Thoughts on healthcare, technology, and the work of explaining both clearly. Mostly about the gap between content that is accurate and content that actually works.
Featured
In plain terms: for people carrying sickle cell trait, the standard diabetes test is meaningfully worse at finding diabetes.
So a Nigerian patient reads an article about HbA1c, written for an American audience, that tells them below 6.5 means they are fine. There is a one in four chance the test is systematically underselling their glucose. Nothing on the page mentions it.
The reviewer’s realistic options are to approve it, or to detonate a week of somebody’s work over a problem that is hard to articulate and does not look like an error.
I have been that reviewer.
Flags → detects → identifies → diagnoses.
A tool that marks a region for review is doing something quite different from a tool that determines what the region is. Diagnosis for most solid tumours happens on tissue, examined by a pathologist.
Medical content needs two competencies that rarely live in the same person.
Companies usually solve for one and hope the other sorts itself out. Hire a brilliant content writer and you get pages that read beautifully and occasionally claim things you cannot support.
Once you understand something, you lose the ability to remember not understanding it. This is not a character flaw. It happens to everyone, and it is the single biggest reason expert-written health content underperforms.
You stop noticing that “contraindicated” is not a normal word. That “presents with” is hospital dialect.
The patients who enrol easily, adhere well, and generate clean continuous data are disproportionately the patients who were already doing well. They have the phone, the power, the data, the literacy, and usually the milder disease.
Your programme then reports strong outcomes. Some of that improvement is real.
The clinician wants to know what happens on a bad shift. Not the accuracy figure. Whether this adds work, whether the alerts will be right often enough to be worth reading, and whether they can override it without writing a justification.
For a lot of health queries, the ideal outcome is the reader leaving your page within thirty seconds because they now know they need to phone someone. That is a successful page. It will look like a bad one in your analytics.
FHIR breaks medical information into small standard pieces instead. There is a standard shape for a patient. A standard shape for a single lab result. One for a prescription, one for an allergy, one for a diagnosis, one for an appointment.
Think of them as Lego bricks.