CoChat
Sign Up
Cited by 158Digital Mental Health InterventionsMental Health Research TopicsEmotion and Mood Recognition

Monitoring Changes in Depression Severity Using Wearable and Mobile Sensors

Open access
Pedrelli, Paola, Fedor, Szymon, Ghandeharioun, Asma, Howe, Esther, Ionescu, Dawn F., Bhathena, Darian, Fisher, Lauren B., Cusin, Cristina +6 more·2020·Frontiers in Psychiatry
DOI 10.3389/fpsyt.2020.584711PMC PMC7775362
Open this paper in CoChat — free accountOpen PDF

Abstract

Background: While preliminary evidence suggests that sensors may be employed to detect presence of low mood it is still unclear whether they can be leveraged for measuring depression symptom severity. This study evaluates the feasibility and performance of assessing depressive symptom severity by using behavioral and physiological features obtained from wristband and smartphone sensors. Method: Participants were thirty-one individuals with Major Depressive Disorder (MDD). The protocol included 8 weeks of behavioral and physiological monitoring through smartphone and wristband sensors and six in-person clinical interviews during which depression was assessed with the 17-item Hamilton Depression Rating Scale (HDRS-17). Results: Participants wore the right and left wrist sensors 92 and 94% of the time respectively. Three machine-learning models estimating depressive symptom severity were developed–one combining features from smartphone and wearable sensors, one including only features from the smartphones, and one including features from wrist sensors–and evaluated in two different scenarios. Correlations between the models' estimate of HDRS scores and clinician-rated HDRS ranged from moderate to high (0.46 [CI: 0.42, 0.74] to 0.7 [CI: 0.66, 0.74]) and had moderate accuracy with Mean Absolute Error ranging between 3.88 ± 0.18 and 4.74 ± 1.24. The time-split scenario of the model including only features from the smartphones performed the best. The ten most predictive features in the model combining physiological and mobile features were related to mobile phone engagement, activity level, skin conductance, and heart rate variability. Conclusion: Monitoring of MDD patients through smartphones and wrist sensors following a clinician-rated HDRS assessment is feasible and may provide an estimate of changes in depressive symptom severity. Future studies should further examine the best features to estimate depressive symptoms and strategies to further enhance accuracy.

AI snapshot

Written from the paper’s full text.

TL;DR

This study evaluated whether passively collected wristband and smartphone sensor data could estimate changes in depression severity among 31 adults with Major Depressive Disorder over 8 weeks. Machine-learning models achieved moderate-to-high correlations with clinician-rated HDRS-17 scores (0.46–0.7) and MAE between 3.88 and 4.74, with a smartphone-only model performing best in the time-split scenario.

Key findings

  • After excluding days with technical problems, participants wore the left and right wrist sensors 94% and 92% of the time, respectively; raw adherence was 77% and 70% when charging/uploading time was included.
  • Across models, correlations between estimated and clinician-rated HDRS ranged from 0.46 (95% CI 0.42, 0.74) to 0.7 (95% CI 0.66, 0.74), and MAE ranged from 3.88 ± 0.18 to 4.74 ± 1.24.
  • In the time-split scenario, the smartphone-only model performed best (MAE = 3.88 ± 0.18; r = 0.7, 95% CI 0.66, 0.74), significantly outperforming the other models [F(2,12) = 19.04, p < 0.002]; in the user-split scenario, models performed similarly [F(2,12) = 0.55, p < 0.59], with the wearable-only model having the lowest MAE.
  • Boruta retained 39 informative features, 54% from mobile phones and 46% from wearables, primarily related to mobile phone engagement, activity level, skin conductance, and heart rate variability; sleep was not found to be a critical feature.

Methods

  • Forty-one participants with MDD were enrolled and 31 completed; participants were aged 19–73 (M = 33.7, SD = 14), primarily female (n = 23, 74%), and had severe depressive symptoms at screening (HDRS-28 = 22.74, SD = 7.38). The protocol included six in-person visits, daily smartphone surveys, and passive monitoring with two E4 Empatica wristbands and the MovisensXS smartphone app.
  • The dependent variable was the HDRS-17 derived from clinician-administered HDRS-28, residualized on the screening HDRS score. The final dataset included 877 features (404 wearable, 473 mobile), calculated for four 6-hour intervals and daily aggregates.
  • Models used an average ensemble of boosting and random forest with 10-fold cross-validation, evaluated in user-split and time-split hold-out scenarios using MAE, RMSE, and correlations; Boruta was used to rank important features.

