Lumiata, a healthcare analytics company, said it will launch a tool called Risk Matrix that takes a patient's electronics record, and combines it with artificial intelligence and risk algorithms to predict health over time for people with chronic conditions.
The software, designed for insurers and healthcare providers, is designed to analyze models based on 175 million patient record years and provide clinical rationale for each prediction.
According to Lumiata, the data should be used to give confidence in clinical decisions. Risk Matrix provides real-time predictions for more than 20 major diseases including congestive heart failure and diabetes.
Risk Matrix first analyzes customer datasets such as health records and lab results. Lumiata, which develops its artificial intelligence via its Lumiata Medical Graph, collects the data and uses the Fast Healthcare Interoperability Resources standard to share the information.
The data is then run through the Risk Matrix deep learning model to generate personalized assessments. Once data is mapped Risk Matrix can assess more than 1 million records in less than three hours.
Healthcare customers can use application programming interfaces to incorporate Risk Matrix predictions into their existing tools.
Lumiata was founded in 2013 and has a team of clinicians, data scientists, and care delivery.
Dr. C. Anthony Jones, Lumiata's chief commercial officer, said the company is looking to apply AI to make better predictions. Insurers and companies are offering programs for chronic conditions, but often identify too many people as risks.
"The primary thing we're trying to do is to give precision on how health is likely to change for people in the next 12 to 18 months," said Jones. "We're offering a higher level of personalization to make decisions.
Jones added that measurements such as BMI don't offer much in the way of making predictions. Lumiata also shows its work to its clients so they can see how predictions were made.
So far, Risk Matrix is in two pilots right now and once they are finished there will be more volume in the models. These pilot customers are also exploring ways to share their results to individuals via a portal. "The early results have been good," said Jones.