Using Data to Steer Health Outcomes

By Sandy K. Johnson

Thirty million Americans have diabetes. Another 83 million are pre-diabetic – and many don’t know it. Imagine if artificial intelligence could help identify people at risk for diabetes and steer them into preventive care or treatment.

It’s already happening. Health companies like Optum are crunching data from insurance claims forms, consumer purchases, electronic medical records, prescription data, social determinants of health and more.

“Those become the ingredients for the different models we build” in predictive analytics, said Steve Griffiths, senior vice president & COO at Optum Enterprise Analytics.

Diabetes is a good example. It has a huge societal cost – to the tune of $322 billion a year. Machine learning, using 290 different markers, can identify people at risk of diabetes with 87 percent accuracy, said Griffiths, a biostatistician by training. On an individual level, a nurse might get the results and call a patient to come in for further testing.

Also good candidates for AI application are chronic diseases like heart failure, hypertension and asthma.

Griffiths said AI can bring speed and efficiency to health care.  On the flip side, health care professionals can be slow to adopt change. He said it typically take 14 to 17 years to change clinical practices.

Are there negatives to AI in health? Griffiths suggested chatbots and virtual assistants providing services and treatment recommendations – something that he said is not likely to happen.

Subscribe on YouTube
More Presentations
Help Make Good Journalists Better
Donate to the National Press Foundation to help us keep journalists informed on the issues that matter most.
DONATE ANY AMOUNT