Data science team’s research is published after presenting at AI conference in Stockholm
TORONTO, ON/NEW YORK, NY – August 29, 2018 – Klick Health today announced that its team of data scientists has developed algorithms to predict blood glucose levels for patients with Type 1 diabetes using gradient-boosted trees as a competitive predictor.
Their research paper, ‘Predicting Glycemia in Type 1 Diabetes Patients: Experiments with XGBoost’ could lay the groundwork for assessing blood glucose levels 30 minutes into the future.
It was published after being presented at the The 3rd International Workshop on Knowledge Discovery in Healthcare Data at the 2018 International Joint Conference on Artificial Intelligence (IJCAI) in Stockholm, Sweden and can be viewed at http://ceur-ws.org/Vol-2148/paper13.pdf.
Authored by Klick’s Cooper Midroni, Peter Leimbigler, Gaurav Baruah, Maheedhar Kolla, Alfred Whitehead, and Yan Fossat, the research is based on the well-validated OhioT1DM dataset containing eight weeks of self-reported data from six patients with Type 1 diabetes. The data includes physiological fitness wristband data, self-reported information, continuous glucose monitoring, and insulin pump data.
The Klick team‘s unique approach to the research included using the XGBoost open-source gradient-boosting software library (a technique applied more in industry than in research), tuning the model using clinically-relevant criteria taken from medical science, and assessing the fitness band’s effect on prediction accuracy.
“We hypothesized that the fitness band could be used as part of a digital biomarker for blood glucose but we ended up proving the opposite,” said Alfred Whitehead, Klick’s SVP, Data Science. “The wearables data induced noise and misled the algorithm, so by discarding the data we were able to improve the accuracy of our algorithms. Thanks to the relentless work of our data science team, we established that past and present blood glucose levels along with self-reported data on meals are the best predictors.”
Today’s news puts the spotlight on the diverse team of data scientists, biomedical engineers, and social scientists at Klick Labs. It also highlights the company’s ongoing investment in research and AI solutions as part of its work to help improve the well-being of patient populations with its proven commercial, scientific, medical and technological expertise.
Klick Health is the world’s largest independent, full-service marketing and commercialization agency for life sciences. By relentlessly conspiring with the future, Klick helps its clients outsmart and outpace disease so people can live richer, longer lives. Klick has been named Agency of the Year five times over the last six years by the industry’s leading publications and was recognized by the 2018 Cannes Lions Health Awards for its client campaigns. Follow Klick Health on Twitter at @KlickHealth.
About Klick Inc.
Klick operates with a fiercely independent spirit at the intersection of technology, creativity, and strategy. The Klick family is a diverse group of people who are driven and connected by a shared culture and commitment to the relentless pursuit of awesome at Klick Health, Katalyst, and Sensei Labs. Established in 1997, Klick has teams in Atlanta, Connecticut, Chicago, Los Angeles, New Jersey, New York, Philadelphia, San Diego, San Francisco, and Toronto. Klick has been consistently recognized as a leading Great Place to Work, Best Workplace for Women, Best Employer, Employee-Recommended Workplace, Fastest Growing Technology Company, and Best Managed Company. For more information on joining Klick, go to http://careers.klick.com.