Princeton University, ODH collaborate on machine learning tools
Applications will seek to find health factors that doctors may miss, identify high-risk members for insurers.
A department at Princeton University has teamed up with data aggregation and analytics vendor ODH to develop new machine learning techniques to help health insurers assess and prioritize mental and social factors underlying members’ conditions, and then propose appropriate interventions.
Work at the university is being conducted by its Operations Research and Financial Engineering Department.
Also See: Radiation departments’ adoption of machine learning accelerates
Machine learning is an artificial intelligence technology that enables computer systems to learn without being explicitly programmed. The technology also supports the scanning of large amounts of data and then uses pattern recognition to make predictions about future patient outcomes.
For example, using traditional medical processes, a physician may assess a patient’s symptoms and conclude a diagnosis of liver failure. But techniques being developed by Princeton and ODH will identify a range of secondary factors overlooked by the initial examination, such as alcohol addiction, other substance abuse and social challenges, which could result in a trip to the hospital or emergency department.
Consequently, additional techniques being developed would help care coordinators identify patients likely to respond to an intervention such as entering a substance abuse treatment program, providing education on medication adherence and offering transportation to a physician office.
“Predictive tools have to be simple enough to yield interpretable health recommendations but also must achieve high accuracy,” says Samory Kpotufe, assistant professor at Princeton. “So, there’s a tension between accuracy and interpretability that has to be addressed, along with practical constraints such as computational tractability, environmental and seasonal changes and patients’ privacy.”
Work at the university is being conducted by its Operations Research and Financial Engineering Department.
Also See: Radiation departments’ adoption of machine learning accelerates
Machine learning is an artificial intelligence technology that enables computer systems to learn without being explicitly programmed. The technology also supports the scanning of large amounts of data and then uses pattern recognition to make predictions about future patient outcomes.
For example, using traditional medical processes, a physician may assess a patient’s symptoms and conclude a diagnosis of liver failure. But techniques being developed by Princeton and ODH will identify a range of secondary factors overlooked by the initial examination, such as alcohol addiction, other substance abuse and social challenges, which could result in a trip to the hospital or emergency department.
Consequently, additional techniques being developed would help care coordinators identify patients likely to respond to an intervention such as entering a substance abuse treatment program, providing education on medication adherence and offering transportation to a physician office.
“Predictive tools have to be simple enough to yield interpretable health recommendations but also must achieve high accuracy,” says Samory Kpotufe, assistant professor at Princeton. “So, there’s a tension between accuracy and interpretability that has to be addressed, along with practical constraints such as computational tractability, environmental and seasonal changes and patients’ privacy.”
More for you
Loading data for hdm_tax_topic #better-outcomes...