Limitations

  • The paper states the sample size was small and participants had low variability of depressive symptoms during the study (average within-user HDRS SD = 3.7; baseline-to-last HDRS decreased by 15% on average), so performance among patients with higher symptom variability is unclear.
  • The paper notes that Boruta evaluates feature importance in the context of all included features, so models with different feature combinations may yield different results; larger studies and other machine-learning models such as Deep Neural Networks are suggested.
  • Evident from the text: technical problems caused missing data and reduced adherence, and the sample was predominantly female (74%) and White (71%), which may limit generalizability; the models also did not outperform the individual-median baseline in the time-split scenario.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3389/fpsyt.2020.584711

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

0 of 65 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    At screening, participants on average had severe depressive symptoms, with a 28-item Hamilton Depression Rating Scale score of 22.74 (SD = 7.38).

    “At screening, participants on average had severe depressive symptoms [28-item Hamilton Depression Rating Scale (30) = 22.74; SD = 7.38].”
    Quote in the paperNot a substantial claim
  • Findingown data

    Features related to asymmetry of right and left skin conductance response may be important in estimating depressive symptoms severity.

    “but that also features related to asymmetry of right and left skin conductance response may be important.”
    Quote in the paperSubstantial claim
  • Findingown data

    Engagement with the phone is an important predictor of depression severity in our model.

    “our finding suggesting that engagement with the phone is an important predictor of depression severity”
    Quote in the paperSubstantial claim
  • Findingnone given

    The model with mobile features in the time-split scenario performed significantly better than the others.

    “the model with mobile features in the time-split scenario performed significantly better than the others”
    Quote in the paperNot a substantial claim
  • Findingown data

    Our models' RMSE values ranged between 4.88 and 5.99.

    “Our models' RMSE values ranged between 4.88 and 5.99”
    Quote in the paperNot a substantial claim
  • May combine two claims. 20
  • Classified as probably not a substantial claim. 35
  • The quote only partly states this claim. 1

  • May combine two claims.

    Findingown data

    54% of the 39 informative features retained by the Boruta analyses were from the mobile phone and 46% were from the wearables.

    “Notably, 54% of the 39 informative features that were retained by the Boruta analyses were from the mobile phone and 46% of all the informative features were from the wearables.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Using the Boruta algorithm, 39 features were defined as important for the estimation of the HDRS scores, one feature remained undecided, and the remaining features were identified as irrelevant to the outcome variable.

    “Using the Boruta algorithm (43) 39 features were defined as important for the estimation of the HDRS scores, one feature remained undecided, and the remaining features were identified as irrelevant to the outcome variable.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [43] Kursa, 2010, Feature selection with the Boruta package, J Stat Softw source

      Not checked in the free run.

  • May combine two claims.

    Findingreasoning

    The normalized MAE ranged between 7.5 and 9.1%, as the HDRS-17 ranges from 0 to 52.

    “Thus, the normalized MAE ranged between 7.5 and 9.1%, as the HDRS-17 ranges from 0 to 52.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Residualized HDRS scores account for the heterogeneous presentation of depressive symptoms.

  • May combine two claims.

    Findingown data

    The correlation between observed and estimated depression in the time-split model including features from the mobile phone was r = 0.7, the strongest among the models evaluated.

    “The correlation between observed and estimated depression in the time-split model including features from the mobile phone (r = 0.7) was the strongest”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    On average, participants uploaded about 17 and 15.5 h of daily data from left and right-hand sensors respectively, corresponding to 77 and 70% adherence.

    “On average, participants uploaded about 17 and 15.5 h of daily data from left and right-hand sensors respectively. This corresponds to 77 and 70% adherence considering that participants were asked to use 2-h each day to charge the E4 and to upload data, which led to a maximum of 22 h of data per day.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The machine learning models provided more accurate estimates than those based on group median and individual screen models but not better than those based on individual median in the time split scenario; these differences were not significant.

