The Final Words of a Father—and What They Reveal About the Patient We Lose in the Data
“It’s not supposed to be happening.” — Joan
M. Griffin, PhD, Mayo Clinic researcher studying lucidity in dementia.
A final moment of clarity in dementia raises a much
bigger healthcare question: What happens when we capture the data—but lose the
context?
The day before her father died, a caregiver came to his
home.
He was rarely awake.
He no longer spoke.
So she did what families often do when medicine has run out
of easy answers.
She kept talking to him.
She leaned into his view.
“Hi, Dad!”
He looked at her.
And answered:
“Hi, Honey!”
It was the last time he spoke.
There is something almost impossible to process about that
moment.
Not because it proves anything supernatural.
Not because dementia suddenly disappeared.
Not because scientists have discovered a cure.
They haven't.
Researchers are still trying to understand why some people
with advanced dementia experience unexpected periods of clarity, communication,
recognition or apparent awareness. Scientists call these episodes paradoxical
lucidity, or terminal lucidity when they occur near the end of life.
The mechanisms remain uncertain.
But the story exposes something much bigger.
The daughter heard two words.
The researchers heard a phenomenon.
A clinician might hear a neurological event.
A stranger might hear nothing more than a sentence.
But the daughter heard her father.
She knew the baseline.
She knew what had changed.
She knew why those two ordinary words were extraordinary.
And that is where this story becomes unexpectedly relevant
to modern healthcare.
Because healthcare has a similar problem.
We are exceptionally good at collecting information.
We are increasingly good at processing information.
We are getting very good at generating information.
But we are surprisingly bad at preserving context.
And without context, information can become almost useless.
The Patient Is Not the Data
Let's start with an uncomfortable proposition.
The patient is not the chart.
The diagnosis isn't the patient.
The CPT code isn't the encounter.
The authorization number isn't the clinical reasoning.
The claim isn't the care.
The denial isn't the explanation.
And the dashboard isn't reality.
Yet somewhere between the exam room and the payment system,
we repeatedly act as though these things are interchangeable.
They're not.
A physician sees a patient.
The patient tells a story.
The physician interprets the story.
A clinical decision is made.
Documentation is created.
Administrative information is added.
The encounter enters a series of systems.
Then the information gets copied, transformed, compressed,
translated, transmitted, coded, validated and eventually turned into a claim.
At every step, something can disappear.
Sometimes it is a field.
Sometimes it is a timestamp.
Sometimes it is an authorization.
Sometimes it is a referral.
Sometimes it is clinical reasoning.
Sometimes it is the relationship between two pieces of
information.
And sometimes what disappears is the thing that mattered
most:
Why.
Healthcare Doesn't Have a Data Shortage
This is where I become deliberately contrarian.
Healthcare does not have a data shortage.
We have more data than almost anyone could reasonably
consume.
Laboratories.
Imaging.
Notes.
Messages.
Claims.
Eligibility responses.
Authorizations.
Referrals.
Medication histories.
Problem lists.
Payer policies.
Portal messages.
Phone calls.
Faxed documents.
Scanned documents.
Spreadsheets.
Dashboards.
And, increasingly, AI-generated summaries.
If data alone solved healthcare's administrative problems,
we should have solved them years ago.
Instead, we keep adding more.
More systems.
More portals.
More integrations.
More dashboards.
More automation.
More AI.
And somehow someone is still saying:
“I know the information is somewhere.”
That sentence should terrify every healthcare technology
founder.
Because it means the data exists.
But the context doesn't travel with it.
Here's the Strange Part
We have built a healthcare system in which a highly trained
physician can spend twenty minutes understanding a patient...
...and a billing department can later spend forty minutes
trying to reconstruct what happened.
That is backwards.
The person closest to the clinical decision often has the
richest context.
The person downstream may have the least.
Yet the downstream person is frequently expected to make the
information work.
So what happens?
Humans become middleware.
The front desk becomes middleware.
The medical assistant becomes middleware.
The nurse becomes middleware.
The biller becomes middleware.
