Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 97 of the 97 references it cites were found in a registry or an index.

Personalized machine learning of depressed mood using wearables

Open access

Abstract

Depression is a multifaceted illness with large interindividual variability in clinical response to treatment. In the era of digital medicine and precision therapeutics, new personalized treatment approaches are warranted for depression. Here, we use a combination of longitudinal ecological momentary assessments of depression, neurocognitive sampling synchronized with electroencephalography, and lifestyle data from wearables to generate individualized predictions of depressed mood over a 1-month time period. This study, thus, develops a systematic pipeline for N-of-1 personalized modeling of depression using multiple modalities of data. In the models, we integrate seven types of supervised machine learning (ML) approaches for each individual, including ensemble learning and regression-based methods. All models were verified using fourfold nested cross-validation. The best-fit as benchmarked by the lowest mean absolute percentage error, was obtained by a different type of ML model for each individual, demonstrating that there is no one-size-fits-all strategy. The voting regressor, which is a composite strategy across ML models, was best performing on-average across subjects. However, the individually selected best-fit models still showed significantly less error than the voting regressor performance across subjects. For each individual’s best-fit personalized model, we further extracted top-feature predictors using Shapley statistics. Shapley values revealed distinct feature determinants of depression over time for each person ranging from co-morbid anxiety, to physical exercise, diet, momentary stress and breathing performance, sleep times, and neurocognition. In future, these personalized features can serve as targets for a personalized ML-guided, multimodal treatment strategy for depression.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41398-021-01445-0

Cites

97 of 97 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Summary of the three main steps involved in the personalized depression modeling pipeline, namely, data ingestion and feature extraction, data preprocessing, and machine learning-based analysis. These steps were carried out separately for each subject and personalized performance reports, prediction reports, and feature importance reports were obtained.
Fig. 1 Summary of the three main steps involved in the personalized depression modeling pipeline, namely, data ingestion and feature extraction, data preprocessing, and machine learning-based analysis. These steps were carried out separately for each subject and personalized performance reports, prediction reports, and feature importance reports were obtained.CC BY · PubMed Central
Fig. 2 Comparisons of actual depression states as self-reported by participants vs. their predicted depression states obtained from the personalized ML pipeline with fourfold CV. Actual and predicted value comparisons are shown over time with each EMA serving as one time-step, and also compared as per their histogram distributions. The bottom row plots show the heatmap and histogram comparisons for actual vs. predicted values across all subjects.
Fig. 2 Comparisons of actual depression states as self-reported by participants vs. their predicted depression states obtained from the personalized ML pipeline with fourfold CV. Actual and predicted value comparisons are shown over time with each EMA serving as one time-step, and also compared as per their histogram distributions. The bottom row plots show the heatmap and histogram comparisons for actual vs. predicted values across all subjects.CC BY · PubMed Central
Fig. 3 Spearman rank correlation coefficients with 95% confidence interval bounds are plotted for the relationship between the predicted and observed depressed state values over time in each individual. The overall correlation obtained by concatenating the actual vs. predicted values across all subjects, is also shown as the last data point. All correlations were significant except in P-18 and P-24. Actual correlation values are shown in Supplementary Table 2.
Fig. 3 Spearman rank correlation coefficients with 95% confidence interval bounds are plotted for the relationship between the predicted and observed depressed state values over time in each individual. The overall correlation obtained by concatenating the actual vs. predicted values across all subjects, is also shown as the last data point. All correlations were significant except in P-18 and P-24. Actual correlation values are shown in Supplementary Table 2.CC BY · PubMed Central
Fig. 4 SHapley additive exPlanations (SHAP) summary plots for each subject showing rank feature importance and the feature effects. The feature importance is depicted by the size of the gray bars that represent mean absolute Shapley values for the top-five features; bar colors simply represent different feature identity. The feature effects are depicted by each colored point on the summary plot which is a Shapley value for a feature and an instance. The position on the y-axis is determined by the feature and on the x-axis by the Shapley value. The color represents the value of the feature from low (blue) to high (pink). Overlapping points are jittered on the y-axis direction, so we get a sense of the distribution of the Shapley values per feature. The features are ordered according to their importance. In most cases, EMA ratings of co-morbid anxiety (“anxious”) best predicted the depressed state. These plots reveal how each individual had different modalities of data as their top-rank predictors, which can then be leveraged for personalized intervention in future studies. Top variables observed were cumm_step_distance/speed/calories/count that depicted the cumulative step features in the past 12 h; Mean Breathing Time and Consistency that were obtained from the 30-s active stress assessment at each EMA, prev_night_sleep or hours of previous night’s sleep; past day sugars/fats/caffeine; exercise_duration/calories over the past 24 h; heart rate within the 30 min window of the EMA; ppg_std that depicted the HRV in the 15 min window of the EMA and time of day. In some cases, neurocognitive metrics also emerged as top-ranking features, including LD_GL_bias and LD_RareG_diff that respectively represented the bias towards frequent gain vs. loss in the reward task and the preference for rare gain choices when they have greater vs. equal expected value in the reward task; GLbias_dACC/left DLPFC that was the neural activity in the dACC/left DLPFC brain region corresponding to bias for frequent gains vs. losses on the reward task; diff_rareLG_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to choices made on the reward task with a contrast of expected values; gw_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Go Wait inhibitory control task; mf_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Middle fish interference processing task; and ls_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Lost Star working memory task.
Fig. 4 SHapley additive exPlanations (SHAP) summary plots for each subject showing rank feature importance and the feature effects. The feature importance is depicted by the size of the gray bars that represent mean absolute Shapley values for the top-five features; bar colors simply represent different feature identity. The feature effects are depicted by each colored point on the summary plot which is a Shapley value for a feature and an instance. The position on the y-axis is determined by the feature and on the x-axis by the Shapley value. The color represents the value of the feature from low (blue) to high (pink). Overlapping points are jittered on the y-axis direction, so we get a sense of the distribution of the Shapley values per feature. The features are ordered according to their importance. In most cases, EMA ratings of co-morbid anxiety (“anxious”) best predicted the depressed state. These plots reveal how each individual had different modalities of data as their top-rank predictors, which can then be leveraged for personalized intervention in future studies. Top variables observed were cumm_step_distance/speed/calories/count that depicted the cumulative step features in the past 12 h; Mean Breathing Time and Consistency that were obtained from the 30-s active stress assessment at each EMA, prev_night_sleep or hours of previous night’s sleep; past day sugars/fats/caffeine; exercise_duration/calories over the past 24 h; heart rate within the 30 min window of the EMA; ppg_std that depicted the HRV in the 15 min window of the EMA and time of day. In some cases, neurocognitive metrics also emerged as top-ranking features, including LD_GL_bias and LD_RareG_diff that respectively represented the bias towards frequent gain vs. loss in the reward task and the preference for rare gain choices when they have greater vs. equal expected value in the reward task; GLbias_dACC/left DLPFC that was the neural activity in the dACC/left DLPFC brain region corresponding to bias for frequent gains vs. losses on the reward task; diff_rareLG_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to choices made on the reward task with a contrast of expected values; gw_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Go Wait inhibitory control task; mf_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Middle fish interference processing task; and ls_leftDLPFC that was the neural activity in the left DLPFC brain region evoked to the Lost Star working memory task.CC BY · PubMed Central
Fig. 5 Personalized ML informed top-five ranking features across individuals. Frequency of top-five ranking Shapley feature domains across participants cumulated based on the personalized ML models in individual subjects are shown.
Fig. 5 Personalized ML informed top-five ranking features across individuals. Frequency of top-five ranking Shapley feature domains across participants cumulated based on the personalized ML models in individual subjects are shown.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

