Atropos Health and Novartis roll out AI trained on real-world data to aid rare disease diagnosis

Atropos Health, in collaboration with Novartis, has developed an AI model trained on electronic health record (EHR) data from 67 million patients to help identify individuals who may warrant further workup for the rare disease, paroxysmal nocturnal hemoglobinuria (PNH).

Atropos Health has announced a collaboration with Novartis to apply AI trained on real-world data (RWD) to improve the detection of rare diseases, beginning with PNH.

The initiative has produced a machine learning algorithm, available for integration into health systems, designed to identify patients who may be considered for diagnostic assessment for PNH, a rare, serious blood disorder that often goes undiagnosed for years due to its varied clinical presentation. By leveraging large-scale EHR data covering 67 million patients across the US, the AI model aims to reduce delays from symptom onset to testing and treatment.

To ensure robustness, the training dataset was restricted to records from 2016–2023 to allow sufficient follow-up. While more than 350 candidate predictive features were evaluated, the final model incorporated 13 variables that achieved performance comparable to the full feature set. This streamlined approach supports integration into clinical workflows while maintaining predictive accuracy.

“Building AI models tested and trained on high-quality real-world data is the next frontier in precision medicine,” said Dr Brigham Hyde (CEO and Co-Founder, Atropos Health). “The accuracy of the models reduces the guesswork and patients who are able to get testing sooner provides a potentially life-changing experience. For providers and health systems, faster time to diagnosis and treatment equate to higher patient satisfaction.”

The model is now available for deployment within health systems via the Atropos Evidence Network and its GENEVA OS® platform, which allows integration of predictive algorithms at the point of care. Future models are planned to address additional rare conditions.

For Novartis, the collaboration underscores a broader commitment to precision health. Rodney Gillespie, Head of Oncology at Novartis US, explained:

“Accelerating diagnosis and treatment through AI and machine learning has the power to significantly improve patient outcomes by enabling faster access to appropriate care.”