meaningful differences occur?
Documentation friction
Where does the clinical record create operational problems?
Human override rate
How often do people disagree with the system?
And here's the interesting one:
Learning velocity
How quickly does the practice turn an error into a better
process?
That may become one of the most important metrics in
healthcare technology.
The Healthcare Technology Test
Before buying another AI tool, ask five questions:
1. What happens before the claim?
2. What signal does the system recognize?
3. Can the human understand why it matters?
4. What happens when the human disagrees?
5. Does the organization learn from the outcome?
If a vendor can't answer those questions clearly, you may be
buying automation without intelligence.
Myth Busters
Myth: Every denial is preventable.
No.
Some are not.
The goal is to reduce avoidable ones.
Myth: More AI means better billing.
No.
Better decisions matter more than more AI.
Myth: More data means better decisions.
Not necessarily.
More irrelevant data creates more noise.
Myth: A dashboard creates intelligence.
A dashboard displays information.
Intelligence requires interpretation.
Myth: AI should make the final decision.
Not necessarily.
The appropriate role depends on the decision, risk and
context.
Myth: Practice-specific learning is unnecessary.
Historical behavior can be highly relevant.
Myth: The goal is maximum reimbursement.
No.
The goal should be accurate representation of care and
appropriate reimbursement within applicable rules.
Three Expert Lessons
1. Jenna VanderWeide: Early recognition matters
Jenna's experience demonstrates how rapidly sepsis can
progress and why warning signs may deserve prompt attention. Elizabeth
Dalchand, MD, and Arun Malhotra, MD, emphasized early recognition and treatment
in the AdventHealth account.
Operational lesson: meaningful signals are most
useful before the crisis.
2. Amanda Ireton: Systems become personal
Ireton understood sepsis professionally before experiencing
it as the daughter of someone with sepsis.
Her experience reinforced for her the importance of
organized sepsis recognition and response.
Operational lesson: a process is never merely a
process to the person living with its consequences.
3. Atul Gawande: Technology should augment judgment
Gawande's recent discussion framed AI as a second opinion
rather than a replacement for doctors.
Technology lesson: the machine can expand human
perception without eliminating human responsibility.
Put those three lessons together:
Recognize earlier.
Build systems that remember.
Keep humans accountable.
That is a pretty good starting point for healthcare
technology.
The Bigger Opportunity for Healthcare Founders
If you're building healthcare technology, stop asking:
“Where can we insert AI?”
Ask:
“Where does healthcare repeatedly recognize the problem
too late?”
That question is much more powerful.
Maybe the answer is billing.
Maybe prior authorization.
Maybe referrals.
Maybe documentation.
Maybe medication reconciliation.
Maybe scheduling.
Maybe care transitions.
The recurring pattern is:
Signal exists → signal is fragmented → recognition is
delayed → humans react → organization pays the price.
That is where innovation gets interesting.
The Bigger Opportunity for Physicians
Physicians understand something technology teams sometimes
miss:
Healthcare is full of exceptions.
The patient who doesn't fit the textbook.
The payer that behaves differently from the policy manual.
The workflow that works beautifully at 9 a.m. and collapses
at 4:45 p.m.
The employee who knows a workaround nobody documented.
The patient whose story changes the meaning of the data.
That's why physicians belong in healthcare technology.
Not because every physician needs to become a programmer.
Because physicians understand context.
And context is what machines often lack.
The Bigger Opportunity for Practice Managers
Practice managers see the operational truth.
They see:
The recurring denial.
The staff workaround.
The payer pattern.
The physician frustration.
The patient complaint.
The spreadsheet nobody wants to maintain.
The process everyone knows is broken.
And the problem nobody has time to fix.
That knowledge is an asset.
The question is whether the organization captures it.
If not, the organization keeps paying tuition for the same
lesson.
A New Definition of Efficiency
We normally define efficiency as:
More work with fewer resources.
I would add:
Fewer avoidable problems created in the first place.
Processing 10,000 denials faster is efficiency.
Preventing 1,000 avoidable denials is something else.
It is prevention.
Prevention is often less visible.
There is no heroic queue.
No dramatic dashboard.
No celebratory email:
“Congratulations! We avoided 1,000 problems that never
happened.”
But maybe there should be.
The Most Interesting Question in the Room
At your next revenue-cycle meeting, don't ask:
“How many claims did we process?”
Ask:
“What did we learn?”
Then ask:
“What will we do differently because of what we learned?”
Then ask:
“How will we know it worked?”
If nobody can answer those questions, you're measuring
activity.
Not learning.
Final Thoughts
Jenna VanderWeide was a nurse.
She understood healthcare.
Then she became the patient.
