A Machine Learning Approach for Continuous Mining of Non-identifiable Smartphone Data to Create a Novel Digital Biomarker Detecting Generalized Anxiety Disorder: A Feasibility Study (Preprint)

Abstract

BACKGROUND Anxiety is one of the leading causes of mental health disability around the world. Currently, a majority of the population who experience anxiety go undiagnosed or untreated. New and innovative ways of diagnosing and monitoring anxiety have emerged using smartphone sensor-based monitoring as a metric for the management of anxiety. This is a novel study as it adds to the field of research through the use of non-identifiable smartphone usage to help detect and monitor anxiety remotely and in a continuous and passive manner. OBJECTIVE The aim of this study is to demonstrate the feasibility of a daily mental health behavioural profiling metric called Mental Health Similarity Score for anxiety generated by mining non-identifiable smartphone data. METHODS Smartphone data and self-reported 7-item GAD anxiety assessments were collected from 229 participants on the Android operating system in an observational study over an average of 14 days (SD=29.8). 34 features were mined to be constructed as a potential digital phenotyping marker from continuous smartphone usage data. We further analyzed the correlation of these digital behavioral markers against each item of the GAD-7 and its influence on the machine learning algorithms predictions. RESULTS A total of 229 participants were recruited in this study that had completed the GAD-7 assessment and had at least one set of passive digital data collected within a 24-hour period. The mean GAD-7 score was 11.8 (SD=5.7). Regression modeling was tested against classification modeling and the highest prediction accuracy was achieved from a binary Random Forest classification model (Precision 89-92%; Recall 85-94%; F188-91%; accuracy 90%; AUC 96%). Non-parametric permutation testing with Pearson correlation results indicated the proposed metric (MHSS) had the strongest relationship between GAD-7 items 1,3, and 7. CONCLUSIONS The proposed Mental Health Similarity Score (MHSS) metric demonstrates the feasibility of using passive non- intrusive smartphone data and machine learning-based data mining techniques to track an individual’s daily anxiety levels with a 90% accuracy that directly related to the Generalized Anxiety Disorder-7 scale.