Researchers at WellDoc® today presented updated results of its hypoglycemia prediction technology at this week’s 2014 American Diabetes Association 74th Scientific Sessions. In previous work, using just one week of blood glucose data, WellDoc’s models were shown to predict correctly 90 percent of the time that hypoglycemia would occur the following day. Today, WellDoc announced an enhanced model that incorporates both glucose and medication data has demonstrated the ability to predict hypoglycemia within a specific hour.
“The challenge in predicting hypoglycemia is that most patients with type 2 diabetes measure their blood glucose once or twice per day. This so-called ‘sparse data’ makes mathematical forecasting difficult,” states WellDoc Chief Data Science Officer, Anand Iyer, Ph.D. “We used machine learning algorithms which allow the computer to detect patterns and make predictions after being trained on thousands of data points.”