Sign in

Wouter van Amsterdam

@vanamsterdam.bsky.social
1.3K followers 254 following 41 posts

machine learning, causal inference, healthcare - assistant professor in dep. of Data Science Methods, Julius Center, of University Medical Center Utrecht, the Netherlands; wvanamsterdam.com

PostsRepliesMedia
Wouter van Amsterdam @vanamsterdam.bsky.social · 04/09/2025
work with Diantha Schipaanboord, Floor B.H. van der Zalm, René van Es, Melle Vessies, Rutger R. van de Leur, Klaske R. Siegersma, Pim van der Harst, Hester M. den Ruijter, N. Charlotte Onland-Moret, on behalf of the IMPRESS consortium
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 04/09/2025
Conclusion: The convolutional neural networks in this study demonstrated resilience to simulated sex-imbalance in training ECG data. pre-print: doi.org/10.1101/2025...
doi.org
ECG classification with convolutional neural networks demonstrates resilience to sex-imbalances in data
Background: Many ECG-AI models have been developed to predict a wide range of cardiovascular outcomes. The underrepresentation of women in cardiovascular disease studies has raised concerns if these m...
120
Wouter van Amsterdam @vanamsterdam.bsky.social · 04/09/2025
Discrimination remained stable across sexes; only calibration shifted in extreme scenarios when prevalence differed by sex, with similar patterns for women and men.
100
Wouter van Amsterdam @vanamsterdam.bsky.social · 04/09/2025
Using ~165k ECGs, we simulated sex-imbalances in representation (women-to-men ratio), outcome prevalence, and misclassification in the training data for LBBB, long QT syndrome, LVH, and physician-labeled “abnormal” ECGs.
100
Wouter van Amsterdam @vanamsterdam.bsky.social · 04/09/2025
Pre-print alert: Many ECG-AI models have been developed to predict a wide range of cardiovascular outcomes. But, underrepresentation of women in cardiovascular studies raises the question: Are ECG-AI models equally predictive for women and men with sex-imbalanced training data?
120
Reposted by Wouter van Amsterdam
Nan van Geloven @gelovennan.bsky.social · 11/08/2025
New paper in @annalsofim.bsky.social "50 ways to misinterpret clinical prediction models for treatment decisions” --> Published version: www.acpjournals.org/doi/10.7326/... --> Open access version: arxiv.org/pdf/2402.17366
acpjournals.org
The Risks of Risk Assessment: Causal Blind Spots When Using Prediction Models for Treatment Decisions | Annals of Internal Medicine
Clinicians increasingly rely on prediction models to guide treatment choices. Most prediction models, however, are developed using observational data that include some patients who have already receiv...
21710
Reposted by Wouter van Amsterdam
bms-aned.bsky.social @bms-aned.bsky.social · 06/06/2025
BMS-ANed Spring Meeting on Thursday, June 19 Time: 13:00–18:00 (CEST) Location: Vredenburg 19, 3511 BB, Utrecht Details and registration: vvsor.nl/biometrics/e...
vvsor.nl
Hans van Houwelingen award ceremony and symposium June 19th 2025 - VVSOR
This spring, the BMS-ANed organises an in-person meeting:
011
Reposted by Wouter van Amsterdam
Oisín Ryan @oisinryan.bsky.social · 28/04/2025
Still some spots available in our summer school on all things causal inference, 7-11 July in Utrecht! Discounts for those working in universities and non-profits, and affordable accommodation offered by @utrechtuniversity.bsky.social summer school!