The practice manager becomes middleware.
Everyone moves information from one disconnected place to
another.
And then we call it a workflow.
Maybe some workflows are simply systems compensating for
missing context.
Meet the World's Most Expensive API
Her name might be Linda.
Or Maria.
Or Steve.
Every medical practice has one.
She's been there for 17 years.
She knows which payer portal actually works.
She knows which authorization form the payer really wants.
She knows which physician forgets which field.
She remembers that one strange patient from 2018.
She knows that if the payer says “pending,” it might mean
three completely different things.
She knows where the missing document probably is.
She knows who to call.
She knows what to say.
She knows when the payer representative is giving her the
standard answer and when something is actually wrong.
She is, essentially, an undocumented API.
And when she retires, the organization discovers something
horrifying:
She was the infrastructure.
That isn't a technology strategy.
That's institutional knowledge trapped inside one human
being.
The answer isn't to replace her.
The answer is to preserve the knowledge she shouldn't
have to repeatedly recreate.
The “Why” Is Usually Missing
Ask a billing team:
Why did this claim deny?
You'll probably get an answer.
Authorization.
Eligibility.
Medical necessity.
Coding.
Modifier.
Coverage.
Documentation.
Timely filing.
Fine.
Now ask:
Why did the underlying problem happen?
That question changes everything.
Maybe the insurance changed before the appointment.
Maybe eligibility was checked against the wrong plan.
Maybe an authorization was obtained but never linked to the
encounter.
Maybe the authorization covered one service but the claim
represented another.
Maybe a referral was required.
Maybe the payer changed its rule.
Maybe the physician documented the clinical reasoning, but
that context never reached the administrative workflow.
Maybe the payer made an error.
Maybe the practice made an error.
Maybe nobody made an error.
Maybe the information was simply fragmented.
That's why I believe the denial itself is often less
interesting than the chain of events that produced it.
The denial isn't necessarily the problem.
The denial is the clue.
Healthcare Loves to Clean Up Messes
This is one of the industry's stranger habits.
We create friction.
Then we build an industry around managing the friction.
A claim rejects.
Someone works the rejection.
An authorization is missing.
Someone calls.
A document can't be found.
Someone searches.
A payer requests information.
Someone uploads it.
A patient doesn't understand the bill.
Someone explains it.
The same patient calls again.
Someone explains it again.
At some point we congratulate ourselves because the workflow
has been “optimized.”
Optimized?
We just got faster at cleaning up the mess.
That's not necessarily innovation.
Sometimes it's high-speed housekeeping.
Prior Authorization Is a Perfect Example
The latest AMA physician survey provides an uncomfortable
picture.
Physicians report completing an average of 40 prior
authorization requests per week.
Those requests consume an average of 13 hours of
physician and staff time every week.
More than nine in ten physicians, 94%, say prior
authorization contributes to burnout.
And 74% report that prior authorization denials have
increased over the past five years.
The AMA survey also found that 40% of physicians employ
staff dedicated exclusively to prior authorization tasks.
Forty requests.
Thirteen hours.
Dedicated staff.
And this is one administrative process.
Not healthcare.
One process.
So perhaps the question isn't:
“How can we process prior authorization faster?”
Perhaps the more important question is:
“Why does the system require this much human
reconstruction in the first place?”
That's a different problem.
And a much more interesting one.
The Hidden Cost Isn't Just Money
We love measuring administrative waste in dollars.
We should.
But there's another currency:
attention.
A physician's attention.
A nurse's attention.
A biller's attention.
A practice manager's attention.
A patient's attention.
A caregiver's attention.
Every unnecessary phone call spends some of it.
Every duplicate entry spends some.
Every manual lookup spends some.
Every “can you send that again?” spends some.
Every portal login spends some.
One interruption isn't catastrophic.
A thousand aren't trivial.
That's how healthcare burnout works.
Not necessarily through one enormous event.
Through thousands of tiny demands that accumulate until
someone finally says:
“I can't do this anymore.”
The Dementia Story Gives Us a Clue
Go back to the father.