    “The machine learning models provided more accurate estimates than those based on group median and individual screen models but not better than those based on individual median in the time split scenario. However, these differences were also not significant.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Residualized HDRS scores account for the heterogeneous presentation of depressive symptoms.

  • May combine two claims.

    Findingown data

    In the user-split scenario, the three models performed about the same, with the lowest MAE obtained by the model using only features from the wearable.

    “When the three models were tested with the user-split scenario, they all performed about the same [F (2,12) = 0.55, p < 0.59] with the lowest MAE obtained by the model using only the features from the wearable.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Residualized HDRS scores account for the heterogeneous presentation of depressive symptoms.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    On average, the HDRS score from baseline to the last assessment decreased by 15%.

    “on average, the HDRS score from baseline to the last assessment decreased by 15%”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The features retained by the Boruta analyses were in the following categories: mobile phone engagement, activity level recorded by the mobile sensors, and skin conductance and HRV features from the wearables.

    “The features that were retained were in the following categories: mobile phone engagement, activity level recorded by the mobile sensors, and skin conductance and HRV features from the wearables (Table 2).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [43] Kursa, 2010, Feature selection with the Boruta package, J Stat Softw source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Inter-rater reliability between DCRP clinicians in diagnosing MDD and measuring severity of depression has yielded kappas greater than 0.75, indicating satisfactory agreement.

    “Recent assessment of inter-rater reliability between DCRP clinicians in diagnosing MDD and measuring severity of depression has yielded kappas > 0.75, indicating satisfactory agreement (based on internal data).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Anhedonia and low energy level can manifest as reduced physical activity, which can be measured using GPS and motion sensors embedded in smartphones or wearable sensors.

    “Anhedonia and low energy level can manifest as reduced physical activity (19, 20), which can be measured using GPS and motion sensors embedded in smartphones or wearable sensors.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [19] Nakonezny, 2015, Evaluation of anhedonia with the snaith-hamilton pleasure scale (SHAPS) in adult outpatients with major depressive disorder, J Psychiatr Res source

      Not checked in the free run.

    • cites [20] Stubbs, 2018, Relationship between sedentary behavior and depression: a mediation analysis of influential factors across the lifespan among 42,469 people in low- and middle-income countries, J Affect Disord source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    MovisensXS has been successfully used to securely and confidentially collect behavioral and self-reported mood data.

    “MovisensXS has been successfully used to securely and confidentially collect behavioral and self-reported mood data (36).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [36] Fritz, 2017, Using a smartphone-based ecological momentary assessment protocol with community dwelling older African Americans, J Gerontol B Psychol Sci Soc Sci source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Prior investigations have primarily focused on detecting presence or absence of depression rather than on assessing its severity, and they have shown overall low accuracy.

    “This has resulted in possible reliability problems, since these investigations have primarily focused on detecting presence or absence of depression rather than on assessing its severity, and they have shown overall low accuracy (14–18, 23–27).”
    Quote in the paperSubstantial claim0 of 10 sources checked
    • cites [14] Beiwinkel, 2016, Using smartphones to monitor bipolar disorder symptoms: a pilot study, JMIR Ment Health source

      Not checked in the free run.

    • cites [15] Faurholt-Jepsen, 2016, Behavioral activities collected through smartphones and the association with illness activity in bipolar disorder, Int J Methods Psychiatr Res source

      Not checked in the free run.

    • cites [16] Faurholt-Jepsen, 2014, Smartphone data as objective measures of bipolar disorder symptoms, Psychiatr Res Neuroimag source

      Not checked in the free run.

    • cites [17] Grünerbl, 2014, Smartphone-based recognition of states and state changes in bipolar disorder patients, IEEE J Biomed Heal Inform source

      Not checked in the free run.

    • cites [18] Saeb, 2015, Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study, J Med Internet Res source

      Not checked in the free run.

    • cites [23] Pratap, 2019, The accuracy of passive phone sensors in predicting daily mood, Depress Anxiety source

      Not checked in the free run.

    • cites [24] Place, 2017, Behavioral Indicators on a mobile sensing platform predict clinically validated psychiatric symptoms of mood and anxiety disorders, J Med Internet Res source

      Not checked in the free run.

