Search for: All records

BackgroundIn the modern economy, shift work is prevalent in numerous occupations. However, it often disrupts workers’ circadian rhythms and can result in shift work sleep disorder. Proper management of shift work sleep disorder involves comprehensive and patient-specific strategies, some of which are similar to cognitive behavioral therapy for insomnia. ObjectiveOur goal was to develop and evaluate machine learning algorithms that predict physicians’ sleep advice using wearable and survey data. We developed a web- and app-based system to provide individualized sleep and behavior advice based on cognitive behavioral therapy for insomnia for shift workers. MethodsData were collected for 5 weeks from shift workers (N=61) in the intensive care unit at 2 hospitals in Japan. The data comprised 3 modalities: Fitbit data, survey data, and sleep advice. After the first week of enrollment, physicians reviewed Fitbit and survey data to provide sleep advice and selected 1 to 5 messages from a list of 23 options. We handcrafted physiological and behavioral features from the raw data and identified clusters of participants with similar characteristics using hierarchical clustering. We explored 3 models (random forest, light gradient-boosting machine, and CatBoost) and 3 data-balancing approaches (no balancing, random oversampling, and synthetic minority oversampling technique) to predict selections for the 7 most frequent advice messages related to bedroom brightness, smartphone use, and nap and sleep duration. We tested our predictions under participant-dependent and participant-independent settings and analyzed the most important features for prediction using permutation importance and Shapley additive explanations. ResultsWe found that the clusters were distinguished by work shifts and behavioral patterns. For example, one cluster had days with low sleep duration and the lowest sleep quality when there was a day shift on the day before and a midnight shift on the current day. Our advice prediction models achieved a higher area under the precision-recall curve than the baseline in all settings. The performance differences were statistically significant (P<.001 for 13 tests and P=.003 for 1 test). Sensitivity ranged from 0.50 to 1.00, and specificity varied between 0.44 and 0.93 across all advice messages and dataset split settings. Feature importance analysis of our models found several important features that matched the corresponding advice messages sent. For instance, for message 7 (darken the bedroom when you go to bed), the models primarily examined the average brightness of the sleep environment to make predictions. ConclusionsAlthough our current system requires physician input, an accurate machine learning algorithm shows promise for automatic advice without compromising the trustworthiness of the selected recommendations. Despite its decent performance, the algorithm is currently limited to the 7 most popular messages. Further studies are needed to enable predictions for less frequent advice labels. 

BackgroundAs mobile health (mHealth) studies become increasingly productive owing to the advancements in wearable and mobile sensor technology, our ability to monitor and model human behavior will be constrained by participant receptivity. Many health constructs are dependent on subjective responses, and without such responses, researchers are left with little to no ground truth to accompany our ever-growing biobehavioral data. This issue can significantly impact the quality of a study, particularly for populations known to exhibit lower compliance rates. To address this challenge, researchers have proposed innovative approaches that use machine learning (ML) and sensor data to modify the timing and delivery of surveys. However, an overarching concern is the potential introduction of biases or unintended influences on participants’ responses when implementing new survey delivery methods. ObjectiveThis study aims to demonstrate the potential impact of an ML-based ecological momentary assessment (EMA) delivery system (using receptivity as the predictor variable) on the participants’ reported emotional state. We examine the factors that affect participants’ receptivity to EMAs in a 10-day wearable and EMA–based emotional state–sensing mHealth study. We study the physiological relationships indicative of receptivity and affect while also analyzing the interaction between the 2 constructs. MethodsWe collected data from 45 healthy participants wearing 2 devices measuring electrodermal activity, accelerometer, electrocardiography, and skin temperature while answering 10 EMAs daily, containing questions about perceived mood. Owing to the nature of our constructs, we can only obtain ground truth measures for both affect and receptivity during responses. Therefore, we used unsupervised and supervised ML methods to infer affect when a participant did not respond. Our unsupervised method used k-means clustering to determine the relationship between physiology and receptivity and then inferred the emotional state during nonresponses. For the supervised learning method, we primarily used random forest and neural networks to predict the affect of unlabeled data points as well as receptivity. ResultsOur findings showed that using a receptivity model to trigger EMAs decreased the reported negative affect by >3 points or 0.29 SDs in our self-reported affect measure, scored between 13 and 91. The findings also showed a bimodal distribution of our predicted affect during nonresponses. This indicates that this system initiates EMAs more commonly during states of higher positive emotions. ConclusionsOur results showed a clear relationship between affect and receptivity. This relationship can affect the efficacy of an mHealth study, particularly those that use an ML algorithm to trigger EMAs. Therefore, we propose that future work should focus on a smart trigger that promotes EMA receptivity without influencing affect during sampled time points. 

Abstract Objectives: This study aimed to evaluate, using wearable sensors, the impact of transitioning from an 8-hour to a 12-hour shift schedule on sleep patterns and well-being in intensive care unit (ICU) nurses with pre-existing sleep disturbances. We also examined differences in outcome based on chronotype. Methods: We conducted an observational study at a university hospital ICU between November 2020 and October 2023, before and after a hospital-wide shift schedule change. Nurses wore wearable sensors and completed daily surveys over 5 weeks under each shift system. Rotating-shift ICU nurses with a Pittsburgh Sleep Quality Index score >5 were eligible. Sleep metrics and subjective well-being were compared using linear mixed models, adjusting for age. Sleep episodes were categorized relative to shift timing, and chronotype-stratified subgroup analyses were performed. Results: Eighty nurses completed the study (12-hour shift: 37; 8-hour shift: 43). The interval between shifts was greater for the 12-hour shift group (36.12 vs 26.78 hours). Total sleep duration did not significantly differ between groups (12-hour shift: 418.5 minutes; 8-hour shift: 398 minutes); however, the 12-hour shift group had less fragmented sleep, higher subjective well-being scores, and lower reported stress and fatigue. Evening chronotypes appeared to benefit more from 12-hour shifts, with longer sleep duration and higher well-being scores, though these differences were not statistically significant. Conclusions: Transitioning to a 12-hour shift schedule was associated with reduced sleep fragmentation and improved well-being, particularly among evening chronotypes. These findings suggest that shift schedule structure and individual chronotype may influence adaptation to shift work in ICU settings. 

Background/Objectives: Nurses are at high risk for burnout. Identification of biomarkers associated with early manifestations of distress is essential to support effective intervention efforts. Methods: Fifty nurses from a large hospital system participated in a 30-day study of biopsychosocial factors that may contribute to burnout. Nurses wore an Oura ring that collected behavioral data and they completed a self-report burnout questionnaire at baseline and the end of the study period. Machine learning models were developed to evaluate whether objective measures could predict burnout states and changes at the end of the study period. Analyses were exploratory and hypothesis-generating for future work. Results: Data for 45 participants were included in the analyses. Participants with burnout had significantly higher sleep variability. Sleep measures provided 75.75% accuracy in ability to discriminate between burnout states. Heart rate-based measures better modeled changes in symptomatic components of burnout (Emotional Exhaustion, Depersonalization) over time. Heart rate-based measures provided a R-squared value of 0.13 (p < 0.05) (RMSE of 7.41) in a regression model of changes in Emotional Exhaustion evaluated in a leave-one-participant-out cross-validation. Conclusions: Sleep measures’ association with a state of burnout may reflect the longer-term manifestations of chronic exposure to workplace stress. Short-term changes in burnout symptoms are associated with disturbances in heart rate measures. Wearable technology may support monitoring/early identification of those at risk for burnout. 

Sleep behavior significantly impacts health and acts as an indicator of physical and mental well-being. Monitoring and predicting sleep behavior with ubiquitous sensors may therefore assist in both sleep management and tracking of related health conditions. While sleep behavior depends on, and is reflected in the physiology of a person, it is also impacted by external factors such as digital media usage, social network contagion, and the surrounding weather. In this work, we propose SleepNet, a system that exploits social contagion in sleep behavior through graph networks and integrates it with physiological and phone data extracted from ubiquitous mobile and wearable devices for predicting next-day sleep labels about sleep duration. Our architecture overcomes the limitations of large-scale graphs containing connections irrelevant to sleep behavior by devising an attention mechanism. The extensive experimental evaluation highlights the improvement provided by incorporating social networks in the model. Additionally, we conduct robustness analysis to demonstrate the system's performance in real-life conditions. The outcomes affirm the stability of SleepNet against perturbations in input data. Further analyses emphasize the significance of network topology in prediction performance revealing that users with higher eigenvalue centrality are more vulnerable to data perturbations. 

Abstract Emotion prediction plays an essential role in mental healthcare and emotion-aware computing. The complex nature of emotion resulting from its dependency on a person’s physiological health, mental state, and his surroundings makes its prediction a challenging task. In this work, we utilize mobile sensing data to predict self-reported happiness and stress levels. In addition to a person’s physiology, we also incorporate the environment’s impact through weather and social network. To this end, we leverage phone data to construct social networks and develop a machine learning architecture that aggregates information from multiple users of the graph network and integrates it with the temporal dynamics of data to predict emotion for all users. The construction of social networks does not incur additional costs in terms of ecological momentary assessments or data collection from users and does not raise privacy concerns. We propose an architecture that automates the integration of the user’s social network in affect prediction and is capable of dealing with the dynamic distribution of real-life social networks, making it scalable to large-scale networks. The extensive evaluation highlights the prediction performance improvement provided by the integration of social networks. We further investigate the impact of graph topology on the model’s performance. 

