Under the Varnish

The Stability of Transgender Identity in Adolescents

For almost two hundred years, people stood in front of Rembrandt’s most famous painting and saw a night scene. They even called it The Night Watch, and the name stuck so completely that it survives to this day, even though Rembrandt never called it that, and even though it is not a painting of the night. When conservators finally removed the layers of darkened varnish after the Second World War, sunlight came pouring across the militia. The gloom that generations of viewers had attributed to the artist’s intention — the mood, the meaning, the whole interpretive tradition built on that darkness — was oxidation. It was grime. Two centuries of people had been reading the varnish and calling it the painting.

In Dresden there is a Vermeer, a girl reading a letter at an open window. For most of its life, the wall behind her was bare. Scholars wrote about that bareness — the privacy, the ambiguity of the letter in her hands. Then X-rays found a painting of Cupid hidden on the wall behind her, and for forty years everyone assumed Vermeer himself had painted it out, a revision by the master’s own hand. It took until 2021 for conservators to establish that the overpainting was done decades after his death, by someone else entirely. When they removed it, the painting changed its meaning. The quiet domestic scene had been a love letter all along. The interpretation everyone trusted was not the artist’s intention. It was a stranger’s hand, mistaken for his.

I immediately thought about these paintings because of a new study out of Canada, and because of what Jack Turban did with it on X.

Photo by Darth Liu on Unsplash

The study, published in the Journal of Adolescent Health, looked at 445 adolescents referred to four Canadian pediatric gender clinics between 2012 and 2017. Most of them — 353 — went on to cross-sex hormones. That is roughly 80%. Turban tweeted the finding like this: “A new study of 445 adolescents treated at pediatric gender clinics in Canada found that after a mean 2.4 years, 94.6% still identified as trans.” (It was a median, not a mean, and 94.6% is not quite the figure for that claim, but never mind. Tweets are written from abstracts, and I could have easily made a similar mistake — mean, median, mode — same thing, right?)

The phrase doing all the work in that study — the phrase I want to stop on — is retrospective chart review.

If you have never worked in medicine, that phrase sounds rigorous. It sounds thorough. People imagine their own medical records as a comprehensive account of their lives — they worry about it, even, this file that holds the things they would not tell anyone except their doctor. Maybe their spouse. Sometimes not. Maybe their priest in confession, if they still went to church. So when the public hears that researchers reviewed the charts of 445 young people, they picture someone reading 445 lives.

When I read the phrase retrospective chart review, my heart sinks. Because I know there are many ways to do one.

Here is what this one reviewed. From each chart, the researchers pulled the age at which the young person said their gender distress began, the dates of referral and visits, the identity stated at the first appointment, whether that stated identity changed in the chart, and the start and stop dates of the medications. That is the painting. That is the whole painting. The clinics’ own intake process generated comprehensive psychological assessments — coexisting mental health conditions, comorbidities, the works — and none of it appears in the study. No mental health outcomes. No measure of whether anyone got better or worse. The study asks one question of each chart: did the label and the prescription continue?

And here is the detail that should be quoted every time this paper is cited: the researchers report that not one of the 445 adolescents was lost to follow-up. Zero. In a field where every honest cohort study hemorrhages patients, this would be a miracle — until you read how they defined follow-up. It was the elapsed time between a patient’s first visit and the day a research assistant opened the file. Not the last time anyone saw the patient. A teenager who came twice in 2015 and vanished still has years of “follow-up” under this definition, and is counted as stable, because the last thing written in the chart is the identity they walked in with. Silence in a chart is not stability. In this study, silence is stability. That is the varnish.

I worked in a pediatric gender clinic for four years. I saw nearly 1,500 patients come through. So when I read retrospective chart review, I don’t think about methodology first. I think about my individual patients — the ones who would make up a subset of any 445-patient file pull.

One young man comes to mind immediately. He came to us at seventeen, from a troubled family background, like a good number of them did. He was identifying as female. He was brought to the clinic by two adult men — partners — who had taken him in. The endocrinologist decided that on the day this young man turned eighteen, it was fine to go ahead and start him on a medical pathway (he waited until that birthday because I stressed that there was no actual legal guardians with this young person at the visit).

He came back for a follow-up visit a few months later, after he had been started on hormones, and he was not doing well. The endocrinologist asked me to go in and check on his mental health, and before I did, the endocrinologist “reviewed his chart” — his words. In the rhythm of a specialty clinic, an endocrinologist has maybe two minutes to review what happened between the last visit and today, and that review is built to answer exactly one question: is anything majorly wrong with an endocrine lab? Nope. Okay. Move on. So the review went like this: he saw that there had been a few emergency room visits between appointments, noted them, and moved on. They weren’t endocrinology visits. They weren’t labs. In his mind, they had nothing to do with him.