    • cites [25] Sano, 2018, Identifying objective physiological markers and modifiable behaviors for self-reported stress and mental health status using wearable sensors and mobile phones: observational study, J Med Internet Res. source

      Not checked in the free run.

    • cites [26] Jacobson, 2020, Passive sensing of prediction of moment-to-moment depressed mood among undergraduates with clinical levels of depression sample using smartphones, Sensors. source

      Not checked in the free run.

    • cites [27] Lu, 2018, Joint modeling of heterogeneous sensing data for depression assessment via multi-task learning, Proc ACM Interact Mob Wearable Ubiquitous Technol source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Severe depressive symptoms are associated with shorter duration of incoming and outgoing calls, and fewer incoming and outgoing phone calls and daily text messages.

    “recent work shows that severe depressive symptoms are associated with shorter duration of incoming and outgoing calls, and fewer incoming and outgoing phone calls and daily text messages (13–18).”
    Quote in the paperSubstantial claim0 of 6 sources checked
    • cites [13] Courtin, 2017, Social isolation, loneliness and health in old age: a scoping review, Health Soc Care Community source

      Not checked in the free run.

    • cites [14] Beiwinkel, 2016, Using smartphones to monitor bipolar disorder symptoms: a pilot study, JMIR Ment Health source

      Not checked in the free run.

    • cites [15] Faurholt-Jepsen, 2016, Behavioral activities collected through smartphones and the association with illness activity in bipolar disorder, Int J Methods Psychiatr Res source

      Not checked in the free run.

    • cites [16] Faurholt-Jepsen, 2014, Smartphone data as objective measures of bipolar disorder symptoms, Psychiatr Res Neuroimag source

      Not checked in the free run.

    • cites [17] Grünerbl, 2014, Smartphone-based recognition of states and state changes in bipolar disorder patients, IEEE J Biomed Heal Inform source

      Not checked in the free run.

    • cites [18] Saeb, 2015, Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study, J Med Internet Res source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductcitation

    The potential problem of overfitting was addressed by reducing the dimensionality of features using linear and non-linear transformations.

    “Similar to our previous work (37), we addressed the potential problem of overfitting by reducing the dimensionality of features using linear and non-linear transformations (see more details in A.2. in Supplementary Material).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [37] Ghandeharioun, 2017, Objective assessment of depressive symptoms with machine learning and wearable sensors data source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductcitation

    Current MDD and exclusionary diagnoses were determined using the Mini International Neuropsychiatric Interview (MINI), a structured diagnostic interview for major psychiatric disorders.

    “Mini International Neuropsychiatric Interview (MINI): The MINI, a structured diagnostic interview for major psychiatric disorders, was administered during the screening visit to determine current MDD and rule out exclusionary diagnoses (32).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [32] Sheehan, 1998, The mini-international neuropsychiatric interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10, J Clin Psychiatry

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The HDRS-17 is one of the methods most commonly employed to measure change in depressive symptom severity in treatment for depression clinical trials.

    “The HDRS-17 (33), one of the methods most commonly employed to measures change in depressive symptoms severity in treatment for depression clinical trials, was derived from the HDRS-28, and used as dependent variable.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [33] Hamilton, 1960, A rating scale for depression, J Neurol Neurosurg Psychiatry source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    Seven participants dropped out after the screening visit, two after visit three, and one after visit two.

    “Seven participants dropped out after the screening visit, two after visit three, and one after visit two.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The HDRS-28 is a 28-item clinician-rated assessment scale to assess depressive symptoms.

    “Hamilton Depression Rating Scale (HDRS): The HDRS-28 is a 28-item clinician-rated assessment scale to assess depressive symptoms (30).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] Thase, 1992, Treatment of imipramine-resistant recurrent depression, III: efficacy of monoamine oxidase inhibitors, J Clin Psychiatry

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    More severe depressive symptoms and worse mood levels are negatively correlated with a higher amount of time the phone screen is on, a higher number of entertainment applications used, and an overall time of interaction with the smartphone.

    “It has also been documented that more severe depressive symptoms and worse mood levels are negatively correlated with a higher amount of time the phone screen is on, a higher number of entertainment applications (apps) used, and an overall time of interaction with the smartphone (18, 21).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [18] Saeb, 2015, Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study, J Med Internet Res source

      Not checked in the free run.

