London rain came sideways against the glass of the Queen Elizabeth II Centre, and Claire watched it streak the window while three hundred cardiologists filed into the auditorium below.
"You're pacing," Ethan said from the dais, adjusting the lapel mic.
"I'm not pacing. I'm—reviewing sightlines."
"You've walked the same eight feet four times." He held out a paper cup. "Tea. Drink it before you spill it on your blouse."
She took the cup, and her fingers brushed his. The pin—Anna's anatomical heart, silver and small—sat pinned to her jacket, above her own heart. She'd worn it to the DMC hearing in Boston, and to the medical board filing, and now, absurdly, to London. It had become a talisman she hadn't meant to make.
"You keep touching it," Ethan said quietly.
"Nervous habit."
"It's not a habit. It's a promise." He straightened the pin without asking, his thumb grazing her collarbone. "She'd have liked this part. The showing-up part."
Claire looked at the auditorium. Rows of blue chairs, name cards in French, German, Portuguese, Dutch. The nonprofit's logo—a simple loop she'd sketched on a napkin two years ago—hung on the banner behind the stage. *Cardiac Equity Initiative.* She still found the name embarrassing. It sounded like a grant application. Which, at first, it had been.
"Doctor Bennett?" A woman with a Dutch accent appeared at the stage door, tablet in hand. "Dr. Vermeer's office confirms he'll ask about the sex-stratified validation cohort during Q&A. He wanted you to have warning."
"Thank you." Claire set down the tea. "Tell him I'll answer in four minutes or less."
The woman smiled. "He timed the last speaker. He'll hold you to it."
She left. Ethan raised an eyebrow. "Vermeer. That's the Leiden group."
"It's the Leiden group that rejected our first dataset in 2023. Said the disparity was 'specific to American care pathways.'" Claire pulled the clicker from her pocket, testing the slide advance. "Today he gets to hear that his own registry reproduced it."
"So this is personal."
"This is science." She pressed the clicker. The first slide bloomed on the screen: *Where Cardiac AI Fails Women: A Cross-National Validation.* "It's also personal."
The auditorium doors opened. The room filled in waves—doctors in suit jackets, residents in scrubs, a knot of nurses who'd clearly come together and claimed a whole row. By the time the moderator stepped to the lectern, every seat was taken and people lined the walls.
Claire and Ethan took the stage. The moderator ran through their credentials too fast for Claire to feel anything but exposed, and then it was her turn.
"Good morning." Her voice came out steady. She'd practiced this in the mirror of the hotel bathroom while Ethan slept. "Two years ago, an algorithm I helped build missed a woman's heart attack. Her name was not released. Her chart number was one-four-six. She died in a Boston emergency department while a machine told her clinicians she was low-risk."
The room went still. She had learned, in two years, that this was the only opening worth using.
"She was not an outlier. She was the visible point on a curve that had been hiding in plain sight for a decade. And when we published the original dataset, we were told the finding was local. American. An artifact of how we bill, how we code, how we practice." She advanced the slide. A world map filled the screen, pins dropping onto capitals one by one. "So we asked other countries to send us their data and tell us we were wrong."
Ethan clicked to the next slide. Numbers. The room leaned in.
"The answer is that we are not wrong." Claire walked to the screen and traced the curve with the tip of the clicker. "In twelve health systems, across four continents, the algorithm under-recognized female patients at rates between eleven and thirty-one percent. The magnitude varies. The direction does not. This is not a billing artifact. This is a training-data problem, and it is global."
A hand went up in the third row. A woman in a bright scarf—Portuguese, by the name card visible on the seatback. "Are you saying every deployment of this class of model should be recalled?"
"I'm saying every deployment should be re-validated, stratified by sex, before it touches a patient. Recalling it is the first step. Rebuilding it is the second. The third is the part nobody wants to fund."
"Which is what?" the woman asked.
"Transparency. Public reporting. A registry of harm." Claire set the clicker down. "We spent two years proving this is real. The next two years we have to spend proving it's fixable."
That got a murmur. Ethan stepped forward, hands in his pockets, the way he always did when he was about to say something uncomfortable.
"I'm a cardiologist," he said. "I've been the person on the other side of an alert. When I was a resident, I trusted the software because I was tired and it was fast and I wanted to go home. That's how these failures happen—not through malice. Through trust. So what I'm asking for is not blind faith. It's verification. Run the numbers on your own patients. Publish what you find. If the model is right for your population, say so. If it isn't, say that too."
The Portuguese doctor wrote something down. Several people did.
The Q&A ran long. Vermeer rose, as promised, and asked about the stratified validation cohort in a tone that suggested he expected them to have fudged it. Claire answered in three minutes and forty seconds, citing the preregistered protocol and the independent biostatistician, and the room laughed when Ethan held up four fingers behind her back.
A German nephrologist asked whether the model could be adapted to renal risk. A Canadian resident asked what a small hospital with no data-science staff should actually do on Monday morning. Claire pulled up the implementation toolkit slide, the one she'd spent six weeks arguing with the nonprofit's board about, and walked the resident through it step by step.
"The toolkit is free," she said. "So is the audit code. If your vendor won't run it, we will. If your regulator has questions, send them to us."
"Us," the resident repeated.
"The Initiative. That's what we're for."
When the last hand fell, the moderator checked his watch and said, "Please join me in thanking Dr. Bennett and Dr. Calloway."
The applause started politely. Then it didn't stop. The whole row of nurses stood first, and the room followed, three hundred clinicians on their feet in an auditorium in London, and Claire stood there with her hands at her sides and felt something she had not let herself feel since the day she'd pressed send on the November email.
Not vindication. Not triumph.
Enough.
Enough people knew. Enough people would carry it back to their hospitals. She looked at Ethan, who was looking at her, and for a second the standing ovation was just noise and the pin on her jacket was just metal and everything that mattered was the fact that they had done the thing and it had worked.
Then the Dutch woman was at her elbow again.
"Dr. Bennett. Dr. Vermeer's office. He'd like a word—privately." She hesitated. "He mentioned the European data-protection review board. There's a scheduled audit of the cross-border registry. He wanted you to hear it from him before the press."
Claire's smile held. Ethan's didn't.
"Which registry?" Ethan asked.
"The pooled one. The one that made today possible." The woman glanced at her tablet. "The board has raised concerns about how the patient data crosses jurisdictions. There's a meeting. Tomorrow morning, Brussels."
Claire looked at the auditorium, at the clinicians still talking in the aisles, at the banner with the loop on it. Then she looked at Ethan.
"Brussels," she said.
"We came for a standing ovation," he said.
"We came for this."
She unpinned nothing. Anna's heart stayed where it was. But her hand found his, briefly, before she turned back to the Dutch woman and started asking questions.