2 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 94 references

Research works printed but not linked: 2

  1. Institute of Medicine (US) Forum on Drug Discovery, Development and T. Clinical Trials in Depression. in Transforming Clinical Research in the United States: Challenges and Opportunities: Workshop Summary. (National Academies Press (US), 2010). https://www.ncbi.nlm.nih.gov/books/NBK50890/.
  2. Heeger, D. & Landy, M. Signal detection theory. in Encyclopedia of perception (ed Goldstein, B.) 887–892 (SAGE Publications, 2009).

Cited by

See all 145

Metadata

Authors
Shah, R. V., Grennan, G., Zafar-Khan, M., Alim, F., Dey, S., Ramanathan, D., & Mishra, J.
Affiliations
University of California San Diego (US)
Corresponding author
Mishra, Jyoti
Venue
Translational Psychiatry
Published
2021-06-09
Volume
11
Issue
1
Pages
338-338
ISSN
2158-3188 (ISSN-L)
OpenAlex
W3170475262
Cited by
123
References
94
Citations by year
2026: 29 · 2025: 28 · 2024: 28 · 2023: 23 · 2022: 12 · 2021: 3
FWCI
15.93
Type
journal article
Language
en
Open access
Licence
Locations
  • Translational Psychiatry (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
Funders
University of California, San Diego
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Mental Health Research Topics
  • Social Sciences › Psychology › Applied Psychology › Digital Mental Health Interventions
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
Keywords
depression, personalized machine learning, wearable devices, ecological momentary assessment, EEG, neurocognitive function, multimodal data integration, voting regressor, Shapley value, feature importance, physical activity, sleep, anxiety comorbidity, precision therapy
MeSH terms
Machine Learning, Ecological Momentary Assessment, Wearable Electronic Devices, Anxiety Disorders, Humans, Exercise

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Personalized machine learning of depressed mood using wearables. https://app.cochat.ai/papers/10.1038/s41398-021-01445-0
The paper
Shah, R. V., Grennan, G., Zafar-Khan, M., Alim, F., Dey, S., Ramanathan, D., & Mishra, J. (2021). Personalized machine learning of depressed mood using wearables. Translational Psychiatry. https://doi.org/10.1038/s41398-021-01445-0

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.