186
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/04/2025
Even if you model a physical system, e.g. avg yearly temperature depending on height, and assume that temp given height is the same everywhere. If you invert it into predicting presence of mountain given temp, you’ll find varying discrimination in diff countries. Example from scholkopf’s talks
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/04/2025
You’ve modeled a system with no meaningful variation across environments. The model may be reliable in the tested environments but you haven’t shown robustness against variation in distributions as you haven’t observed any
200
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/04/2025
A question that remains is how these differences in environments may come about and what to do with this in practice? On this, I wrote a paper titled, available here: arxiv.org/abs/2409.01444 fin!
arxiv.org
A causal viewpoint on prediction model performance under changes in case-mix: discrimination and calibration respond differently for prognosis and diagnosis predictions
Prediction models need reliable predictive performance as they inform clinical decisions, aiding in diagnosis, prognosis, and treatment planning. The predictive performance of these models is typicall...
021
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/04/2025
if the distribution of outcome given features remains the same (Y|X), calibration is preserved. If both are the same, the environments were not meaningfully different to begin with! a more lengthy explanation is in this blog post: wvanamsterdam.com/posts/250425...
wvanamsterdam.com
211
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/04/2025
as promised (so all of you can breathe normally again), here's my TLDR answer: Environments must differ with respect to something. If the distribution of features given outcome remains the same (X|Y), discrimination is preserved;
100
Wouter van Amsterdam @vanamsterdam.bsky.social · 24/04/2025
tagging some prediction modelers / statisticians, @maartenvsmeden.bsky.social @benvancalster.bsky.social @gelovennan.bsky.social @f2harrell.bsky.social @lucystats.bsky.social @miguelhernan.org @gscollins.bsky.social (I will answer tomorrow)
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 24/04/2025
Which is stronger evidence for robustness? When evaluating predictive performance of one model in several different environments (e.g. regions / hospitals): A. stable discrimination (AUC) and calibration in all environments B. stable discrimination, varying calibration vote with 👍=A; ❤️=B
311
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/04/2025
ask chatGPT o3 this before submitting your next paper to, I got ~10 usable comments out of it: you're a reviewer for <journal>; review the attached paper when you're either:
020
Wouter van Amsterdam @vanamsterdam.bsky.social · 11/04/2025
what are the exceptions?
100
Wouter van Amsterdam @vanamsterdam.bsky.social · 09/04/2025
2. an external reproduction of the PROTECT method from Manchester University with Charlie Cuniffe, Matt Sperrin and Gareth Price (www.nature.com/articles/s41...) 3. a 'causal' meta-analysis method using only aggregate data, exciting work with Qingyang Shi from Groningen University
nature.com
Individual treatment effect estimation in the presence of unobserved confounding using proxies: a cohort study in stage III non-small cell lung cancer - Scientific Reports
Scientific Reports - Individual treatment effect estimation in the presence of unobserved confounding using proxies: a cohort study in stage III non-small cell lung cancer
010
Wouter van Amsterdam @vanamsterdam.bsky.social · 09/04/2025
Very excited for my first (belated) visit to #EuroCIM2025! I'm here with 3 bits of work: 1. a poster on a causal understanding of prediction model performance under shifts in 'case-mix' (or covariate / outcome drift); I show how discrimination and calibration respond differently bit.ly/ccm-arxiv
bit.ly
A causal viewpoint on prediction model performance under changes in case-mix: discrimination and calibration respond differently for prognosis and diagnosis predictions
Prediction models inform important clinical decisions, aiding in diagnosis, prognosis, and treatment planning. The predictive performance of these models is typically assessed through discrimination a...
2102
Wouter van Amsterdam @vanamsterdam.bsky.social · 03/03/2025
this seems pretty cool: an overview of llms for statisticians arxiv.org/abs/2502.17814
arxiv.org
An Overview of Large Language Models for Statisticians
Large Language Models (LLMs) have emerged as transformative tools in artificial intelligence (AI), exhibiting remarkable capabilities across diverse tasks such as text generation, reasoning, and decis...
040
Reposted by Wouter van Amsterdam
Nan van Geloven @gelovennan.bsky.social · 30/12/2024
Vacancy for a postdoc position. Improve the transparency of decision support algorithms by figuring out how we can quantify and communicate uncertainty in individual causal predictions. With Marleen Kunneman, Daniala Weir and me. Three more days to apply 👇 www.lumc.nl/en/about-lum...
lumc.nl
Postdoc Biomedical Data Scientist / Biostatistician | LUMC
In this postdoc position at LUMC, you will work on groundbreaking research that enhances the transparency and trustworthiness of decision support algorithms in healthcare. This position allows you to ...
065
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/12/2024
Building in the physics is one way to potentially get the right causal mechanisms In sofar as the model is trained on real world patient data, you'll still have to ensure no biases e.g. related to confounding creep in
120
Wouter van Amsterdam @vanamsterdam.bsky.social · 22/12/2024
Digital twins are useful insofar as they reflect causal mechanisms Don't think a generative model ('digital twin') can inform treatment decisions just because it procudes different outputs when you give it different inputs. Doesn't matter if it's 'AI' or not.
160
Wouter van Amsterdam @vanamsterdam.bsky.social · 17/12/2024
saliency maps are the new table 2 fallacy
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 16/12/2024
Not sure about overfitting, results seemed robust to 5-site cross validation. It just learns correlations, what's wrong with that? The words 'confounders' and 'bias' make it sound they expected the model to yield some causal understanding. Maybe these heatmaps are the new table 2 fallacy
120
Wouter van Amsterdam @vanamsterdam.bsky.social · 16/12/2024
Awesome, congrats!
110
Wouter van Amsterdam @vanamsterdam.bsky.social · 06/12/2024
Liking this interaction with @mmbronstein.bsky.social and Denis Danilov so much I'm reposting it here
3374
Reposted by Wouter van Amsterdam
Oisín Ryan @oisinryan.bsky.social · 04/12/2024
Interested in how to use non-experimental data to answer causal research questions? Mystified by DAGs and counterfactuals? Want to learn what Target Trial Emulation is all about? Sign up now for the 2nd edition of our summer school, 7-11 July in Utrecht, with @vanamsterdam.bsky.social & BPdeVries
utrechtsummerschool.nl
Introduction to Causal Inference and Causal Data Science | Utrecht Summer School
The course takes an interdisciplinary approach and is suitable for applied researchers across health, social and behavioural sciences.
15414
Wouter van Amsterdam @vanamsterdam.bsky.social · 27/11/2024
Probably more like "the average of an infinite sequence of throws hits the bulls eye"
110
Wouter van Amsterdam @vanamsterdam.bsky.social · 17/11/2024
@oisinryan.bsky.social and I are developing a julia package for target trial emulation with a student, happy to be added to the list
030
Wouter van Amsterdam @vanamsterdam.bsky.social · 05/11/2024
Hahaha; btw an hour after dinner, that’s quite some energy to spend!
110
Wouter van Amsterdam @vanamsterdam.bsky.social · 29/10/2024
note that this is not meant as a critique of anyone; I'm curious if other people on this platform feel the same way. Also, maybe toning down on politics may help keep bots and trolls away from a time-line / network?
210
Wouter van Amsterdam @vanamsterdam.bsky.social · 29/10/2024
Can someone make a "science-not-politics" starter pack? Great respect for academics who are politically engaged, and yes, bsky/X are places for political discussion / news sharing. But science and politics take different parts of my brain; can we get science without political distractions?
4284
Wouter van Amsterdam @vanamsterdam.bsky.social · 24/10/2024
Look at @maartenvsmeden.bsky.social go, he's won the VIDI grant! (arguably the most prestigious personal grant in the Netherlands at his career stage). Congrats!
0162
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/10/2024
Anyone else here afraid we’re enjoying a VC backed honeymoon on this platform, and that once we’ve (re-)established a valuable network, the money-making ad-serving attention-grabbing algorithms will kick in and spoil the fun?
110
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/10/2024
We're still accepting applications for this PhD position, send in your application by Oct 29!
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 21/10/2024
hahaha, "Granger Danger"
000
Reposted by Wouter van Amsterdam
Maarten van Smeden @maartenvsmeden.bsky.social · 21/10/2024
Created a new group replacing and updating a list I enjoyed following on the bird site. Mostly people posting on medical stats and DS/AI ICYI: go.bsky.app/ArqEz36
go.bsky.app
Medical stats/ds/ai
Join the conversation
76732
Wouter van Amsterdam @vanamsterdam.bsky.social · 15/10/2024
new PhD position with @maartenvsmeden.bsky.social and me! are you: - interested in the intersection of "science" and deep learning? - keen to work with electrocardiography (ECG) data - eager to learn and be part of a vibrant data science team of the UMC Utrecht? see bit.ly/3UaSN23
bit.ly
Vacancy — PhD student SciML4Medicine
PhD student SciML4Medicine: electrocardiography analysis with mathematical modeling and deep learning.
043
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/09/2024
And Doranne Thomassen finishing up with estimands in RCTs with an application in oncology
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/09/2024
Followed-up by Dimitris Rizopoulos on using joint models for dynamic treatment policies in salvage therapy in prostate cancer
110
Wouter van Amsterdam @vanamsterdam.bsky.social · 23/09/2024
Joost van Rosmalen from #umcutrecht kicking of the Causal Inference for AI meetup with his talk on using data from historical controls for trials with dynamic borrowing methods
120
Wouter van Amsterdam @vanamsterdam.bsky.social · 12/09/2024
this tweet isn't too hopeful about the migration: x.com/emollick/sta...
x.com
x.com
100
Wouter van Amsterdam @vanamsterdam.bsky.social · 12/09/2024
free entry, no sign-up needed, location and talk-titles are here: www.uu.nl/en/events/ca... on-site only
uu.nl
Causal Inference for AI in Health Meeting
In this quarterly meeting, researchers from a network of 4 universities come together to present recent work on methods and applications of causal inference, followed by an informal networking session...
020
Wouter van Amsterdam @vanamsterdam.bsky.social · 12/09/2024
We're hosting a Causal AI meet-up with Nan van Geloven, Jesse Krijthe, @oisinryan.bsky.social and @jeremylabrecque.bsky.social - this time focussed on Health, Speakers: Joost van Rosmalen, Dimitri Rizopoulos and Doranne Thomassen University Library Utrecht USP (Boothzaal) Sep 23, 14.30-17.00
121
Reposted by Wouter van Amsterdam
Oisín Ryan @oisinryan.bsky.social · 08/04/2024
Want to learn how observational data can give insight into causal effects? How to specify a target trial & use prediction models for causal insight? Sign-up now for "Intro to Causal Inference and Causal Data Science", Aug 5-9 Utrecht: utrechtsummerschool.nl/courses/heal... @vanamsterdam.bsky.social
utrechtsummerschool.nl
Introduction to Causal Inference and Causal Data Science | Utrecht Summer School
The course takes an interdisciplinary approach and is suitable for applied researchers across health, social and behavioural sciences.
1136
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/03/2024
still much to be done, see also e.g. the planned 'extensions' very curious to get feedback and tips on this post, so tagging some researchers / developers: @kdpsingh.bsky.social @maartenvsmeden.bsky.social @benvancalster.bsky.social @casperalbers.nl
000
Wouter van Amsterdam @vanamsterdam.bsky.social · 25/03/2024
Ever need to do many repeated analyses? e.g. power calculations, sensitivity analyses, then speed becomes a crucial! In this blog post I compare newcomers #Julia google's #JAX versus #R for logistic regression TLDR: Julia wins, ~10x faster than R vanamsterdam.github.io/posts/240308...
vanamsterdam.github.io
Wouter van Amsterdam - The need for speed, performing simulation studies in R, JAX and Julia
Wouter van Amsterdam’s academic home page
120