I am Vesela Kovacheva from Brigham and Women's Hospital Brigham and Women’s Anesthesiology Brigham and Women’s OB Anesthesia. I am a physician-scientist interested in using #AI #MachineLearning and genetics to investigate and improve maternal health. Very excited to join Dr Anastasia Mihailidou FAHA FCSANZ FESC Hypertension #HYPHIP .

some Hypertension House rules before we starrt!

1- Number your questions and/or comments (Q1, Q2…);

2- Don’t forget to tag the author, use #HYPHIP in all answers and Qs and 3) Remember that this is a friendly chat, so be respectful as always, enjoy and ask your Qs! Hypertension…

Welcome everyone to our chat with #HYPHIP author Vesela Kovacheva 👋

I am Dr Anastasia Mihailidou FAHA FCSANZ FESC from Northern Sydney Local Health District Kolling Institute Macquarie University

This is such a joy for me for this timely discussion on #Preeclampsia Prediction Using #MachineLearning & Polygenic Risk Scores from…

Upfront before we chat TOMORROW a warm ❤️WELCOME 💃👠 Tali Elfassy to the Hypertension X family Dr Anastasia Mihailidou FAHA FCSANZ FESC Carmen De Miguel, PhD, MS, FAHA Georg Ehret, MD FESC FAHA Francisco Rios Andrew Michael South, MD, MS Brandi Wynne, MS PhD FAHA Mariane Bertagnolli Charles German Eugene Yang Dr. Martha Gulati ♥️🫀❤️‍🩹🇨🇦 Guto Montezano The Touyz Lab David G Harrison

Congratulations Francisco Rios RI-MUHC CHALProgram for the Best Oral Presentation Award Hypertension Canada Congress!! Fran gave an amazing talk on Mg TRPM7 #inflammation #fibrosis in #Hyperaldosteronism 👏👏👏

Next is Guto Montezano representing the The Touyz Lab presenting on potential mediators of brown adipose tissue biology

#HYPHIP Q10:What about overfitting with deep learning (despite your huge sample size) or could any parameters that you put in explain your striking results ?

Shishir Rao Kazem Rahimi Hypertension Société Française d'Hypertension Artérielle Cardiology Geneva University Hospitals HUG

Francisco Rios Kazem Rahimi Hypertension Georg Ehret, MD FESC FAHA #HYPHIP A8: Not naive at all! This is a good question. The T-BEHRT model is a two stage framework for association analyses. 1) it extracts confounding variables in raw EHR before baseline and then uses the latent features learned to predict propensity score and the outcome 1/2

LiVe TwItTer ChAt: What did you want to know about #hypertension targets and #diabetes ?
To follow, simply use #HYPHIP

Please number the questions (so that the answer can be marked with the same number)

#HYPHIP Q1: Can you tell us what your paper is about ? Shishir Rao Kazem Rahimi Hypertension

Georg Ehret, MD FESC FAHA Kazem Rahimi Hypertension #HYPHIP A1: the association of systolic bp and CVD in patients with diabetes is uncertain. In our work, we showed that using AI modelling to adjust for latent confounding in rich UK EHR, the association of systolic BP and CVD in diabetic patients is a continuous relationship!

#HYPHIP Q1:In the past less aggressive BP target values have been recommended for diabetics (e.g. Mancia et al, Diabetologica 2018). Can you comment and put into context (j-curve vs. linear relationship).

Shishir Rao Kazem Rahimi Société Française d'Hypertension Artérielle Hypertension

Hi Shishir Rao Kazem Rahimi Thanks for join us today. Congratulations for your paper at @hyperAHA. Here is Francisco Rios Francisco Rios from RI-MUHC (Montreal, CA). #hypHIP

#HYPHIP Q2 (apologies): In the past less aggressive BP target values have been recommended for diabetics (e.g. Mancia et al, Diabetologica 2018). Can you comment and put into context (j-curve vs. linear relationship).

Shishir Rao
Kazem Rahimi
Société Française d'Hypertension Artérielle
Hypertension

Georg Ehret, MD FESC FAHA Shishir Rao Société Française d'Hypertension Artérielle Hypertension #HYPHIP A2: Previous recommendations have in part been due to limited evidence from trial, and traditional observational studies showing a J-curve relationship (ie, implying that lowering the BP below a level might be harmful).

Welcome Fran ! Fantastic !

#HYPHIP Shishir Rao Kazem Rahimi

#HYPHIP Q3: In a nutshell: what did you discover regarding the J-shaped relationship of BP and outcome ?

Shishir Rao @kazemr@HyperAHA

Hi everyone! This is Guto Montezano from the @RIMUHC in Montreal! Excited to learn more about the #HYPHIP from Shishir Rao Kazem Rahimi Hypertension Thank you for hosting us Georg Ehret, MD FESC FAHA

#HYPHIP Q4:You did great work to overcome potential shortcomings of observational studies using new methods. Can you outline the most important biases of the previous methods for us please ?

Shishir Rao Kazem Rahimi Hypertension

Georg Ehret, MD FESC FAHA #HYPHIP A3: We learned that the J-shaped curve found in previous studies (which we replicated in our work) was a product of insufficient confounding adjustment. When more comprehensive confounding adjustment was conducted using AI modelling on rich EHR, the J-shape vanished!