Nassib Chamoun
Founder, President & CEO
Earlier this year, the Centers for Medicare and Medicaid Services (CMS) announced the AI Health Outcomes Challenge, a national initiative to harness healthcare’s collective creativity to develop better ways of predicting and preventing unplanned admissions and complications. In our application, the Health Data Analytics Institute (HDAI) outlined how we would address these adverse events, which lead to billions of dollars in unnecessary spending every year—as well as immeasurable anguish for patients and their families. More than 300 other organizations submitted their proposed approaches as well.
Today, I’m thrilled that CMS has selected HDAI as one of the 25 participants in the next phase of the Challenge. Between now and February, we will develop an analytic proof of concept for predictors of certain high-value adverse events, along with written guidance outlining how these models can improve the delivery of care in hospitals and physicians’ offices around the country.
In our application to CMS, we wrote in detail how we intend to build and deploy highly accurate, calibrated, and transparent predictive models for critical healthcare outcomes. I am eager to learn about the approaches others plan to take, and I’m sure each of us will bring a different analytic approach and clinical perspective.
I re-read our initial application since CMS invited us to move forward. The animating beliefs we described in that application resonate with me today more than ever, and they will remain our touchstones as we move forward in the Challenge:
Success measured by real-world clinical utility and patient impact, not raw AI performance. We aspire to develop predictive models that supplement and enhance a clinician’s judgment, enabling timely and targeted interventions and more informed shared decision-making. AI models are only as valuable as the clinical improvements they enable.
Solutions created in partnership. Measurable improvements in cost and outcomes will only come from close collaboration with the full constellation of hospitals, clinicians, and staff members who influence the course of a patient’s care. We are honored to partner with the Cleveland Clinic, New England Baptist Hospital, and others on a pilot to deploy AI predictors to prevent or improve the response to adverse events in the perioperative setting.
Transparency in AI models. Each prediction we generate will include its underlying drivers, so clinicians know not just a patient’s risk of an outcome but also how it can be managed. We are committed to transparency of performance as well, and we will share full operating characteristics for each model so providers know when and for whom they work, as well as the limitations of the models as they arise.
Commitment to activating researchers. We will make all of our predictive models available to the research community—at no cost—via an API. We can help academics and public health experts unlock the insights in their research databases by providing baseline predictions of cost and risk, allowing more precise measurement of the effects of novel policies and interventions.
The change we are all hoping for in healthcare will require renewed collaboration, transparency, and data-driven intelligence. We’re excited and grateful for the opportunity to help move the industry forward as part of the next phase of the CMS AI Challenge!
-Nassib
