Summary: Data from fitness trackers can detect mood episodes in individuals with bipolar disorder with up to 89.1% accuracy for mania and 80.1% for depression. Researchers used passively collected, noninvasive data and machine learning algorithms to identify mood changes, demonstrating the potential for real-time monitoring.
These findings could enhance clinical care by alerting healthcare providers to mood episodes between appointments, enabling faster intervention. The approach moves toward personalized algorithms that could broadly support patients without requiring specialized devices or invasive data sharing.
Key Facts:
- Fitness tracker data detected mania with 89.1% accuracy and depression with 80.1%.
- The study used passively collected, noninvasive data for real-world clinical application.
- Algorithms could alert clinicians to mood episodes, improving bipolar disorder treatment.
Source: Brigham and Women’s Hospital
Investigators from Brigham and Women’s Hospital, a founding member of the Mass General Brigham healthcare system, evaluated whether data collected from a fitness tracker could be used to accurately detect mood episodes in people with bipolar disorder.
Their findings, published in Acta Psychiatrica Scandinavica, indicate that it is possible to detect time intervals when patients with bipolar disorder are experiencing depression or mania with high accuracy using data from fitness tracking devices.
“Most people are walking around with personal digital devices like smartphones and smartwatches that capture day-to-day data that could inform psychiatric treatment.
“Our goal was to use that data to identify when study participants diagnosed with bipolar disorder were experiencing mood episodes,” said corresponding author Jessica Lipschitz, PhD, an investigator in the Brigham’s Department of Psychiatry. “In the future, our hope is that machine learning algorithms like ours could help patients’ treatment teams respond fast to new or unremitting episodes in order to limit negative impact.”
Bipolar disorder (BD) is a chronic psychiatric disorder characterized by extreme mood swings, including depression, mania, and hypomania followed by periods of remission. Identification and treatment of new and unremitting mood episodes is essential for limiting the impact of BD on patients’ lives.
While previous research has indicated that personal digital devices can accurately detect mood episodes, previous studies have not used methods designed for broad application in clinical settings.
As an implementation scientist, Lipschitz, together with colleagues, focused on using methods that could be broadly implemented in clinical practice. Specifically, they used commercially available personal digital devices, limited data filtering, and entirely passively collected and noninvasive data.
Applying a new type of machine learning algorithm, they were able to detect clinically significant symptoms of depression with 80.1% accuracy and clinically significant symptoms of mania with 89.1% accuracy.
The researchers note that, “overall, results move the field a step toward personalized algorithms suitable for the full population of patients, rather than only those with high compliance, access to specialized devices, or willingness to share invasive data.”
Their next step is to apply these predictive algorithms in routine care where they could be used to improve BD treatment by informing clinicians when their patients are experiencing depressive or manic episodes between scheduled appointments. The researchers have also been working on extending this work to major depressive disorder.
Authorship: In addition to Lipschitz, Mass General Brigham authors include Chelsea K Pike and Katherine E. Burdick. Additional authors include Sidian Lin and Soroush Saghafian.
Disclosures: Burdick serves as the chair of the steering committee and as the Scientific Director for the Integrated Network of the non-profit foundation, Breakthrough Discoveries for thriving with Bipolar Disorder (BD^2) and receives grant funding and honoraria in this capacity and also received honorarium as a scientific advisory board member for Merck in the past 12 months, but declares no financial competing interests.
Lipschitz is a consultant to Solara Health Inc., but declares no financial competing interests. All other authors declare no financial or non-financial competing interests.
Funding: This research was supported by a Young Investigator Grant from the Brain & Behavior Research Foundation (#28537; to JML), a grant from the Harvard Brain Science Initiative Bipolar Disorder Seed Grant Program, and a Pathways Research Award from Alkermes, Inc. The data collection for the longitudinal study was supported in part by the Baszucki Brain Research Fund (to KEB) and the Harvard Brain Science Initiative Bipolar Disorder Seed Grant Program (to KEB).
Additionally, Lipschitz’s time was partially supported by the National Institute of Mental Health (NIMH) Grant MH120324. Saghafian’s time was partially supported through a grant from Harvard’s Middle East Initiative Kuwait Science Program, which is aimed at improving population health via machine learning and AI-enabled mobile health interventions.
The funding organizations played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.
About this bipolar disorder and neurotech research news
Author: Cassandra Falone
Source: Brigham and Women’s Hospital
Contact: Cassandra Falone – Brigham and Women’s Hospital
Image: The image is credited to Neuroscience News
Original Research: Closed access.
“Digital phenotyping in bipolar disorder: Using longitudinal Fitbit data and personalized machine learning to predict mood symptomatology” by Jessica Lipschitz et al. Acta Psychiatrica Scandinavica
Abstract
Digital phenotyping in bipolar disorder: Using longitudinal Fitbit data and personalized machine learning to predict mood symptomatology
Background
Effective treatment of bipolar disorder (BD) requires prompt response to mood episodes. Preliminary studies suggest that predictions based on passive sensor data from personal digital devices can accurately detect mood episodes (e.g., between routine care appointments), but studies to date do not use methods designed for broad application.
This study evaluated whether a novel, personalized machine learning approach, trained entirely on passive Fitbit data, with limited data filtering could accurately detect mood symptomatology in BD patients.
Methods
We analyzed data from 54 adults with BD, who wore Fitbits and completed bi-weekly self-report measures for 9 months. We applied machine learning (ML) models to Fitbit data aggregated over two-week observation windows to detect occurrences of depressive and (hypo)manic symptomatology, which were defined as two-week windows with scores above established clinical cutoffs for the Patient Health Questionnaire-8 (PHQ-8) and Altman Self-Rating Mania Scale (ASRM) respectively.
Results
As hypothesized, among several ML algorithms, Binary Mixed Model (BiMM) forest achieved the highest area under the receiver operating curve (ROC-AUC) in the validation process. In the testing set, the ROC-AUC was 86.0% for depression and 85.2% for (hypo)mania.
Using optimized thresholds calculated with Youden’s J statistic, predictive accuracy was 80.1% for depression (sensitivity of 71.2% and specificity of 85.6%) and 89.1% for (hypo)mania (sensitivity of 80.0% and specificity of 90.1%).
Conclusion
We achieved sound performance in detecting mood symptomatology in BD patients using methods designed for broad application. Findings expand upon evidence that Fitbit data can produce accurate mood symptomatology predictions. Additionally, to the best of our knowledge, this represents the first application of BiMM forest for mood symptomatology prediction.
Overall, results move the field a step toward personalized algorithms suitable for the full population of patients, rather than only those with high compliance, access to specialized devices, or willingness to share invasive data.