Shift work disrupts sleep and causes chronic stress, resulting in burnout syndrome characterized by emotional exhaustion, depersonalization, and decreased personal accomplishment. Continuous biometric data collected through wearable devices contributes to mental health research. However, direct prediction of burnout risk is still limited, and interpreting machine learning (ML) models in healthcare poses challenges. In this paper, we develop machine learning models that utilize wearable and survey data, including rhythm features, to predict burnout risk among shift workers. Additionally, we employ the DiCE (Diverse Counterfactual Explanations) framework to generate interpretable explanations for the ML model, aiding in the management of burnout risks. Our experiments on the AMED dataset show that incorporating rhythm features significantly enhances the predictive performance of our models. Specifically, sleep and heart rate features have emerged as significant indicators for accurately predicting burnout risk 

Physiological and behavioral data collected from wearable or mobile sensors have been used to estimate self-reported stress levels. Since stress annotation usually relies on self-reports during the study, a limited amount of labeled data can be an obstacle to developing accurate and generalized stress-predicting models. On the other hand, the sensors can continuously capture signals without annotations. This work investigates leveraging unlabeled wearable sensor data for stress detection in the wild. We propose a two-stage semi-supervised learning framework that leverages wearable sensor data to help with stress detection. The proposed structure consists of an auto-encoder pre-training method for learning information from unlabeled data and the consistency regularization approach to enhance the robustness of the model. Besides, we propose a novel active sampling method for selecting unlabeled samples to avoid introducing redundant information to the model. We validate these methods using two datasets with physiological signals and stress labels collected in the wild, as well as four human activity recognition (HAR) datasets to evaluate the generality of the proposed method. Our approach demonstrated competitive results for stress detection, improving stress classification performance by approximately 7% to 10% on the stress detection datasets compared to the baseline supervised learning models. Furthermore, the ablation study we conducted for the HAR tasks supported the effectiveness of our methods. Our approach showed comparable performance to state-of-the-art semi-supervised learning methods for both stress detection and HAR tasks. 

Background Shift workers are at high risk of developing sleep disorders such as shift worker sleep disorder or chronic insomnia. Cognitive behavioral therapy (CBT) is the first-line treatment for insomnia, and emerging evidence shows that internet-based CBT is highly effective with additional features such as continuous tracking and personalization. However, there are limited studies on internet-based CBT for shift workers with sleep disorders. Objective This study aimed to evaluate the impact of a 4-week, physician-assisted, internet-delivered CBT program incorporating machine learning–based well-being prediction on the sleep duration of shift workers at high risk of sleep disorders. We evaluated these outcomes using an internet-delivered CBT app and fitness trackers in the intensive care unit. Methods A convenience sample of 61 shift workers (mean age 32.9, SD 8.3 years) from the intensive care unit or emergency department participated in the study. Eligible participants were on a 3-shift schedule and had a Pittsburgh Sleep Quality Index score ≥5. The study comprised a 1-week baseline period, followed by a 4-week intervention period. Before the study, the participants completed questionnaires regarding the subjective evaluation of sleep, burnout syndrome, and mental health. Participants were asked to wear a commercial fitness tracker to track their daily activities, heart rate, and sleep for 5 weeks. The internet-delivered CBT program included well-being prediction, activity and sleep chart, and sleep advice. A job-based multitask and multilabel convolutional neural network–based model was used for well-being prediction. Participant-specific sleep advice was provided by sleep physicians based on daily surveys and fitness tracker data. The primary end point of this study was sleep duration. For continuous measurements (sleep duration, steps, etc), the mean baseline and week-4 intervention data were compared. The 2-tailed paired t test or Wilcoxon signed rank test was performed depending on the distribution of the data. Results In the fourth week of intervention, the mean daily sleep duration for 7 days (6.06, SD 1.30 hours) showed a statistically significant increase compared with the baseline (5.54, SD 1.36 hours; P=.02). Subjective sleep quality, as measured by the Pittsburgh Sleep Quality Index, also showed statistically significant improvement from baseline (9.10) to after the intervention (7.84; P=.001). However, no significant improvement was found in the subjective well-being scores (all P>.05). Feature importance analysis for all 45 variables in the prediction model showed that sleep duration had the highest importance. Conclusions The physician-assisted internet-delivered CBT program targeting shift workers with a high risk of sleep disorders showed a statistically significant increase in sleep duration as measured by wearable sensors along with subjective sleep quality. This study shows that sleep improvement programs using an app and wearable sensors are feasible and may play an important role in preventing shift work–related sleep disorders. International Registered Report Identifier (IRRID) RR2-10.2196/24799.