Her husband, Ryan, and their five daughters watched a
routine day become a terrifying fight for her life.
Her condition deteriorated so severely that her daughters
were brought to the hospital because they might have to say goodbye.
She survived.
Months later, the people who cared for her remembered her
even though she could not remember them.
She realized she had been more than a number.
She had been a whole person.
Then there is Amanda Ireton.
She worked in infection prevention.
Then she became the daughter beside her mother's hospital
bed.
She knew the medicine.
But knowing medicine did not make her immune to
helplessness.
These stories remind us why healthcare systems matter.
Not because systems are fascinating.
Because people live inside them.
That is also why I think medical billing deserves a
different conversation.
The claim isn't the beginning.
The denial isn't the beginning.
The dashboard isn't the beginning.
The beginning is the information that existed before those
things happened.
The signal.
The context.
The decision.
The human judgment.
The opportunity.
Maybe the future of medical billing isn't about becoming
better at chasing yesterday's problems.
Maybe it's about becoming better at recognizing tomorrow's
problems before they arrive.
That is what I mean by Upstream Billing Intelligence.
And perhaps the most important question for every
physician-owned practice is not:
“How do we work denials faster?”
It is:
“What are we repeatedly fixing downstream that we should
have recognized upstream?”
That is the question worth answering.
Your Turn
If you are a physician, practice owner, practice manager or
medical biller:
What recurring problem does your practice keep fixing
that you suspect should have been caught earlier?
Tell me in the comments.
Not the politically correct answer.
Not the consultant answer.
The real answer.
Because somewhere inside that recurring annoyance may be the
next important workflow improvement.
And if you know another physician or practice manager who
spends too much time fixing yesterday's problems, share this with them.
Maybe the next breakthrough isn't another AI model.
Maybe it's recognizing the signal sooner.
About the Author
Dr. Daniel Cham is a physician, healthcare strategist
and entrepreneur focused on the intersection of medical practice management,
healthcare technology and medical billing.
As founder of OnnX, Dr. Cham is developing an
approach to Upstream Billing Intelligence for small and medium-sized
physician practices—focused on recognizing potential problems earlier,
understanding practice-specific context, supporting human decision-making and
learning from operational outcomes.
His perspective is grounded in a simple belief:
Healthcare technology should reduce complexity, not
manufacture more of it.
Connect with Dr. Cham on LinkedIn for perspectives on
healthcare operations, medical billing, practice management and healthcare
innovation.
Continue the Conversation
LinkedIn: Connect with Dr.
Daniel Cham
Website: Dr.
Daniel Cham
Spotify: Listen to the
podcast
YouTube: Watch
on YouTube
X: Follow on X
Facebook: Follow on
Facebook
Free LinkedIn resource: Visit the Featured section of
my LinkedIn profile for a free resource on thinking about upstream billing,
operational signals and better practice decisions.
Disclaimer
This article is intended for general educational and informational
purposes. It does not constitute medical, legal, coding, compliance,
reimbursement or financial advice. Healthcare professionals and organizations
should evaluate their own circumstances and consult appropriately qualified
professionals for specific guidance.
References
1. AdventHealth — “Sepsis nearly took a mother's life.
Now, she wants others to know the warning signs.”
Published September 16, 2026. The account documents Jenna VanderWeide's sepsis
experience, her husband Ryan VanderWeide, their five daughters and her long
recovery.
Read
the Jenna VanderWeide story
2. AdventHealth — “Turning an unimaginable loss into a
lifesaving mission.”
Published September 21, 2026. The story describes Amanda Ireton's experience
losing her mother to sepsis and her subsequent advocacy for sepsis awareness
and recognition.
3. CDC — “Sepsis Burden.”
Updated August 17, 2026. The CDC reports approximately 1.7 million adult sepsis
cases annually in the United States and at least 350,000 adult deaths or
hospice discharges.
4. Atul Gawande — “What's Worth Living For?”
September 14, 2026, 10% Happier with Dan Harris. The conversation covers
aging, difficult diagnoses, what matters to patients and the role of AI as a
second opinion rather than a replacement for physicians.
Listen
to the Atul Gawande conversation
One Last Thought
Healthcare has no shortage of information.
It has no shortage of software.
It has no shortage of alerts.
What it may need more of is better recognition.
Because the most expensive problem may not be the problem we
can see.
It may be the signal we saw—
and didn't understand soon enough.
#MedicalBilling #HealthcareManagement #PracticeManagement
#RevenueCycleManagement #HealthcareInnovation #UpstreamBilling
#HealthcareTechnology #PhysicianPractice #MedicalCoding #PhysicianEntrepreneur
#PatientSafety #HealthTech #RevenueCycle #OnnX
No comments:
Post a Comment