Just published @ Eur Heart J: Combined clinical, metabolomic, and polygenic scores for cardiovascular risk prediction academic.oup.com/eurheartj/ad... tl;dr these next-gen risk factors could boost the number of CVD events prevented by +80% while keeping treatment efficiency (NNT) the same.

ICYMI 👉 Quantifying the fairness of polygenic risk scores: our modelling and analysis for CVD risk prediction www.medrxiv.org/content/10.1... Overall PRS fairness levels similar to or better than traditional predictors (age, sex, BP, lipids) h\t @iamslambert.bsky.social @clairecoffey.bsky.social

Current polygenic risk scores are unlikely to exacerbate unfairness in cardiovascular disease risk prediction

Background Current cardiovascular disease (CVD) risk prediction models place many individuals in an intermediate risk category where clinical decision-making remains uncertain, highlighting a critical...

medrxiv.org

Awesome work by @clairecoffey.bsky.social @iamslambert.bsky.social @sritchie73.bsky.social Lisa Pennells and Angela Wood

There's much more in the preprint - a great team effort from @clairecoffey.bsky.social @sritchie73.bsky.social, Lisa Pennells, Angela Wood & @mikeinouye.bsky.social all from @cambridge-ceu.bsky.social!

Clinical statements (AHA, ESC) have suggested that PRS may be most useful as a risk-enhancing factor for those at intermediate risk. We used PRS to reclassify individuals at intermediate risk by QRISK3 and showed improved sensitivity and fairness benefits (particularly for age & sex).

We evaluated PRS in two contexts, as a standalone risk predictor and as a risk-enhancing factor. Given the known performance differences across ancestries, we were surprised to find that the fairness of PRS predictions across ethnicities was equal to or lower other CVD risk factors when used alone.

Algorithmic fairness metrics have made big inroads in the AI/ML fields to investigate model bias. A clever PhD student (@clairecoffey.bsky.social) thought, why don't you see these methods applied to CVD risk prediction or polygenic risk scores? Enter our new preprint: www.medrxiv.org/content/10.1...

Current polygenic risk scores are unlikely to exacerbate unfairness in cardiovascular disease risk prediction

Background: Current cardiovascular disease (CVD) risk prediction models place many individuals in an intermediate risk category where clinical decision-making remains uncertain, highlighting a…

medrxiv.org

📣 Latest from the lab: Performance of deep-learning-based approaches to improve polygenic scores www.nature.com/articles/s41... Its thought deep learning will substantially improve PGS but the reality is MANY have tried but no/little gain has been seen so far. Here we report our negative results.

What a great first day at #HAICON25 - I had a fab time in the poster session chatting about my work on AI for environmental health this afternoon, and I look forward to learning lots more in the next couple of days! @helmholtzai.bsky.social

📣 Interested in the environmental (un)sustainability of AI (and how to support positive change)? Or know someone who is? I have an open position for a postdoctoral researcher to join my group in Cambridge, UK 🌱 🚨Closing June 5th www.jobs.cam.ac.uk/job/51255/ Topics below 👇 #AcademicSky #GreenSky

Healthcare data is vulnerable to biases. Working with @acceleratescience.bsky.social and @clairecoffey.bsky.social, we are building an open source educational resource hub to mitigate algorithmic bias. Read the blog post below to find out more!

Accelerate Programme for Scientific Discovery@acceleratescience.bsky.social · 2y ago

📢 New blog! Read about @lekaa-shree.bsky.social's work on dataset scarcity & algorithmic bias in AI in healthcare. With collaborators in Cambridge & Uganda, Lekaa has examined challenges & is planning a virtual resources hub to improve bias mitigation strategies. bit.ly/41Shq7X

Excited to share this blog by my fab collaborator @lekaa-shree.bsky.social on tackling dataset scarcity & algorithmic bias in healthcare AI 🌍🧠 We're building an open-source hub of bias mitigation resources with @acceleratescience.bsky.social - stay tuned!👀 📖 science.ai.cam.ac.uk/machine-lear...

How can we … address dataset scarcity challenges and mitigate algorithmic bias to build better AI healthcare tools?

Artificial Intelligence models are only as fair as the data on which they are trained. They are vulnerable to biases in training data, often reflecting deep-rooted systemic inequalities, human prejudi...

science.ai.cam.ac.uk

I passed my PhD viva! 🎉 Thank you to everyone who has been a part of this incredible journey, especially my supervisors @iamslambert.bsky.social, Angela Wood, and @mikeinouye.bsky.social 😄

Congratulations @clairecoffey.bsky.social on passing your PhD viva (it's not everyday your first PhD student graduates)! It's been a pleasure to supervise your work applying algorithmic fairness techniques to risk prediction models and polygenic scores with Angela Wood & @mikeinouye.bsky.social.

I’m excited to share that I’ve joined @www.helmholtz-munich.de this week! 😄 Here, I’ll be using machine learning to better understand the relationships between environmental changes, socioeconomic factors, and health - building on my PhD research by integrating algorithmic fairness throughout. 🤓

Happy new year everyone! 💫 I'm pleased to share that I submitted my PhD thesis a couple of weeks back - just in time for the holidays - now to wait for the viva! It’s been an amazing journey and I'm excited for the new year, which marks the start of my next chapter! 🎓

Happy to share this report summarising the public’s views on improving healthcare through the use of healthcare data, technology, and artificial intelligence from the workshops we conducted in the UK and Uganda this year: zenodo.org/records/1450...