Invisible Women: default male risk

Sep 5

Hello my very dear GFPs, and a very happy September to you all!It's been...some weeks, I'm not entirely sure how many. In any case, I'm here now! I hope you've all been having wonderful Augusts. Skip to the end for some pics of how Poppy has been handling her August.September is a busy month for me. First up, and most excitingly, we are launching season 2 of Visible Women, which I know you've all been extremely impatient to hear. Well, the wait is over! First episode drops THIS WEDNESDAY and it's on a topic I know is close to many GFP hearts: murderous menstrual blood, which faithful GFPs may remember we covered briefly in this newsletter about a year ago. Available wherever you get your podcasts -- can't wait to hear what you think!I am also going to be in NORWAY this month. I have been to Norway once before, in early March 2020, so shortly before the first lockdown. I remember that I was definitely doing a lot of nervous hand sanitisering -- and feeling quite foolish about it, feeling like I needed to do it surreptitiously in case people thought I was hysterical or implying they were dirty. Two weeks later we were in our first lockdown. *Shudder*.The trip was a flying one to give a talk in a city called Tønsberg and it was (obviously) my last international work trip before the world closed down, so it seems somehow kind of fitting for Norway to be my first international work trip now the world has opened up again. I'll be doing an event at the House of Literature in Oslo on the evening of Friday the 16th, and another the next day in Trondheim at 1pm. If there are any Norwegian GFPs reading this who are free, nearby and able to come -- please do! Would be lovely to have some friendly faces in the audience for my first international event in...over two years! Suddenly feeling quite nervous... 😬Right, that's enough housekeeping, let's check in on how the gender data gap has been getting on in our absence...

[caption align="alignnone" width="980"]

[/caption]

Gender data gap of the week

Women denied best treatment for heart attacks ‘because of sexism’ | News | The Times — www.thetimes.co.uk Thousands of British women are being denied the best treatment for heart attacks as a result of sexism in medicine, a major study has revealed.Research suggests

Sigh. I'm not angry, I'm just disappointed.

jk obviously I'm jolly cross. Although, it has to be said, not even slightly surprised because of, well, literally everything I wrote about in Invisible Women the book, and of course the research we did for episode 3 of Visible Women the podcast. In this episode we looked specifically at bias in healthcare and spoke to Sally Bee, who had multiple heart attacks misdiagnosed. Here's a clip from that episode:

[embed https://youtu.be/a6zSdfU2Z-s]

You can listen to the rest of the episode here or wherever you get your podcasts.

So, anyway no, I was not surprised to discover that "nearly 12,000 women in the UK have missed out on appropriate care over the past two decades, with some dying as a result."

The research, led by scientists at Imperial College London and the University of Zurich, and published in The Lancet seems to be, if you can believe this, the first sex-disaggregated analysis of the Global Registry of Acute Coronary Events (GRACE) 2.0 score (lol jk of course you can believe it.)

The GRACE score is used to risk assess the likelihood of death in patients who come in with symptoms of Acute Coronary Syndrome. Naturally, it has been "derived from and validated in predominantly male patient populations." And, also naturally, it "does not include sex among its determinants," and "is used in both sexes alike," without accounting for sex-specific disease characteristics.

So far, so Invisible Women, specifically p.218 for male-dominated risk prediction models for heart disease.

Then of course there was that AI that I wrote about in TIME magazine (and also had a minor rant about in this episode), which was hailed around the press a few years ago under headlines like this one:

AI to predict heart attacks at least five years before they occur | The Scotsman — www.scotsman.com Artificial intelligence could be saving lives within a year by making it possible to identify people at high risk of a heart attack at least five years before it strikes, scientists claim.

...but provided barely any sex disaggregated data and trained its algorithm on heavily male biased data. 🤪

Anyway, these researchers from Imperial and Zurich have now done the analysis that should have been done decades ago and they've found some, er, slight problems with old Gracey.

It turns out that GRACE 2.0 has been systematically underestimating the risk of death in women for years. ikr, this is unprecedented.

[caption align="alignnone" width="980"]

just to cheer you all up 😘[/caption]

Here's just one example of how this played out.

So, using the GRACE score, kidney function in female patients kept coming out as better than in male patients. Slight problem with this: the threshold they used to score renal function was based on male data.

Let me explain (or try to, I am not a doctor). So, basically the GRACE score uses creatine concentration as "a surrogate of renal function." Which is fine, except men and women have a different physiological range, and as previously discussed, the GRACE score doesn't account for sex. Using this one-size-fits-men score-card, female patients appeared to have better kidney function than males. BUT, "the estimated glomerular filtration rate (eGFR)," which is a sex-adjusted measure of renal function, ie it takes sex into account, "indicated the opposite." ie, if you account for sex, women's renal function was in fact lower than men's. Enthusiastic double thumbs up all round.

The researchers also found that, in the UK, female patients were

more likely than male patients to have preserved ejection fraction at baseline, another factor with potentially different prognostic implications after acute coronary syndrome in women and men and not included in the GRACE score.

And so on.

The result of this systematic underestimation of mortality risk for female patients is that women were less likely to receive "early invasive management." They were less likely "to receive coronary angiography and to undergo early invasive therapy than were male patients."

Obviously, this is infuriating. But GFPs, do not despair. This is a good news story, because new sex disaggregated data is ALWAYS a good news story, especially if it's highlighting a disparity like this; it all adds to the evidence base, making it harder and harder to continue to ignore this problem.

Not only that, but these authors claim to have found a solution:

They proposed replacing it with an algorithm that uses artificial intelligence to score a patient on how likely their heart attack is to be deadly, eliminating gender bias.

The lead author, Professor Thomas F Lüscher, a consultant at the Royal Brompton Hospital in London, said: “I hope the implementation of this novel score will reduce sex inequalities and eventually improve the survival of patients with heart attack, both male and female.”

Dr Florian Wenzl, his co-author, said: “The study shows that established risk models which guide current patient management are less accurate in females and favour the undertreatment of female patients.

“Using a machine learning algorithm and the largest datasets in Europe we were able to develop a novel artificial intelligence-based risk score which accounts for sex-related differences in the baseline risk profile and improves the prediction of mortality in both sexes.”

The really interesting thing is that this new risk prediction model, using sex disaggregated analysis, performed better not just for women, but for men too. Almost like sex disaggregation is better for everyone or something??

GRACE 3.0 does still perform better in men -- I suspect this may be down to the fact that women do remain underrepresented -- so there is still progress to be made, but this is nevertheless a huge step forward. And it will save women's lives. Actual enthusiastic double thumbs up.