Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Wearable and mobile devices that capture multi-modal data have the potential to identify risk factors for self-reported high stress and poor mental health and provide information to improve health and wellbeing. Objective: We developed new tools that provide objective, physiological, and behavioral measures using wearable sensors and mobile phones, together with methods that improve their data integrity. Our goal was to examine, using machine learning, how accurately these measures can identify conditions of self-reported high stress and poor mental health, and which of the underlying modalities and measures are most accurate in identifying the conditions. Methods: We designed and conducted the one month SNAPSHOT study whose goal was to investigate how daily behaviors and social networks influence stress, mood, and other health/wellbeing-related factors. We collected over 180,000 hours of data from 251 college students, all recruited within self-identified social groups. For approximately one month, each student filled out standardized pre- and post-questionnaires on mental health and stress; completed twice-daily e-diaries; wore two wrist-based sensors that recorded continuous physical activity, light exposure and autonomic physiology; and installed an app on their phone that recorded call and texting behaviors, and geolocation patterns. We developed tools to make data collection more efficient including (i) automatic sensor and mobile phone data-check systems and (ii) an e-diary administrative module for study investigators to locate possible errors in the e-diaries and communicate with participants to correct their entries promptly that reduced the time taken to clean e-diary data by 69%. We constructed features and applied machine learning techniques to the multimodal data to identify factors associated with self-reported stress and mental health reported at the end of the study, including behaviors that can be possibly modified by the individual to improve these measures. Results: We identified physiological sensor, phone, and mobility features that were best predictors for classifying self-reported high/low stress and mental health. In general, features derived from wearable wrist-worn sensors performed better than features derived from mobile phone or modifiable behaviors. Features derived from wearable sensors, including skin conductance and temperature, reached 78% accuracy for classifying students into high/low stress groups and 86% accuracy for classifying high/low mental health groups. Features derived from modifiable behaviors, including number of naps, studying duration, phone calls, mobility patterns, and phone-screen-on time, reached 74% accuracy for classifying students into high/low stress groups and 78% accuracy for classifying high/low mental health groups. Conclusions: New semi-automated tools improved the efficiency of long-term ambulatory data collection using mobile and wearable devices. Applying machine learning to the resulting data revealed a set of both objective features and modifiable behavioral features that can classify self-reported high/low stress and mental health groups in a college student population better than previous studies and showed new insights into digital phenotyping.

While accurately predicting mood and wellbeing could have a number of important clinical benefits, traditional machine learning (ML) methods frequently yield low performance in this domain. We posit that this is because a one-size-fits-all machine learning model is inherently ill-suited to predicting outcomes like mood and stress, which vary greatly due to individual differences. Therefore, we employ Multitask Learning (MTL) techniques to train personalized ML models which are customized to the needs of each individual, but still leverage data from across the population. Three formulations of MTL are compared: i) MTL deep neural networks, which share several hidden layers but have final layers unique to each task; ii) Multi-task Multi-Kernel learning, which feeds information across tasks through kernel weights on feature types; and iii) a Hierarchical Bayesian model in which tasks share a common Dirichlet Process prior. We offer the code for this work in open source. These techniques are investigated in the context of predicting future mood, stress, and health using data collected from surveys, wearable sensors, smartphone logs, and the weather. Empirical results demonstrate that using MTL to account for individual differences provides large performance improvements over traditional machine learning methods and provides personalized, actionable insights.

Recently, wearable devices have allowed for long term, ambulatory measurement of electrodermal activity (EDA). Despite the fact that ambulatory recording can be noisy, and recording artifacts can easily be mistaken for a physiological response during analysis, to date there is no automatic method for detecting artifacts. This paper describes the development of a machine learning algorithm for automatically detecting EDA artifacts, and provides an empirical evaluation of classification performance. We have encoded our results into a freely available web-based tool for artifact and peak detection.

What can wearable sensors and usage of smart phones tell us about academic performance, self-reported sleep quality, stress and mental health condition? To answer this question, we collected extensive subjective and objective data using mobile phones, surveys, and wearable sensors worn day and night from 66 participants, for 30 days each, totaling 1,980 days of data. We analyzed daily and monthly behavioral and physiological patterns and identified factors that affect academic performance (GPA), Pittsburg Sleep Quality Index (PSQI) score, perceived stress scale (PSS), and mental health composite score (MCS) from SF-12, using these month-long data. We also examined how accurately the collected data classified the participants into groups of high/low GPA, good/poor sleep quality, high/low self-reported stress, high/low MCS using feature selection and machine learning techniques. We found associations among PSQI, PSS, MCS, and GPA and personality types. Classification accuracies using the objective data from wearable sensors and mobile phones ranged from 67-92%.

We present the first quantitative characterization of electrodermal activity (EDA) patterns on the wrists of healthy adults during sleep using dry electrodes. We compare the new results on the wrist to the prior findings on palmar or finger EDA by characterizing data measured from 80 nights of sleep consisting of 9 nights of wrist and palm EDA from 9 healthy adults sleeping at home, 56 nights of wrist and palm EDA from one healthy adult sleeping at home, and 15 nights of wrist EDA from 15 healthy adults in a sleep laboratory, with the latter compared to concurrent polysomnography. While high frequency patterns of EDA called "storms" were identified by eye in the 1960s, we systematically compare thresholds for automatically detecting EDA peaks and establish criteria for EDA storms. We found that more than 80% of the EDA peaks occurred in non-REM sleep, specifically during slow-wave sleep (SWS) and non-REM stage 2 sleep (NREM2). Also, EDA amplitude is higher in SWS than in other sleep stages. Longer EDA storms were more likely to occur in the first two quarters of sleep and during SWS and NREM2. We also found from the home studies (65 nights) that EDA levels were higher and the skin conductance peaks were larger and more frequent when measured on the wrist than when measured on the palm. These EDA high frequency peaks and high amplitude were sometimes associated with higher skin temperature, but more work is needed looking at neurological and other EDA elicitors in order to elucidate their complete behavior.

In order to model students' happiness, we apply machine learning methods to data collected from undergrad students monitored over the course of one month each. The data collected include physiological signals, location, smartphone logs, and survey responses to behavioral questions. Each day, participants reported their wellbeing on measures including stress, health, and happiness. Because of the relationship between happiness and depression, modeling happiness may help us to detect individuals who are at risk of depression and guide interventions to help them. We are also interested in how behavioral factors (such as sleep and social activity) affect happiness positively and negatively. A variety of machine learning and feature selection techniques are compared, including Gaussian Mixture Models and ensemble classification. We achieve 70% classification accuracy of self-reported happiness on held-out test data.