One Drop uses predictive analytics for individual users
One Drop presented long-term A1C, blood pressure and weight outcomes forecasts for individuals, along with overnight hypoglycemia predictions for CGM users based on the use of its One Drop app.
The app collects health data through manual input, automatic upload from connected devices, HealthKit, GoogleFit, and direct integrations with Dexcom, Fitbit, Apple Watch, Companion Medical and other partners. Users can receive 8-hour glucose forecasts along with real-time advice for exercise, diet and lifestyle adjustments.
New York-based One Drop noted that it has collected more than 11 billion health data points from nearly 3 million One Drop app users in 195 countries. The company uses machine learning to analyze this pooled population data to produce its predictive insights.
In a study presented at the conference, One Drop trained a suite of patent-pending supervised learning models to accurately forecast changes in glucose, blood pressure and weight up to 6 months in advance. Sources for the study included health and self-care data from a sample of approximately 55,000 users. Another study focused on overnight hypoglycemia predictions for people using CGMs. The patent-pending overnight hypoglycemia risk algorithms, trained on over 500,000 nights of data from app users, are up to 87% accurate, the company said.
“The real power of our predictive models is how individualized they are,” said Dan Goldner, One Drop VP of data science, in a news release. “We don’t bother looking for broad patterns of factors that might help in distinct groups. Instead, we jump straight to predicting the factors most likely to help you.”
In April, One Drop acquired the assets and intellectual property of Sano Intelligence and announced plans to develop a multisensorial patch that will generate thousands of new data points per user per day. The company expects to release new predictions to subscribers within the year.