Improving Care: Using Smartphones and Wearables to Identify Individuals with Depression

October 23, 2024
Ruta Nonacs, MD PhD
Artificial intelligence may be able to use data passively collected from wearable devices and smartphones to improve mental health care by providing real-time updates to clinicians.

October is National Depression and Mental Health Screening Month.  According to recent data, major depressive disorder affects about 8.3% of adults, with an even higher prevalence in young adults (aged 18-25).  While depression is a leading cause of disability, treatment rates vary widely, with many affected individuals not receiving adequate care.  Given the current demand for mental health services, many are looking for strategies to increase our capacity to screen for depressive disorders and to monitor ongoing treatment.

The following article, previously published on our website, suggests that smartphones and wearable devices may help to increase our ability to provide care and monitor patients with depression.

Smartphones and Wearables Able to Identify and Monitor Depressive Symptoms

The vast majority of Americans – 85% according to a report from the Pew Research Center –  own a smartphone of some kind.  In addition, about 45% of Americans regularly wear a smartwatch or fitness tracker – most commonly Gen Zers (70%) and millennials (57%).  These devices collect a tremendous amount of potentially useful information, and medical researchers have begun to explore how to use this kind of data to make more accurate diagnoses and to monitor patients from a distance.  

In a collaborative project with Rosalind Picard, ScD from MIT’s Abdul Latif Jameel Clinic for Machine Learning in Health, Paola Pedrelli, PhD and colleagues from the Depression Clinical and Research Program have been working to develop machine learning algorithms that could help diagnose and monitor symptoms in patients with major depressive disorder.

In a pilot study, Pedrelli and Picard used wearable devices and smartphones to gather data from 31 study participants with major depressive disorder (MDD).  Participants were instructed to wear two E4 Empatica wristbands (one on each wrist) for 22 hours each a day, gathering information on skin conductance and temperature, heart rate, physical activity, and sleep patterns.  Mobile-based social interactions (e.g., number of calls, texts), activity patterns , and number of apps used were tracked through a smartphone app.

Readings gathered from the wristbands and smartphone were compared to clinicians’ ratings of depressive symptom severity assessed using the 28-item Hamilton Depression Rating Scale (HDRS-28).

Machine-learning models were developed to correlate behavioral and psychological data with clinician-rated depressive symptom severity.  Correlations between the best models’ estimate of depression severity and clinician-rated HDRS scores were high (0.7, CI: 0.66, 0.74) and had moderate accuracy.  The most predictive features were related to mobile phone engagement, activity level, skin conductance, and heart rate variability.

What’s Next?

The current study supports the feasibility of measuring changes in depressive symptoms severity using information collected passively by smartphone and wrist sensors. Overall, adherence was high but was reduced by technological issues.  The researchers estimated that with improvements in the reliability of network access, connectivity, and hardware would lead to more than 90% adherence.

This study highlights the potential of digital phenotyping as a tool for continuous mental health monitoring. By leveraging mobile and wearable technology, clinicians could receive timely updates on patients’ depressive symptoms without relying solely on patient self-reporting or infrequent clinical visits.  For example, in patients starting treatment, this passively gathered, real-time information may facilitate the early detection of response or non-response to treatment, allowing clinicians to adjust medications more quickly, ideally decreasing time to remission.  In addition, this monitoring may facilitate early detection of depression relapse and allow for the expeditious delivery of effective treatments to patients.

The findings indicate that behavioral, as well as physiological features, contributed to the model’s accuracy.  Future studies will focus on refining these models by identifying the most critical features for accurate symptom assessment and will explore larger sample sizes to enhance model performance.  While the widespread use of smartphones makes their use in monitoring symptoms highly scalable, passive collection of physiological parameters may be less feasible given the high cost of wearables.  Further studies are critical to identify which features, or aggregate of features, are essential to developing models that are cost-effective, feasible, and scalable.

Read More

Curtiss JE, Mischoulon D, Fisher LB, Cusin C, Fedor S, Picard RW, Pedrelli P. Rising early warning signals in affect associated with future changes in depression: a dynamical systems approach. Psychol Med. 2021 Dec 23:1-9. doi: 10.1017/S0033291721005183. 

Pedrelli P, Fedor S, Ghandeharioun A, Howe E, Ionescu DF, Bhathena D, Fisher LB, Cusin C, Nyer M, Yeung A, Sangermano L, Mischoulon D, Alpert JE, Picard RW. Monitoring Changes in Depression Severity Using Wearable and Mobile Sensors. Front Psychiatry. 2020 Dec 18; 11:584711. doi: 10.3389/fpsyt.2020.584711. 

In the News

Paola Pedrelli, PhD is the Director Of Dual Diagnoses Studies at the Depression Clinical and Research Program.  She is an Assistant in Psychology at Massachusetts General Hospital and an Assistant Professor of Psychiatry at Harvard Medical School.  Her research focuses on investigating the etiology, assessment, and treatment of comorbid Affective Disorders and Alcohol Use Disorders (AUDs).

Share this with your network: