GenHealth.ai unveils medical model, surpassing industry benchmarks

GenHealth.ai has published a study detailing an advanced generative AI model that decisively outperforms established industry benchmarks by over 14% on healthcare cost and risk prediction. This model challenges traditional approaches by firms like Milliman, Cotiviti, and Johns Hopkins, said the company, “setting a new standard for precision in healthcare prediction while enabling an unprecedented lens into the future of a patient’s healthcare journey”.

The large medical model (LMM) combines a unique vocabulary tailored specifically for healthcare, the technology of neural network transformers (which underpins large language models), and data from trillions of healthcare events across 140 million patients. The resulting model achieves a similar leap in performance for the healthcare domain akin to the improvement in LLMs over traditional natural language processing approaches.

Ricky Sahu, CEO of GenHealth.ai, commented: “We’ve created the large medical model to address a broad range of healthcare applications where LLMs and traditional analytics fall short. Our healthcare-specific AI is built from the ground up to support everything from pop health analytics, to automating prior authorisations, and detecting fraud/waste/abuse.”

Highlights of the study include:

• A new tokenisation scheme: the paper introduces a new use of generative AI on healthcare-specific data and tokens to predict patient futures holistically
• State of the art (SOTA) performance on cost prediction achieves unprecedented accuracy in predicting patient total cost of care and risk factors, substantially surpassing legacy systems used by current industry leaders
• SOTA performance on chronic condition prediction: in addition to predicting total cost of care more accurately, the paper details how the same model is used to predict a wide variety of chronic conditions
• Impact on healthcare: by delivering more accurate predictions at an event level detail, the AI model promises to be more actionable and explainable. These features help insights make it from research into practice to reduce wasteful spending and improve patient management.

Healthtech partner Spectrum.Life recently announced a strategic rebrand to a clinically backed digital health, mental health, and wellbeing solutions provider.