The daughter says:
“Hi, Dad.”
He says:
“Hi, Honey.”
Those words by themselves are ordinary.
The context makes them extraordinary.
That is the point.
Context changes meaning.
Without context, the words are merely data.
With context, they become a moment.
Healthcare understands this intuitively.
Doctors know that a lab value means different things
depending on the patient.
A symptom means different things depending on the history.
A medication means different things depending on the
indication.
A diagnosis means different things depending on the
encounter.
Yet our administrative systems often flatten all of that
complexity into discrete fields.
Then we act surprised when the downstream system doesn't
understand the story.
What Joan Griffin's Research Really Makes Me Think About
Joan M. Griffin, PhD, a Mayo Clinic researcher who studies
lucidity in dementia from the caregiver perspective, told the Washington Post:
“It’s not supposed to be happening.”
That sentence is haunting.
But it also reveals something important about science.
When reality doesn't fit the model, the answer isn't
necessarily to ignore reality.
Sometimes the model is incomplete.
Healthcare technology needs the same humility.
When the same claim keeps denying...
When the same authorization keeps failing...
When staff repeatedly search for the same information...
When the same patient has to explain the same thing...
Maybe the people aren't the problem.
Maybe the model is incomplete.
Another Remarkable Observation
The Washington Post reported that Griffin has seen
caregivers change how they interact with people with dementia after witnessing
moments of lucidity.
Some physicians, she said, began speaking directly to the
patient rather than only to the caregiver, recognizing that the patient
might be processing information even when communication appeared limited.
That is a profound shift.
The person's apparent inability to communicate had been
interpreted as absence.
Then an unexpected moment challenged the assumption.
And the behavior changed.
There is a lesson here for healthcare technology:
What you cannot observe easily is not necessarily what
does not exist.
The same principle applies to context.
If your billing system can't see the relationship between
eligibility, authorization, documentation and the eventual claim, that doesn't
mean the relationship isn't there.
It means your architecture isn't preserving it.
This Is the Data Quality Problem Nobody Wants to Own
Everyone says they want clean data.
But “clean data” is usually interpreted as:
No missing fields.
No duplicate records.
Valid formatting.
Consistent codes.
Correct dates.
That's necessary.
But it isn't enough.
You can have perfectly formatted data that is completely
disconnected from its meaning.
Imagine a spreadsheet with:
Patient ID.
Date.
Procedure.
Diagnosis.
Payer.
Authorization number.
Everything is technically clean.
But nobody knows whether the authorization actually
corresponds to the procedure.
The data is clean.
The context is broken.
That's a much more subtle problem.
And potentially a much more expensive one.
Clean Data vs. Meaningful Data
I would separate the two.
Clean data
Is accurate, structured and valid.
Contextual data
Preserves relationships, timing, provenance and meaning.
We need both.
The next generation of healthcare infrastructure should be
designed around both.
Because AI cannot reason reliably about context that has
already been discarded.
This Is Where AI Gets Interesting
The healthcare industry is rushing toward AI.
That's understandable.
AI can summarize.
Classify.
Extract.
Predict.
Route.
Flag.
Compare.
Generate.
But here's my contrarian warning:
AI does not magically repair bad context.
It can process bad context faster.
It can summarize incomplete context beautifully.
It can produce an extremely confident answer from
information that should never have been trusted.
That's not intelligence.
That's automated ambiguity.
And healthcare cannot afford to confuse the two.
The Real AI Question
Instead of asking:
“Which AI model should we use?”
Ask:
“What information will the model actually receive?”
Then:
“Where did that information come from?”
Then:
“When was it verified?”
Then:
“What context surrounds it?”
Then:
“What happens when two sources disagree?”
Then:
“What does the system do when the answer is unknown?”
Those questions are considerably less glamorous than talking
about agents.
They are also more important.
AI Should Preserve Context, Not Invent It
This principle matters enormously in healthcare.
If eligibility is unknown:
Don't guess.
If authorization is uncertain:
Flag it.
If documentation conflicts:
Surface the conflict.
