OHDSI Home | Forums | Wiki | Github

OHDSI collaboration opportunity for the CMS AI Challenge

(Patrick Ryan) #1


As an FYI: last week, on behalf of our community, @jennareps, @Rijnbeek, @hripcsa and I submitted an OHDSI entry into the CMS AI Health Outcomes Challenge (https://www.cmschallenge.ai/). The CMS Challenge “is an opportunity for innovators to demonstrate how AI tools, such as deep learning and neural networks, can be used to predict unplanned hospital and skilled nursing facility (SNF) admissions and adverse events. The challenge will prioritize explainable AI solutions, which can work with front-line clinicians and patients to improve data feedback and drive quality improvements.” Our submission focused on the differentiated strengths of our community, including the PatientLevelPrediction framework and open-source software that we have developed to support analyses on any observational data using the OMOP common data model, and our ability to train and externally validate large-scale prediction models across our international data network. We thought this competition is very nicely aligned with OHDSI’s mission and also a fun opportunity to collaborate on a shared problem of public health importance.

It appears the CMS Challenge has been a quite popular competition, so it will be interesting to see if our OHDSI open-science approach stands out amongst the rest of the submissions. We should find out in early August if we’ve been selected to be one of the 25 Stage 1 participants invited to train models against the CMS claims data and present our results. That’ll be the point where the fun will really start! If chosen, we’ll coordinate a kickoff with anyone in our community who is interested in joining the journey through model development and evaluation. If you would like to participate, just let us know by replying to this thread.

The CMS AI Challenge comes with cash prizes for the winners. While the odds are against us, there’s no harm in planning for the positive: should we advance in the competition, we will use any prize money awarded to support our community’s Patient-Level Prediction workgroup activities. And if we win the $1 million grand prize, then you can bet we will have one heckuva fun OHDSI face-to-face meeting focused on prediction for all those who participate.

Fingers crossed, and I will keep you posted when we hear more. I look forward to collaborating with all of you on this exciting opportunity!

(Benjamin Skov Kaas Hansen) #2

Hi Patrick et al.,

This is such a cool endeavour, and I’d love to join the fun! How could anyone not?
As an MD, I really think there’s too many useless AI pseudo-solutions out there that break when faced with real life, and I’d really love for the truly open OHDSI approach to show that it doesn’t need to be like that.

And again, what an awesome opportunity.