Alcohol use in young adults is common, with high rates of morbidity and mortality largely due to periodic, heavy drinking episodes (HDEs). Behavioral interventions delivered through electronic communication modalities (e.g., text messaging) can reduce the frequency of HDEs in young adults, but effects are small. One way to amplify these effects is to deliver support materials proximal to drinking occasions, but this requires knowledge of when they will occur. Mobile phones have built-in sensors that can potentially be useful in monitoring behavioral patterns associated with the initiation of drinking occasions. The objective of our work is to explore the detection of daily-life behavioral markers using mobile phone sensors and their utility in identifying drinking occasions. We utilized data from 30 young adults aged 21-28 with past hazardous drinking and collected mobile phone sensor data and daily Experience Sampling Method (ESM) of drinking for 28 consecutive days. We built a machine learning-based model that is 96.6% accurate at identifying non-drinking, drinking and heavy drinking episodes. We highlight the most important features for detecting drinking episodes and identify the amount of historical data needed for accurate detection. Our results suggest that mobile phone sensors can be used for automated, continuous monitoring of at-risk populations to detect drinking episodes and support the delivery of timely interventions.

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

Depression, a prevalent and complex mental health issue affecting millions worldwide, presents significant challenges for detection and monitoring. While facial expressions have shown promise in laboratory settings for identifying depression, their potential in real-world applications remains largely unexplored due to the difficulties in developing efficient mobile systems. In this study, we aim to introduce FacePsy, an open-source mobile sensing system designed to capture affective inferences by analyzing sophisticated features and generating real-time data on facial behavior landmarks, eye movements, and head gestures - all within the naturalistic context of smartphone usage with 25 participants. Through rigorous development, testing, and optimization, we identified eye-open states, head gestures, smile expressions, and specific Action Units (2, 6, 7, 12, 15, and 17) as significant indicators of depressive episodes (AUROC=81%). Our regression model predicting PHQ-9 scores achieved moderate accuracy, with a Mean Absolute Error of 3.08. Our findings offer valuable insights and implications for enhancing deployable and usable mobile affective sensing systems, ultimately improving mental health monitoring, prediction, and just-in-time adaptive interventions for researchers and developers in healthcare.

Background: Digital just-in-time adaptive interventions can reduce binge-drinking events (BDEs; consuming ≥4 drinks for women and ≥5 drinks for men per occasion) in young adults but need to be optimized for timing and content. Delivering just-in-time support messages in the hours prior to BDEs could improve intervention impact. Objective: We aimed to determine the feasibility of developing a machine learning (ML) model to accurately predict future, that is, same-day BDEs 1 to 6 hours prior BDEs, using smartphone sensor data and to identify the most informative phone sensor features associated with BDEs on weekends and weekdays to determine the key features that explain prediction model performance. Methods: We collected phone sensor data from 75 young adults (aged 21 to 25 years; mean 22.4, SD 1.9 years) with risky drinking behavior who reported their drinking behavior over 14 weeks. The participants in this secondary analysis were enrolled in a clinical trial. We developed ML models testing different algorithms (eg, extreme gradient boosting [XGBoost] and decision tree) to predict same-day BDEs (vs low-risk drinking events and non-drinking periods) using smartphone sensor data (eg, accelerometer and GPS). We tested various “prediction distance” time windows (more proximal: 1 hour; distant: 6 hours) from drinking onset. We also tested various analysis time windows (ie, the amount of data to be analyzed), ranging from 1 to 12 hours prior to drinking onset, because this determines the amount of data that needs to be stored on the phone to compute the model. Explainable artificial intelligence was used to explore interactions among the most informative phone sensor features contributing to the prediction of BDEs. Results: The XGBoost model performed the best in predicting imminent same-day BDEs, with 95% accuracy on weekends and 94.3% accuracy on weekdays (F1-score=0.95 and 0.94, respectively). This XGBoost model needed 12 and 9 hours of phone sensor data at 3- and 6-hour prediction distance from the onset of drinking on weekends and weekdays, respectively, prior to predicting same-day BDEs. The most informative phone sensor features for BDE prediction were time (eg, time of day) and GPS-derived features, such as the radius of gyration (an indicator of travel). Interactions among key features (eg, time of day and GPS-derived features) contributed to the prediction of same-day BDEs. Conclusions: We demonstrated the feasibility and potential use of smartphone sensor data and ML for accurately predicting imminent (same-day) BDEs in young adults. The prediction model provides “windows of opportunity,” and with the adoption of explainable artificial intelligence, we identified “key contributing features” to trigger just-in-time adaptive intervention prior to the onset of BDEs, which has the potential to reduce the likelihood of BDEs in young adults. Trial Registration: ClinicalTrials.gov NCT02918565; https://clinicaltrials.gov/ct2/show/NCT02918565

Hospital readmissions are a major problem facing health care systems today, costing Medicare alone US$26 billion each year. Being readmitted is associated with significantly shorter survival, and is often preventable. Predictors of readmission are still not well understood, particularly those under the patient's control: behavioral risk factors. Our work evaluates the ability of behavioral risk factors, specifically Fitbit-assessed behavior, to predict readmission for 25 postsurgical cancer inpatients. Our results show that sum of steps, maximum sedentary bouts, frequency, and low breaks in sedentary times during waking hours are strong predictors of readmission. We built two models for predicting readmissions: Steps-only and Behavioral model that adds information about sedentary behaviors. The Behavioral model (88.3%) outperforms the Steps-only model (67.1%), illustrating the value of passively collected information about sedentary behaviors. Indeed, passive monitoring of behavior data, i.e., mobility, after major surgery creates an opportunity for early risk assessment and timely interventions.