    • cites [21] Alvarez-Lozano, 2014, Tell me your apps and I will tell you your mood: correlation of apps usage with bipolar disorder state

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Forty-one participants with MDD were enrolled and 31 completed the study.

    “Forty-one participants with MDD were enrolled and 31 completed the study.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The protocol was approved by the Institutional Review Boards of Massachusetts General Hospital and Massachusetts Institute of Technology.

    “The protocol was approved by the Institutional Review Boards of Massachusetts General Hospital and Massachusetts Institute of Technology.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The time-split mobile phone model's correlation (r = 0.7) was higher than a previous model combining features from the fitbit and from smartphones (best r2 = 0.44 or r = 0.66) and a model aggregating mobile-based and physiological features (r = 0.58).

    “was higher than the one of a previous model combining features from the fitbit and from smartphones (the best model yielded an r2 = 0.44 or r = 0.66) (27) and the one of a model aggregating mobile-based and physiological features (r = 0.58) (26).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [27] Lu, 2018, Joint modeling of heterogeneous sensing data for depression assessment via multi-task learning, Proc ACM Interact Mob Wearable Ubiquitous Technol source

      Not checked in the free run.

    • cites [26] Jacobson, 2020, Passive sensing of prediction of moment-to-moment depressed mood among undergraduates with clinical levels of depression sample using smartphones, Sensors. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    The Boruta method evaluates the importance of the features in the context of all the features in the model.

    “It is important to highlight that the Boruta method evaluates the importance of the features in the context of all the features in the model.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    O'Brien et al. (45) documented that physical activity was low among individuals with late life depression related to healthy controls.

    “O'Brien et al. (45) documented that physical activity was low among individuals with late life depression related to healthy controls”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [45] O'Brien, 2017, A study of wrist-worn activity measurement as a potential real-world biomarker for late-life depression, Psychol Med source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    Monitoring of MDD patients through smartphones and wrist sensors following a clinician-rated HDRS assessment is feasible and may provide an estimate of changes in depressive symptom severity.

    “Monitoring of MDD patients through smartphones and wrist sensors following a clinician-rated HDRS assessment is feasible and may provide an estimate of changes in depressive symptom severity.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Residualized HDRS scores account for the heterogeneous presentation of depressive symptoms.

  • May combine two claims.

    Interpretationreasoning

    Indices of associations do not allow a granular evaluation of the accuracy of the models and of the magnitude of the difference between estimated and actual values, impacting scalability.

    “However, indices of associations do not allow a granular evaluation of the accuracy of the models and of the magnitude of the difference between estimated and actual values, impacting scalability.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Evidence is accumulating that these models may have moderate accuracy.

    “while evidence is accumulating that these models may have moderate accuracy”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Evidence is accumulating suggesting that physiological sensing may also be included in the effort to objectively characterize changes in depression severity.

    “Evidence is accumulating suggesting that physiological sensing may also be included in the effort to objectively characterize changes in depression severity.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Right and left asymmetry may be a better indication of arousal than one-sided EDA measurements.

    “Previously we have posited that right and left asymmetry may be a better indication of arousal than one-sided EDA measurements (46).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [46] Picard, 2015, Multiple arousal theory and daily-life electrodermal activity asymmetry, Emotion Rev source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Digital phenotyping has included the collection of behavioral data such as mobility patterns (via accelerometer) and socialization (via call and texts).

    “To date, digital phenotyping has included the collection of behavioral data such as mobility patterns (via accelerometer) and socialization (via call and texts) (12).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [12] Torous, 2016, New tools for new research in psychiatry: a scalable and customizable platform to empower data driven smartphone research, JMIR Mental Health source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    Digital phenotyping of depression can be leveraged as a clinical tool and may allow a more seamless continuous treatment.

    “Digital phenotyping of depression can be leveraged as a clinical tool and may allow a more seamless continuous treatment.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Saeb et al. (18) first showed that mobility pattern, recorded by phone sensors, is associated with depression.

    “Saeb et al. (18) first showed that mobility pattern, recorded by phone sensors, is associated with depression”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] Saeb, 2015, Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study, J Med Internet Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Cao et al. (44) documented that activity level was associated with low mood among adolescents.