I have a different way of reviewing a chart. I click in, and then I click in all the way through. I look at what time of day a kid showed up in an emergency room- that matters. I read the chief complaint. I read the nursing notes, and the social work notes, and I keep reading, every encounter, because if you dig long enough it is all in there. And what I found, digging through this young man’s visits, was devastating. He had gone to the emergency room in real mental distress because he was being sexually abused — by the men who had brought him to our clinic. He was being harmed in the home he lived in, cycling toward homelessness, unsure whether he would have enough to eat or where he would sleep or whether the people sheltering him would hurt him again. That was what was under the visits the endocrinologist had glanced at and scrolled past. We had moved forward and medicalized this young man on the strength of a snapshot — a stated identity, a checked box, a chart unread.

If you do a chart review with a storyline already in hand — trans-identified youth, mental health struggles, that’s the story, that’s why they’re in the ER, and transition is the thing that will fix it — then the ER visits confirm what you already believed, and you never click in. The narrative functions exactly like varnish. It darkens everything to one color, and people stand in front of it for years and mistake the darkness for the painting.

And here is the part I keep coming back to, because the two-minute lab check is not an aberration in this care model — it is the care model, in miniature. We took a population of young people in psychological distress and assigned their care to hormone prescribers, whose training, whose appointment structure, whose entire clinical attention is organized around one question: are the labs in range? We were trying to treat a mental health condition by physically changing the body with endocrine chemicals. The endocrinologist scrolling past the emergency room visits was not failing at his job. He was doing his job. His job was simply never the right tool for what was actually wrong with the children in front of him. That mismatch was built into the model from the beginning, which is why the model was always destined to fall apart. It was never going to work. It never could have worked.

And it is not only the devastating stories. It is the nuance a real review would carry. Not just did they show back up at the gender center two years later, but: did they end up at the sleep specialist in the interim? At the urologist, because they now have urinary dysfunction? What happened to their cholesterol, their hematology labs — did any of this make them healthier, or worse? None of that is in the Canadian study, because none of it was extracted, because the question asked of the charts was never how are they — only are they still ours.

At our clinic, we lost about thirty percent of our patients to follow-up. We knew only that they stopped coming back. And a lot of clinicians wrote that off with a rosy picture: they’ve moved on, they’re in college now, surely they’re continuing care through some campus health system. But if you have the time, and the willingness, the fuller picture is findable. In Epic there is a function called Care Everywhere — records linked across systems — and if you sit and click and care enough about a patient as a human being, you see where the quiet ones went. They stopped coming to see us, and they started going to a substance abuse treatment center. They stopped coming to see us, and they’re back with their general pediatrician, where they would never use the word detransitioner, but you can see they are no longer injecting testosterone. They removed themselves from our picture — from our screen, from our chart review — and cobbled together some kind of medical home somewhere else. They didn’t announce anything. They just left. The literature on detransitioners says most of them never tell the clinic that treated them, and I believe it, because I watched it happen one quiet chart at a time.

Those young people are in studies like this one. They are inside the 94.6%. Their charts went silent, and their silence was tallied as persistence, and their persistence was tweeted as reassurance.

A medical chart is not a snapshot. It is an old painting that many hands have worked on. The patient’s own account is the underdrawing. Every clinician who touches the chart adds a layer — and some of those layers are honest pigment, and some are overpainting done by people with their own ideas about what the picture should show, and over all of it settles the varnish of the prevailing story, darkening the details into a single mood that viewers learn to call the truth — and often the layers aren’t even fresh paint; clinicians copy and paste the old visit into the new. A retrospective chart review can mean restoration — the slow, patient work of lifting each layer to find out what is actually there. Or it can mean standing in front of the varnish with a clipboard, recording its color, and publishing the color as a finding.

The Canadian study recorded the color. Four hundred forty-five paintings, and the question put to each one was whether the surface had visibly changed. The researchers report that it mostly had not, and they are right: the surfaces had not changed. The Night Watch did not change either, for two hundred years. It was a day scene the entire time.

When conservators cleaned the Vermeer, the meaning of the painting reversed, because what everyone had taken for the artist’s intention turned out to be someone else’s hand. I think about that when I think about these charts — about how much of what is written in them is the patient, and how much is the institution’s hand painting over the patient, and how a study that reads only the top layer cannot tell the difference and does not try. The 94.6% is a true description of 445 charts. It is not a description of 445 children.

What got reported was the persistence of paperwork, and what got tweeted was the persistence of people, and the entire reassurance lives in the space between those two things — the same space where a young man’s emergency room visits sat unread, three clicks down, the whole time.