Summary: A first-of-its-kind clinical pilot found that a personalized, machine-learning–guided lifestyle coaching program substantially increased remission rates for people with mild-to-moderate depression. By using smartwatch data and frequent self-reports to identify each person’s top behavioral drivers of low mood, the intervention nearly doubled typical remission outcomes and produced lasting gains in anxiety, cognition, and quality of life.
Using consumer smartwatches and short daily mood logs, researchers created individualized Mood Augmentation Plans (iMAPs) informed by N-of-1 machine learning. Delivered remotely with brief weekly coaching, the program produced a 55% remission rate and significant reductions in anxiety, suggesting a scalable path toward personalized, data-driven mental health care.
Key Facts
- The problem with generic advice: More than 21% of U.S. adults experience depression. Standard clinical recommendations—like improving sleep, exercise, and diet—are useful in principle but often fail in practice because they are broad and overwhelming. Depression varies widely between individuals, so a one-size-fits-all approach is frequently ineffective.
- Two-week biometric baseline: Fifty adults with mild-to-moderate depression completed a two-week monitoring period. Participants wore smartwatches to capture heart rate and movement and completed ecological momentary assessments (EMAs) up to four times per day to record sleep quality, diet, activity, social contact, and current mood.
- The iMAP approach: Researchers at UC San Diego’s Neural Engineering and Translation Labs (NEATLabs) used personalized machine learning models to identify the single or small set of lifestyle factors that best predicted each person’s low moods. Health coaches then translated those insights into focused, evidence-based behavioral interventions tailored to each participant.
- Improved clinical outcomes: Traditional behavioral programs typically yield remission rates around 30%. The machine-learning–guided iMAP approach produced a 55% remission rate after six weeks, meaning over half of participants no longer met criteria for depression on the PHQ-9 screening tool.
- Broader health benefits: Participants also experienced a 36% reduction in anxiety symptoms, meaningful improvements in self-reported quality of life, and measurable gains on brief tests of memory, attention, and executive function.
- Durable effects: Follow-up assessments showed that improvements in mood, anxiety, and cognition persisted for at least three months after the active six-week coaching period ended.
Source: UCSD
More than 21% of U.S. adults live with depression, which substantially lowers quality of life. Many people with mild-to-moderate depression can improve their symptoms by changing daily habits such as sleep, activity, diet, and social engagement. But because depression manifests differently for each person, generalized lifestyle advice can feel overwhelming and is often ineffective.
In a novel pilot trial, Jyoti Mishra, PhD, and colleagues developed a machine-learning–guided lifestyle coaching program that uses personal device data and frequent mood reports to tailor behavioral treatments to each individual. Participants who followed their personalized plans reported large reductions in depressive symptoms after six weeks, suggesting a promising remote model for individualized depression care. The trial was published in NPP – Digital Psychiatry and Neuroscience.
During the initial two-week monitoring phase, participants wore a smartwatch that captured heart rate and physical activity while answering short mood and lifestyle surveys four times daily. The research team built a machine learning model for each person to identify which lifestyle variables best forecasted their low moods. Those insights informed an individualized Mood Augmentation Plan (iMAP) that participants implemented with weekly video coaching for six weeks.
Coaching and interventions were selected from established behavioral therapies and matched to the participant’s top predictive factor. For example, participants whose low mood correlated most strongly with poor sleep received targeted cognitive behavioral therapy for insomnia; those whose mood tracked with social isolation received interventions to increase meaningful social contact; others focused on structured physical activity or dietary adjustments aimed at mood stabilization.
After six weeks of guided implementation, outcomes included:
- A substantial reduction in depressive symptoms: 55% of participants no longer met depression criteria on the PHQ-9.
- A 36% decrease in anxiety symptoms measured by the GAD-7.
- Significant gains in reported quality of life.
- Improvements on brief cognitive tests assessing memory, selective attention, and interference processing.
The trial also demonstrated that these benefits were sustained during a three-month follow-up period after the coaching ended. Mishra notes that targeting individual, data-driven triggers is likely less overwhelming and more feasible for people in a depressed state than attempting broad lifestyle overhauls.
Although this pilot study involved a modest sample size and lacked a randomized control arm, it offers compelling early evidence that combining digital monitoring, individualized machine learning insights, and brief remote coaching can deliver clinically meaningful improvements for people with mild-to-moderate depression. The authors recommend larger, controlled trials to confirm and extend these findings.
Co-authors include Jason Nan, Suzanna Purpura, Satish Jaiswal, Houtan Afshar, Vojislav Maric, James K. Manchanda, Charles T. Taylor, and Dhakshin Ramanathan. The study received partial funding from a seed grant by the Hope for Depression Research Foundation.
Key Questions Answered:
A: When someone is depressed they frequently operate in survival mode, focused on day-to-day functioning. A long list of broad lifestyle changes can be overwhelming. Because depression’s causes and expressions differ widely across individuals, targeting the single most predictive, data-backed factor for each person is more manageable and effective than trying to change everything at once.
A: Over two weeks, the model analyzes patterns between biometric signals (like activity and heart rate) and frequent mood reports to identify which lifestyle variables best predict low mood for that individual. If low mood is most strongly associated with poor sleep, coaching focuses on sleep interventions; if social contact is the key predictor, the plan emphasizes rebuilding meaningful social connections.
A: In this trial, the reductions in depression and anxiety and the cognitive gains persisted through a three-month follow-up period after the six weeks of video coaching concluded, indicating durable therapeutic effects.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full by the editorial team.
- Additional context was added by staff to clarify methods and clinical implications.
About this AI and depression research news
Author: Susanne Bard
Source: UCSD
Contact: Susanne Bard – UCSD
Image: The image is credited to Neuroscience News
Original Research: Open access. “Personalized machine learning guided intervention for optimizing lifestyle behaviors in depression: a pilot study” by Jason Nan, Suzanna Purpura, Satish Jaiswal, Houtan Afshar, Vojislav Maric, James K. Manchanda, Charles T. Taylor, Dhakshin Ramanathan & Jyoti Mishra. NPP—Digital Psychiatry and Neuroscience. DOI: 10.1038/s44277-026-00062-3
Abstract
Personalized machine learning guided intervention for optimizing lifestyle behaviors in depression: a pilot study
Personalized, data-driven interventions for depression are urgently needed. This pilot leveraged N-of-1 machine learning to optimize lifestyle behavior targets for individuals with depression. Fifty adults with mild-to-moderate depression enrolled in the single-arm, open-label Personalized Mood Augmentation (PerMA) trial (ClinicalTrials.gov NCT05662254).
Participants completed a two-week digital monitoring phase using smartphone ecological momentary assessments (EMAs) four times daily, combined with smartwatch tracking of sleep, exercise, diet, social connection, and mood. Personalized ML models identified the lifestyle factors most predictive of each participant’s mood, and those results were translated into individualized Mood Augmentation Plans (iMAPs). Participants implemented iMAPs for six weeks with weekly health coach support.
Among intervention completers (n = 40), depression symptoms decreased significantly on self-rated PHQ-9 and clinician-rated HDRS scales, with effects maintained through a 12-week follow-up. Anxiety scores fell and quality of life improved. Objective cognitive measures including selective attention, interference processing, and working memory also showed significant improvement. EMA analysis confirmed that mood improvement was specifically predicted by improvements in the individually targeted lifestyle factors. Decision algorithms and a large language model could replicate human coach iMAP assignments with high accuracy, supporting future automation and scale-up. The PerMA trial supports further testing in larger, controlled studies to establish efficacy.
PERMA was registered with ClinicalTrials.gov under registry number NCT05662254.