The CMS Artificial Intelligence (AI) Health Outcomes Challenge was an opportunity for innovators to demonstrate how AI tools – such as deep learning and neural networks – can be used to accelerate development of AI solutions for predicting patient health outcomes for Medicare beneficiaries for potential use in CMS Innovation Center innovative payment and service delivery models.
The Challenge was operated by CMS in partnership with the American Academy of Family Physicians and Arnold Ventures.
Challenge Objectives
- For Stage 1, use AI, including but not limited to deep learning methodologies, to predict unplanned hospital and SNF admissions, and adverse events within 30 days for Medicare beneficiaries, based on a data set of Medicare administrative claims data, including Medicare Part A (hospital) and Medicare Part B (professional services).
- For Stage 2, use AI, including but not limited to deep learning methodologies, to predict unplanned hospital and SNF admissions, and adverse events, within 30 days for Medicare beneficiaries, as well as 12-month mortality for all Medicare beneficiaries, based on a Part A and Part B data set.
- For both Stage 1 and Stage 2, develop innovative strategies and methodologies to: explain the AI-derives predictions to front-line clinicians and patients to aid in providing appropriate clinical resources to model participants; and increase use of AI-enhanced data feedback for quality improvement activities among model participants.
Participants also were required to address implicit algorithmic biases that impact health disparities in their submissions.
Winner & Runner-Up
CMS announced 2 Participants as the competition winner and runner-up on April 30, 2021.
Winner
Runner-Up
- Stage 2
CMS announced 7 Participants to advance to Stage 2 on October 29, 2020. The 7 Participants, titles of proposed solutions and geographic locations are listed below:
Participant: Ann Arbor Algorithms Inc.
Proposed Solution: Generalizing Time-to-event Algorithms to Deep Learning-based Prediction for CMS Data
Geographic Location: Sterling Heights, MichiganParticipant: ClosedLoop.ai
Proposed Solution: Healthcare's Data Science Platform
Geographic Location: Austin, TexasParticipant: Consulting LLP
Proposed Solution: Further, Faster: The Deloitte Team’s Approach to Harnessing the Power of AI to Improve Health Outcomes
Geographic Location: Arlington, VirginiaParticipant: Geisinger
Proposed Solution: Reducing Adverse Events and Avoidable Hospital Readmissions by Empowering Clinicians and Patients
Geographic Location: Danville, PennsylvaniaParticipant: Jefferson Health
Proposed Solution: Using AI to Improve Medicare Population Health, Optimize Ambulatory Scheduling, and Reduce Adverse Events at Hospitals
Geographic Location: Philadelphia, PennsylvaniaParticipant: Mathematica Policy Research, Inc.
Proposed Solution: The CPC+ AI Model by Mathematica
Geographic Location: Princeton, New JerseyParticipant: University of Virginia Health System
Proposed Solution: Actionable AI
Geographic Location: Charlottesville, Virginia
- Stage 1
CMS announced 25 Participants to advance to Stage 1 on October 30, 2019. The 25 Participants, titles of proposed solutions and geographic locations are listed below:
Participant: Accenture Federal Services
Proposed Solution: Accenture Federal Services AI Challenge
Geographic Location: Arlington, VirginiaParticipant: Ann Arbor Algorithms Inc.