If the clinical record doesn't support a conclusion:
Don't manufacture support.
If the system infers something:
Distinguish inference from fact.
If a human needs to decide:
Escalate to the human.
The smartest system may sometimes be the one that says:
“I don't know. Here's what needs to be checked.”
That is not weakness.
That's governance.
The Opportunity Before the AI
This is where my thinking about OnnX begins.
I don't believe the biggest opportunity in healthcare
billing is simply building another AI layer on top of the existing revenue
cycle.
There are already plenty of tools trying to make downstream
work smarter.
My question is more fundamental:
What if we improve the information before it becomes a
billing problem?
Healthcare billing is, at its core, a data-quality problem.
Not merely a tooling problem.
And certainly not merely a staffing problem.
Most of the Problem Starts Upstream
Think about a claim.
By the time it reaches the billing queue, a tremendous
amount has already happened.
The patient was scheduled.
Insurance was identified.
Eligibility may have been checked.
Referral requirements may have existed.
Authorization requirements may have existed.
Clinical information was collected.
A service was selected.
Documentation was created.
Coding decisions were made.
Information was transmitted.
By the time a claim denies, the industry often focuses on
the last step.
But the last step may only be where the problem becomes
visible.
Visibility is not causation.
That's a distinction healthcare should take much more
seriously.
The OnnX Concept
The operating philosophy behind OnnX is:
Capture → Structure → Preserve → Propagate → Act → Learn
Capture
Capture relevant information as early as possible.
Structure
Turn it into usable, consistent information.
Preserve
Keep the relationship between information and the encounter.
Propagate
Carry relevant context forward instead of forcing people to
recreate it.
Act
Identify exceptions before they become expensive downstream
problems.
Learn
Use outcomes to improve the system.
The idea isn't complicated.
The implementation is.
Healthcare has a remarkable talent for turning simple ideas
into complicated workflows.
The Canonical Context Record
One concept I keep coming back to is the canonical
context record.
Not another giant chart.
Not another database where everything goes to die.
A structured representation of the relevant facts and
relationships surrounding an encounter.
For example:
Patient
↓
Coverage
↓
Eligibility
↓
Referral
↓
Authorization
↓
Clinical indication
↓
Service
↓
Documentation
↓
Coding
↓
Claim
↓
Payer response
↓
Payment / denial
The value isn't simply storing each item.
The value is preserving the relationships between them.
That's the missing layer.
Imagine the Difference
Today:
“Why did this claim deny?”
Someone investigates.
Tomorrow:
“The claim denied because the authorization covered
procedure A, while the submitted claim represented procedure B. Eligibility was
verified four days before service. The mismatch originated at the
authorization-to-claim transition.”
That's a different system.
Not because the AI is magical.
Because the context survived the journey.
The Metric I Want to See
We measure denial rates.
Clean claims.
A/R.
Days to payment.
Collection rates.
Useful.
But I would add another metric:
Context Loss Rate
How often does your organization have to rediscover
information that should already be available?
Ask:
How often does staff search for an authorization?
How often does someone re-enter patient information?
How often does someone call the payer because the answer
wasn't preserved?
How often does a physician get asked for information already
documented?
How often does the patient repeat information?
How often does someone say:
“I know it's somewhere.”
That's a context-loss event.
Start counting them.
You may discover that your biggest revenue-cycle problem
isn't your billing software.
It's the number of times your employees become detectives.
The Sherlock Holmes Problem
A practice should not require a detective to submit a claim.
Yet that's what happens.
“Where is the authorization?”
“Who verified eligibility?”
“When did we verify it?”
“Was the referral required?”
“Which payer representative confirmed it?”
“Did the patient change plans?”
“Did anyone upload the document?”
“Where's the clinical note?”
“Which version?”
“What did the payer actually ask for?”
This is not sophisticated work.
It is reconstruction work.
And reconstruction is expensive.
The Human Middleware Tax
Here's another metric I'd like healthcare organizations to
think about:
Human Middleware Tax.
How much employee time is spent moving information between
systems because the systems themselves don't carry the context?
