Thread by @dvir_a

Getting ready to #DMCI after dark from the top of Tel-Aviv. I’ll try to live tweet the talks.

View from the venue.

Udi Goori, Anthem Israel general manager, opening #DMCI. Why are we here? To connect academia and industry to solve the pressing healthcare problems.

The #DMCI community

Dr. Yona Vaisbuch @RambamHCC moderating the meeting.

Dr. Noa Dagan @ClalitResearch @BenGurionUni talking about prediction models in healthcare (no COVID19 allowed). Integrating models in the clinic is an ongoing struggle. Clalit developed novel platform to solve this.

Calibrating is a critical performance measure. Calibration may be noisy on subpopulation. Developed a multicalibration method to solve this. ncbi.nlm.nih.gov/pmc/articles/P…

How do we choose candidates for intervention? Should we add features to increase diversity? No good answers yet, but we have to be accountable and fair.

Now up Dr. David Golan #VizAI on identifying stroke as soon as possible. If we capture it early we can save from its devastating outcomes. Every minute of delay 1.9 neurons die, 4.2 days of disability, $1060 additional costs.

How do you evaluate the impact of an AI solution? For causality you need an RCT, but hard expensive and logistical nightmare. Here isn’t really possible to randomize. Stepped wedged cluster RCT can solve this - randomly install the technology in hospitals.

Testing the prediction model across all hospitals, not just top academic centers. Important to understand that quality of data might be very different.

Dr. Irit Gat-Viks @TelAvivUni on creating baseline clinical states of health to develop risk predictors of complex diseases.

Its getting dark outside. Tel-Aviv lights from above.

The 2d map of clinical states can differentiate between patients in different conditions. Focused on atherosclerosis and found that after correcting to its clinical states the biomarkers are not indicative.

Ending with a panel with the speakers. How to go from basic research to clinical impact?