Every essay, in one place.
Four lanes. Pick the one that fits what you came for.
- Doctrine: how the system works
- Field Notes: practical lessons
- Translations: explanations for patients
- The Lab: try the ideas
The mammogram that didn't count
She got the mammogram. The scorecard says she didn't. Quality in primary care rarely fails at the doing. It fails on the trip back.
Don't build a pipe on a lie
Automation doesn't fix bad information. It delivers it faster, to more people. That's why I built the boring layers first and the AI last.
A discharge summary is a story
The next doctor doesn't need a form. They need to know what happened. Here's how to write a discharge summary someone can actually use.
Two vials, one arm
If two doctors order your blood work, you may get stuck twice, and neither doctor may see the other's results. Here's why it happens and what you can do about it.
Data doesn't add hours. It aims them.
I built a simulation of one year in a 2,500-patient panel and ran it four ways. The result I didn't expect: aiming the hours mattered more than adding them.
Who holds the key
Ten years after Congress outlawed information blocking, enforcement finally has a pulse. And a primary care practice is on both sides of it.
The 27-hour day
Give 2,500 adults the care the guidelines recommend, and the math comes to 26.7 hours a day. That number explains more about primary care than any survey on burnout.
The inbox is the second clinic
Nobody schedules it. Nobody bills for most of it. It runs every day after the last patient leaves, and it is getting bigger.
Red tape, itemized
Six data pipes already exist. Almost none reach the exam room in time. If I could fix one first, here is which one, and why.
What replaces the visit: a continuous clinical surface
If the fifteen-minute visit is no longer the unit of care, something has to be. And it has to be precise enough that a payer can contract against it and an auditor can review it.
How to get your whole medical record
It is your record. You have a legal right to it, there is a deadline, and there is a cheaper way than paying for copies.
MEAT, when an auditor walks in
A diagnosis you treated but didn't document is a diagnosis you didn't treat, as far as the audit is concerned. The four-letter check that keeps the chart honest.
The prior-auth clock, 2026 edition
Since January, most government-program payers have had to answer within days and tell you why they said no. Here's how to use the clock instead of waiting on it.
How to read a risk score without getting fooled
A number with a percent sign looks like a fact. It is a guess with a method. Five questions to ask before you act on one.
You saw your results before I called. Here's why.
Your test results now arrive in your patient portal as soon as they're ready, often before your doctor has seen them. That's on purpose. Here's how to read them without panicking.
The 15-minute visit is a settlement, not a standard
Primary care didn't choose the fifteen-minute visit; it inherited it. The number is a billing artifact now asked to carry a workload it was never priced for. The difference gets absorbed by the clinician in the room.
A letter between computers
What HL7, ADT, and FHIR actually are, explained without jargon, with one patient's trip through a hospital as the running example.
Your blood pressure is a verb, not a number
Why the single reading on the cuff is the least interesting thing about it, and what your body is actually doing between visits.
The differential isn't a list. It's a probability cloud.
The single mental upgrade that separates students who memorize differentials from clinicians who actually use them.
Value-based care, explained without a single buzzword
Strip the jargon and it's almost embarrassingly simple. The hard part was never the concept. It's the operational reality.
What 'preventive care' actually prevents
Spoiler: not everything. A clear-eyed look at what screening can and can't do, without the wellness-industrial-complex hype.
Reading a troponin without panicking
A high-sensitivity troponin is not a yes/no light. Treating it like one is how good students order bad workups.
I let an AI draft my notes. Here's the line I won't let it cross.
On building clinical tooling that's genuinely useful without quietly outsourcing the part that makes you a doctor.