That might be:
Copying.
Pasting.
Calling.
Faxing.
Uploading.
Downloading.
Searching.
Re-entering.
Reconciling.
Confirming.
Repeating.
The employee is doing work.
But the organization isn't necessarily creating value.
It is compensating for architectural limitations.
This Is Not an Argument Against Humans
Quite the opposite.
The goal shouldn't be:
Remove the humans.
It should be:
Stop wasting the humans.
A great biller should be solving exceptions.
A great practice manager should be improving operations.
A physician should be practicing medicine.
A nurse should be caring for patients.
An experienced employee should be handling the complicated
cases where judgment actually matters.
Not searching three portals for a PDF that somebody already
uploaded.
The Janice Test
Here's a test for any healthcare technology company.
Take your product into a clinic.
Find the person who knows everything.
Ask them to demonstrate the current workflow.
Then ask:
“What do you do when this goes wrong?”
They'll show you.
Now ask:
“How many times a week?”
Then:
“How long does it take?”
Then:
“What information are you looking for?”
Then:
“Why isn't that information already available?”
That final question is where product discovery gets
interesting.
Stop Asking, “Would You Buy It?”
Healthcare founders love asking:
“Would this solve a problem for you?”
Of course.
The person wants to be nice.
Instead ask:
“Show me the last time this happened.”
Then:
“What did you do next?”
Then:
“Who else got involved?”
Then:
“How long did it take?”
Then:
“What happens if you're out sick?”
Then:
“How do you know the problem is fixed?”
Now you're getting somewhere.
The Practice Manager's Conversation
Imagine this:
Biller: “The claim denied.”
Manager: “Why?”
Biller: “Authorization.”
Manager: “Didn't we have one?”
Biller: “I think so.”
Manager: “Where is it?”
Biller: “Payer portal.”
Manager: “Which portal?”
Biller: “Let me check.”
That's not a billing workflow.
That's an escape room.
And someone is paying for the privilege.
The Real Definition of Automation
Automation should not mean:
“We moved the work faster.”
It should mean:
“We eliminated unnecessary work.”
Those are different.
If a person previously searched five systems for 20 minutes
and an AI searches five systems in 30 seconds, congratulations.
You've built a faster searcher.
But if the information had been structured correctly at the
beginning, perhaps nobody needed to search at all.
That is the more interesting innovation.
Healthcare Needs Fewer Detective Stories
The best healthcare infrastructure should make the boring
things boring.
Eligibility should be boring.
Authorization status should be boring.
Documentation availability should be boring.
Claim context should be boring.
Payment reconciliation should be boring.
The interesting work should be:
Clinical judgment.
Complex patients.
Difficult diagnoses.
New treatments.
Research.
Human relationships.
Not:
“Where did that authorization PDF go?”
The Ethical Line
There is another reason this matters.
When AI enters healthcare administration, convenience cannot
become the only objective.
A system shouldn't fabricate missing documentation.
It shouldn't silently infer facts and represent them as
verified.
It shouldn't conceal uncertainty.
It shouldn't turn a probability into a fact.
It shouldn't make a physician responsible for an
AI-generated conclusion without appropriate visibility.
And it shouldn't encourage practices to treat payer rules as
obstacles to be manipulated.
The objective should be:
accurate information, transparent reasoning, traceable
actions and appropriate human oversight.
That isn't bureaucracy.
That's trust.
The Patient Is Paying for the Fragmentation
We often discuss administrative burden as though it belongs
to the healthcare organization.
It doesn't.
Patients experience it.
They wait.
They call.
They repeat information.
They receive confusing bills.
They wonder why one doctor knows something another doesn't.
They wait for authorizations.
They delay treatment.
They become frustrated.
And eventually they say something familiar:
“Why can't you people just talk to each other?”
That may be the most accurate healthcare interoperability
assessment ever written.
Independent Practices Feel This Differently
A large health system may have departments dedicated to
fixing administrative problems.
A small physician-owned practice may have:
One physician.
A few staff.
One practice manager.
One biller.
