The advent of technology and mobile health apps has transformed the way people monitor, manage, and communicate health-related information. For pain, “Manage My Pain” is a mobile health app used by thousands of individuals to measure and manage their pain.
It also has the capacity for large-scale, real-world data collection to advance pain research, treatment, and policy. In particular, data mining and machine learning methods can be used to analyze multimodal and dynamic features as well as build models for prediction. This may ultimately help us understand how pain changes across time within users and to allow for the development of effective coping strategies.
Using data collected from users of the Manage My Pain app, researchers at York University, including UTCSP members Drs. Joel Katz and Hance Clarke, and corresponding author Tahir Janmohamed, defined a new pain volatility measure and used machine learning to predict future pain volatility levels. Pain volatility measures the fluctuation or variability in pain over time and it is important because it takes into account intra- and interindividual differences in pain that may not be captured using the mean or median. In this study, pain volatility was newly defined as the mean of absolute changes in pain severity. A total of 782 users with 329,070 pain records were included in the study.
The authors collected 130 features from the first month of engagement with the app to predict high versus low levels of pain volatility at the sixth month mark of app engagement. They found that the prediction model using Random Forests performed the best – an approximate 70% accuracy level was achieved for both high and low volatility classes. The results of this study point to how large, real-world datasets collected from users of mobile health apps provide an important opportunity to study clinical syndromes and identify strategies to better measure and manage health symptoms. Specifically, the development of prediction models using machine learning may allow earlier intervention to prevent the development of critical health symptoms such as high pain volatility.
Pain volatility measures the fluctuation or variability in pain over time and it is important because it takes into account intra- and interindividual differences in pain that may not be captured using the mean or median. In this study, pain volatility was newly defined as the mean of absolute changes in pain severity.
