Glenn Kunsman, DO — Founder, Medico PC | Founder, Medico Disability Services
I have spent sixteen years in rooms with people who were genuinely disabled.
Not people claiming to be disabled. People whose bodies had stopped cooperating with the life they were trying to live — whose backs had given out, whose joints had deteriorated past the point of recovery, whose pain had quietly reorganized every hour of every day around what could not be fixed.
I saw them as a physician performing consultative examinations for the Social Security Administration. Medico PC has been providing those examinations for the State of Arizona since 2010. In that time I evaluated thousands of claimants — reviewed their records, examined their function, documented my findings for SSA adjudication.
And in that time I watched something happen over and over that I could not stop from inside the examination room.
The wrong documents were driving the outcomes.
The Documentation Problem
Not fraudulent records. Not fabricated claims. Real medical records from real treating providers — records that simply did not reflect the functional reality of the person sitting across from me.
SSA adjudication is not a clinical process. It is an evidentiary one. Adjudicators evaluate whether the record supports a specific legal and functional threshold. A claimant can be genuinely, severely disabled and lose because the documentation arriving at the decision point does not tell that story in the language SSA requires.
I watched that happen across three phases that almost every claimant’s record moves through. The acute phase — dense records from the moment of diagnosis or crisis — that captures the worst moment but says nothing about what life looks like two years later. The treatment phase — documentation of everything attempted — that implies progress even when function never meaningfully recovered. The ongoing care phase — sparse notes, stable medication lists — that reads to an adjudicator as a managed, functional patient rather than someone whose ceiling has permanently lowered.
Each phase misrepresents the disability when read alone. Together, without a framework connecting them, they still miss the functional reality that lives between the appointments — the hours, days, and weeks no provider ever documents.
I understood the problem clearly for years before I had any idea what to do about it.
The Failed Attempts
I tried to solve it the way you solve a business problem: hire people who can build what you cannot.
I engaged technology firms. Explained the problem. Walked through the clinical logic. Watched them build approaches that failed — and failed for reasons I did not fully understand until I understood the problem more precisely myself. The failures were not their fault. I had not yet found the right articulation of what the system needed to do. You cannot outsource a solution to a problem you have not fully defined.
The early conceptual attempts were instructive in their own way. The intuitive approach — read the clinical narrative, assign meaning, identify diagnoses, synthesize a conclusion — was essentially trying to automate what a reviewing physician does. It failed completely. Clinical reasoning depends on implicit judgment that does not transfer to code. Every attempt to map narrative meaning to diagnostic conclusions broke down across the variability of real medical records.
What finally worked was smaller and more disciplined. Single fact identification. Discrete, atomic extraction — this value, this date, this source, this context. No interpretation layer. No concept mappers attempting to bridge clinical meaning. Just facts, and then signal detection built on deviation from expected parameters. Abnormal findings as the primary evidentiary driver, normal findings as pertinent negatives. The strict prohibition on any layer that tried to interpret rather than extract turned out to be the discipline the whole system depended on.
I knew that. I had worked it out conceptually. What I could not do was build it.
When the Technology Finally Caught Up
My last serious code was .NET 4.
The gap between that and building a functioning clinical document analysis system is not a gap you close by brushing up on syntax. It is a different world. And in 2024 I could not have crossed it.
In 2025 I could.
AI-assisted development changed the equation in a way I did not anticipate and still find genuinely remarkable. Not because it writes code autonomously — it does not, not reliably, not for something this specific. But because it created a platform where deep domain expertise became the primary driver of development rather than a requirement to be translated through an engineering team and lost in the translation.
Sixteen years of understanding what SSA actually evaluates. Years of working through what medical records actually contain versus what they appear to contain. A precise understanding of where the evidentiary gaps live and why they form. That knowledge, which had no development home before, suddenly had one.
The project started in 2025. The failures were real and the iterations were many. The combination that held up — fact extraction, abnormal signal analysis, strict prohibition on interpretation layers — was not obvious from the outside. It required being wrong about the approach multiple times before understanding why the simpler, more disciplined framework was the one that worked at scale.
The Work Continues
The technology landscape this work sits inside is moving faster than any stable description can track. Capabilities available today did not exist eighteen months ago. What is a hard problem now will be a different problem in another eighteen months — differently shaped, with different failure modes, different opportunities, and different risks of getting it wrong.
Staying current is not optional. It is the work. Domain expertise without technical awareness drifts. Technical capability without domain grounding builds the wrong thing precisely. The combination is what makes the difference — and maintaining it in a rapidly evolving environment is an ongoing commitment, not a solved problem.
Medico PC continues to provide consultative examinations for SSA. That work grounds everything else — direct, ongoing contact with the evidentiary problem, the claimants carrying incomplete records, the functional realities that documentation does not capture.
Medico Disability Services was built from what sixteen years of that work revealed. Physician-led medical record review and documentation support for SSA claimants and the attorneys who represent them.
The technology finally caught up to the problem.
It took sixteen years of watching the wrong documents drive the wrong outcomes to know exactly what to build when it did.
Glenn Kunsman, DO — Founder, Medico PC | Founder, Medico Disability Services
medicodisabilityservices.com · Glendale, AZ