And a phone that never seems to stop ringing.
Every unnecessary administrative task has an opportunity
cost.
A physician fighting an authorization isn't seeing a
patient.
A biller searching for an old document isn't working another
claim.
A manager reconstructing a workflow isn't improving the
practice.
Scale doesn't eliminate friction.
It simply determines how much friction you can afford.
What I Would Fix First
If I were sitting with an independent practice tomorrow, I
wouldn't start with AI.
I'd start with one service.
One payer.
One workflow.
One recurring problem.
Then I'd trace it.
Appointment.
Eligibility.
Authorization.
Clinical encounter.
Documentation.
Coding.
Claim.
Payment.
Denial.
And I'd ask:
Where was the first preventable information failure?
Not where did the denial occur.
Where did the problem begin?
That distinction can change an entire product strategy.
The Five Questions
For every recurring administrative problem, ask:
1. What happened?
Document the event.
2. Why did it happen?
Find the upstream cause.
3. What information existed?
Identify the evidence.
4. Where did the context disappear?
Find the handoff.
5. How do we prevent reconstruction next time?
That's where technology belongs.
The Future of Healthcare AI May Be Boring
This may sound strange coming from someone building an AI
healthcare company.
But I don't want AI to feel impressive.
I want it to feel boring.
I want the practice manager to say:
“Wait. I didn't have to call them?”
I want the biller to say:
“The authorization was already attached?”
I want the physician to say:
“Why did I get this request? The information was already
there.”
I want the patient to say:
“I didn't have to explain that again?”
That's success.
Not a flashy demo.
Not a futuristic robot.
Not a 47-slide investor deck.
Just fewer unnecessary problems.
The Best AI May Be the AI Nobody Notices
A patient doesn't care that an AI model processed their
eligibility data.
They care that their appointment wasn't delayed.
A physician doesn't care that an algorithm classified a
payer rule.
They care that the patient got the treatment.
A biller doesn't care that a model used a sophisticated
embedding architecture.
They care that the claim didn't require three hours of
detective work.
The technology should disappear into the workflow.
The outcome should be visible.
Three Lessons From the Dementia Story
The first is simple:
1. Absence of visible communication is not necessarily
absence of awareness.
Healthcare should be careful about confusing what a system
cannot observe with what does not exist.
2. Context changes meaning.
Two ordinary words—“Hi, Honey”—can become extraordinary when
someone knows the story surrounding them.
3. Systems should preserve what humans need to
understand.
The daughter didn't need more data.
She needed the context that made the data meaningful.
Healthcare needs the same thing.
A Different Way to Think About “Patient-Centered”
Patient-centered healthcare shouldn't simply mean:
Be nice to the patient.
It should mean:
Don't make the patient repeatedly compensate for the
system's inability to preserve information.
That is harder.
But it is measurable.
How many times does a patient repeat information?
How many times does a patient call?
How many administrative delays occur?
How many bills require explanation?
How many times does a patient become the messenger between
two parts of the healthcare system?
Those are patient-experience metrics too.
What Healthcare Founders Should Stop Building
Maybe we should stop building:
Another dashboard nobody checks.
Another portal nobody likes.
Another chatbot that doesn't know the patient's context.
Another AI summarizer that creates another document.
Another workflow tool that requires staff to manually feed
it information.
Another “single pane of glass” that somehow creates a second
pane of glass.
Healthcare doesn't need more software for software's sake.
It needs better information architecture.
What We Should Build Instead
Systems that:
Capture information once.
Preserve its meaning.
Maintain provenance.
Carry context forward.
Detect conflicts.
Surface exceptions.
Ask humans when uncertainty matters.
Learn from outcomes.
That sounds less sexy.
Good.
Sexy software makes great demos.
Reliable infrastructure makes great businesses.
The OnnX Thesis in One Sentence
Healthcare billing should become more deterministic
because the information entering the revenue cycle becomes more structured,
contextual and trustworthy.
Not because we hire more people to chase problems.
Not because we throw AI at denials.
Not because we build another RCM dashboard.