Proposed Solution: Generalizing Time-to-event Algorithms to Deep Learning-based Prediction for CMS Data
Geographic Location: Sterling Heights, MichiganParticipant: Booz Allen Hamilton
Proposed Solution: Booz Allen Launch Stage Submission
Geographic Location: McLean, VirginiaParticipant: ClosedLoop.ai
Proposed Solution: Healthcare's Data Science Platform
Geographic Location: Austin, TexasParticipant: Columbia University Department of Biomedical Informatics
Proposed Solution: The CLinically Explainable Actionable Risk (CLEAR) Model from Columbia University Department of Biomedical Informatics
Geographic Location: New York, New YorkParticipant: CORMAC
Proposed Solution: CORMAC Response to Challenge Questions
Geographic Location: Columbia, MarylandParticipant: Deloitte Consulting LLP
Proposed Solution: Further, Faster: The Deloitte Team’s Approach to Harnessing the Power of AI to Improve Health Outcomes
Geographic Location: Arlington, VirginiaParticipant: Geisinger
Proposed Solution: Reducing Adverse Events and Avoidable Hospital Readmissions by Empowering Clinicians and Patients
Geographic Location: Danville, PennsylvaniaParticipant: Health Data Analytics Institute
Proposed Solution: HDAI’s Analytic Platform Technology for Healthcare Improvement
Geographic Location: Dedham, MassachusettsParticipant: HealthEC, LLC
Proposed Solution: Leveraging Artificial Intelligence to Predict and Improve Health Outcomes, Maximize Quality Improvement, and Reduce Costs
Geographic Location: Edison, New JerseyParticipant: Hospital of the University of Pennsylvania
Proposed Solution: The Intelligent Risk Project
Geographic Location: Philadelphia, PennsylvaniaParticipant: IBM Corporation
Proposed Solution: AI for Explainable Adverse Event Prediction: Empowering Beneficiaries and Providers to Improve Health Outcomes
Geographic Location: Yorktown, New YorkParticipant: Innovative Decisions Inc. (IDI)
Proposed Solution: Multi-Modeling with Augmented Datasets for Positive Health Outcomes (MADPHO)
Geographic Location: Vienna, VirginiaParticipant: Jefferson Health
Proposed Solution: Using AI to Improve Medicare Population Health, Optimize Ambulatory Scheduling, and Reduce Adverse Events at Hospitals
Geographic Location: Philadelphia, PennsylvaniaParticipant: KenSci Inc.
Proposed Solution: Assistive Intelligence for Unplanned Admissions and Adverse Events Prediction
Geographic Location: Seattle, WashingtonParticipant: Lightbeam Health Solutions, LLC
Proposed Solution: AI Risk Predictions- preventing hospital, ER and SNF admissions
Geographic Location: Irving, TexasParticipant: Mathematica Policy Research, Inc.
Proposed Solution: The CPC+ AI Model by Mathematica
Geographic Location: Princeton, New JerseyParticipant: Mayo Clinic
Proposed Solution: Claims-based Learning Framework (CBLF)
Geographic Location: Rochester, MinnesotaParticipant: Mederrata
Proposed Solution: Boosting medical error and readmission prediction by leveraging Deep Learning, Topological Data Analysis, and Bayesian modeling
Geographic Location: Columbus, OhioParticipant: Merck & Co., Inc.
Proposed Solution: Actionable AI to Prevent Unplanned Admissions and Adverse Events
Geographic Location: Kenilworth, New JerseyParticipant: North Carolina State University (NCSU)
Proposed Solution: Multi-Layered Feature Selection and Dynamic Personalized Scoring
Geographic Location: Raleigh, North CarolinaParticipant: Northrop Grumman Systems Corporation (NGSC)
Proposed Solution: Reducing Patient Risk through Actionable Artificial Intelligence: AI Risk Avoidance System (ARAS)
Geographic Location: Herndon, VirginiaParticipant: Northwestern Medicine
Proposed Solution: A human-machine solution to enhance delivery of relationship-oriented care
Geographic Location: Chicago, IllinoisParticipant: Observational Health Data Sciences and Informatics (OHDSI)
Proposed Solution: OHDSI Submission
Geographic Location: New York, New YorkParticipant: University of Virginia Health System
Proposed Solution: Actionable AI
Geographic Location: Charlottesville, Virginia
COVID-19 3-Month Pause
In order to support providers responding to the COVID-19 Public Health Emergency (PHE), the Centers for Medicare & Medicaid Services (CMS) temporarily relaxed some requirements for health care providers participating in innovative payment and service delivery models and other initiatives. By relaxing certain initiative requirements, in combination with the other efforts already underway by other federal agencies, CMS worked to better support health care providers in directing their resources towards caring for patients, ensuring the safety of staff, and reducing the spread of transmission. In order to accommodate those priorities, CMS temporarily paused the Artificial Intelligence Health Outcomes Challenge (the Challenge) for three months, and restarted the Challenge on Monday, June 29, 2020. The Challenge resumed normal activities at that time.
Learn More
- Listen to the May 12, 2021 blog
- Watch the April 18 Informational Overview webinar
- Official public notice (PDF) (PDF)of the Challenge (Updated August 2020) (CMS reserves the right to change or update the Challenge requirements at any time.)
Prizes (subject to change)
Total prizes up to $1.65 million
- 7 finalists progressed to Stage 2 and received awards of up to $60,000
- 1 grand prize winner received up to $1 million and the runner-up received up to $230,000