Machine Learning in Medication Adherence Detection

Artificial Intelligence (AI) and Machine Learning (ML) research and its application in various fields is reaching new heights everyday. There are numerous applications of these advanced technology from consumer products e.g. Apples FaceID, Personalization (recommendation systems) and education. A recent research published in Journal of Diabetes Science and Technology investigated ML based approach of measuring and detecting adherence in diabetes patients using once-daily basal insulin injections with an ultimate goal of developing an early alarm system for medication adherence detection. 

Authors have utilized data obtained from in-silico continuous glucose monitoring (CGM) technology to simulate a cohort of type 2 diabetes patients on a daily dose of insulin injection. Results generated from a state-of-the-art ML modeling technique called “Deep Learning” and simple feature-engineered ML classification models were compared. Following observation was made by the authors –

“The three classification models based on expert-engineered features obtained mean accuracies of 78.6%, 78.2%, and 78.3%. The classification model based purely on learned features obtained a mean accuracy of 79.7%. The two classification models fusing expert-engineered and learned features obtained mean accuracies of 79.7% and 79.8%. All the mentioned results were obtained 16 hours after time of injection.”

“..there is a clear benefit of using simple models in contrast to more complex models.”

This research can, if reproduced in real world CGM data, open new possibilities for implementing early adherence detention tools for helping decision makers and patients alike. E.g. output of the model (i.e. adherence alarm) can be used to create a notification system in the health care management apps by providers. Scaling of this application would benefit from richer real world dataset to further improve the model