Because we reduce the number of things that go wrong
upstream.
The Denial Is Still Useful
Remember:
The denial isn't the enemy.
It is information.
A denial tells you that something in the system didn't line
up.
Maybe the payer's rule.
Maybe the documentation.
Maybe the authorization.
Maybe eligibility.
Maybe coding.
Maybe timing.
Maybe the system.
The question is not:
“How quickly can we make this denial disappear?”
The better question is:
“What is this denial trying to teach us?”
That's where revenue cycle becomes more than collections.
It becomes a feedback system.
Capture Once. Preserve Context.
This is ultimately the idea.
A physician shouldn't have to recreate information.
A nurse shouldn't have to recreate information.
A biller shouldn't have to recreate information.
A patient shouldn't have to recreate information.
A practice manager shouldn't have to recreate information.
Capture once.
Preserve context.
Let the information travel with the encounter.
That is a very different vision of healthcare
infrastructure.
The Father at the End of the Story
The Washington Post reported that after experiencing these
unexpected moments, some caregivers changed how they interacted with people
with dementia.
They began to reconsider what they assumed the patient could
understand.
They spoke directly to them.
They paid closer attention.
They stopped equating silence with absence.
That is the lesson I want healthcare technology to take
seriously.
Because our systems can make a similar mistake.
When information isn't visible, we assume it isn't there.
When context isn't structured, we assume it doesn't matter.
When the workflow breaks, we blame the person downstream.
Maybe we should question the system upstream.
“Hi, Dad.”
“Hi, Honey.”
Two ordinary sentences.
An extraordinary moment.
And perhaps an uncomfortable lesson for healthcare:
The thing we fail to capture is not necessarily the thing
that isn't there.
Sometimes it is simply the thing our system wasn't designed
to preserve.
And that's the problem I think we should be solving.
Not:
How do we collect more data?
But:
How do we stop losing the meaning of the data we already
have?
Because the patient was never just the data.
And the claim was never just the claim.
The story was there all along.
We just didn't carry it forward.
My Challenge to Physicians and Clinic Owners
Tomorrow, ask your team one question:
“What do you have to explain more than once?”
Don't ask what annoys them.
Ask what gets lost.
Eligibility.
Authorization.
Referral information.
Clinical reasoning.
Documentation.
Coding rationale.
Payer instructions.
Patient financial information.
You may discover that the biggest problem in your practice
isn't something you need to add.
It is something you need to stop losing.
My Challenge to Healthcare Founders
Stop asking:
“Where can I put AI?”
Ask:
“Where is context disappearing?”
Then:
“What does that loss cost?”
Then:
“Who is compensating for it?”
Then:
“Could the system preserve it earlier?”
That is where interesting companies begin.
My Challenge to Healthcare Leaders
Don't only measure the work your employees complete.
Measure the work your systems force them to do.
Those are not the same thing.
A brilliant employee can make a broken workflow look
functional.
That doesn't mean the workflow is good.
It means the employee is good.
Don't accidentally turn your best people into human
error-correction systems.
The Question I Want to Leave You With
What if the next major breakthrough in healthcare isn't
more data?
What if it is less data loss?
What if the next great healthcare AI company doesn't win
because its model is slightly smarter?
What if it wins because its system knows:
where the information came from,
what it means,
when it was verified,
what it is connected to,
and what should happen next?
That is a much less glamorous pitch.
But perhaps a much more useful one.
Continue the Conversation
I'm interested in hearing from physicians, billers, practice
managers, healthcare operators and founders:
Where does your organization lose context?
At scheduling?
Eligibility?
Prior authorization?
Documentation?
Coding?
Billing?
Claims?
Denials?
Patient communication?
Or somewhere between all of them?
Tell me what you've seen.
And if this article made you think of someone who has spent
years quietly working around a broken healthcare workflow, send it to them.
Sometimes the person closest to the problem already knows
the solution.
They just haven't been asked the right question.
Connect with Dr.
Daniel Cham on LinkedIn
About the Author
Dr. Daniel Cham is a physician, medical consultant
and healthcare entrepreneur focused on the intersection of medicine,
healthcare operations, medical technology and medical billing.
He is the founder of OnnX, an emerging healthcare
technology venture focused on helping physician-owned practices rethink how
clinical and administrative information moves through the revenue cycle.
His central thesis is simple:
Healthcare billing is a data-quality problem, not merely
a tooling problem.
Dr. Cham writes about healthcare operations, physician
entrepreneurship, medical technology, AI, medical billing and the practical
realities of building better healthcare infrastructure.
Disclaimer
This article is provided for educational and
informational purposes only. It is not medical, legal, coding, compliance,
reimbursement, financial or other professional advice.
Healthcare requirements vary by patient, payer, contract,
organization and jurisdiction. Appropriate professional guidance should be
obtained for specific clinical, legal, reimbursement, privacy, security or
compliance decisions.
A Note on the Human Story
The father and daughter described in the opening Washington
Post account were not publicly identified by full name, and I have
intentionally not invented names for them.
That matters.
The story belongs to a real family.
The researchers can be identified because their professional
roles and statements were publicly reported. The caregiver accounts themselves
remain anonymous in the published reporting.
The second account includes a man named Joe, who was
identified in the caregiver's account as her husband. The Washington Post did
not provide his surname.
That distinction is worth preserving.
Accuracy is part of context too.
Sources & Further Reading
Margaret Hetherman, The Washington Post — “Some dementia
patients grow lucid near the end. Scientists hope it’s a clue.” — September 25,
2026.
The primary current-week source for the human stories, Joan
Griffin's research and the LEAD Network conference.
American Medical Association — 2025 Prior Authorization
Physician Survey, published May 2026.
The current AMA survey reports an average of 40 prior
authorization requests per physician per week, 13 hours of physician/staff time
per week, 94% reporting a contribution to burnout, and 74% reporting increased
denials over five years.
Duke University — Heather Whitson and the Lucidity in
Alzheimer’s and Dementia Network.
Useful background on the research effort studying unexpected
lucidity in dementia.
Featured Resource
For physicians and clinic owners interested in the broader
problem of upstream data quality, context preservation and deterministic
revenue-cycle workflows, visit the Featured section of my LinkedIn
profile.
No hype.
No “AI will replace everyone.”
No magic wand.
Just a different question:
What if we fixed the information before we had to fix the
claim?
Final Thought
The healthcare industry has spent decades becoming better at
processing information.
The next decade may require us to become much better at preserving
meaning.
Because somewhere between the patient and the claim,
something important gets lost.
And sometimes the most important thing in the entire system
is the thing that never made it into the database.
The patient.
References
- Hetherman,
Margaret. “Some dementia patients grow lucid near the end. Scientists hope
it’s a clue.” The Washington Post, September 25, 2026.
Primary source for the opening story, Joan M. Griffin’s research, the “Hi, Dad / Hi, Honey” account, and the current discussion of paradoxical and terminal lucidity. Read the Washington Post article - Duke
University School of Medicine. “When the fog lifts: Pioneering a new
frontier in Alzheimer’s and dementia research.” August 11, 2026.
Provides current background on Heather Whitson’s research and the scientific questions surrounding unexpected periods of lucidity in advanced dementia. Read the Duke University research article - American
Medical Association. “AMA survey: Prior authorization reform pledge falls
short with physicians.” 2026.
Source for the current physician survey figures cited in the article, including an average of 40 prior authorizations per week, 13 hours of physician/staff time, 94% reporting a contribution to burnout, and 74% reporting increased denials over five years. Read the AMA survey
#Healthcare #MedicalBilling #HealthcareAI #HealthTech
#DigitalHealth #HealthcareInnovation #PhysicianEntrepreneur #MedicalTechnology
#RevenueCycleManagement #IndependentPractice #PracticeManagement
#HealthcareOperations #PatientExperience #ClinicalDocumentation #AIinHealthcare
#HealthcareLeadership #Dementia #PatientCenteredCare #Physicians
#HealthcareTransformation