    “Cao et al. (44) documented that activity level was associated with low mood among adolescents.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [44] Cao, 2020, Tracking and predicting depressive symptoms of adolescents using smartphone-based self-reports, parental evaluations, and passive phone sensor data: development and usability study, JMIR Ment Health source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationcitation

    The finding of the role of HRV in predicting depressive symptoms is consistent with and extends previous reports.

    “is consistent and extends previous reports.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductcitation

    To avoid overfitting, multiple dimensionality reduction and feature transformation techniques were applied to the raw features and 10-fold cross-validation was performed during training.

    “To avoid overfitting, multiple dimensionality reduction and feature transformation techniques were applied to the raw features and we performed 10-fold-cross-validation (41) during training (see more details in A.2. in Supplementary Material);”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [41] Géron, 2017

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Cao et al. (44) reported that in a model including HR features recorded by the fitbit, light exposure, and sleep, one of the most important features to estimate mood variation among 18 individuals with MDD was HR.

    “Moreover, Cao et al. (44) reported that in a model including HR features recorded by the fitbit, light exposure, and sleep one of the most important features to estimate mood variation among 18 individuals with MDD was HR.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [44] Cao, 2020, Tracking and predicting depressive symptoms of adolescents using smartphone-based self-reports, parental evaluations, and passive phone sensor data: development and usability study, JMIR Ment Health source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The finding that activity level was among the most important features is consistent with previous reports.

    “is consistent with previous reports (23, 24).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [23] Pratap, 2019, The accuracy of passive phone sensors in predicting daily mood, Depress Anxiety source

      Not checked in the free run.

    • cites [24] Place, 2017, Behavioral Indicators on a mobile sensing platform predict clinically validated psychiatric symptoms of mood and anxiety disorders, J Med Internet Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable). The quote only partly states this claim.

    Interpretationown data

    The MAE range of 3.8 to 4.74 may be too high of an inaccuracy for the model to be scalable.

    “which may be too high of an inaccuracy for the model to be scalable.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    A model including a combination of features different from ours may yield different results.

    “Thus, a model including a combination of features different from ours may yield different results.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The patients/participants provided their written informed consent to participate in this study.

    “The patients/participants provided their written informed consent to participate in this study.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Passive collection of physiological indices may be less feasible due to the high cost of wearables.

    “passive collection of physiological indices may be less feasible due to the high cost of wearables.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Wearables-based recording of skin conductance could be leveraged to detect high stress and to distinguish between high and low mental health groups of individuals.

    “Previously, it had been shown that wearables-based recording of skin conductance could be leveraged to detect high stress and to distinguish between high and low mental health groups of individuals (25).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [25] Sano, 2018, Identifying objective physiological markers and modifiable behaviors for self-reported stress and mental health status using wearable sensors and mobile phones: observational study, J Med Internet Res. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The finding that engagement with the phone is an important predictor of depression severity is consistent with previous studies.

    “is consistent with previous studies (18).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] Saeb, 2015, Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: an exploratory study, J Med Internet Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The difference in results may be due to the fact that the studies employed different measures of depression.

    “The difference in results may be due to the fact that the studies employed different measures of depression.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The studies involving human participants were reviewed and approved by MGH and MIT IRBs.

    “The studies involving human participants were reviewed and approved by MGH and MITIRBs.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    After the parameters of the model were learnt, the model was tested on a separate hold-out test set of data.

    “After the parameters of the model were learnt, the model was tested on a separate hold-out test set of data.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The finding that sleep was not a critical feature may have been due to the fact that sleep features in our model were derived from up to 48 h prior to the HDRS assessments and features capturing a longer time period may be needed to characterize depressive symptoms.

    “This finding may have been due to the fact that sleep features in our model were derived from up to 48 h prior to the HDRS assessments and features capturing a longer time period may be needed to characterize depressive symptoms.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous models estimating mood states with moderate accuracy have included measures of heart rate (HR) and HRV.

    “Previous models estimating mood states with moderate accuracy have included measures of heart rate (HR) (27, 44) and HRV (26).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [27] Lu, 2018, Joint modeling of heterogeneous sensing data for depression assessment via multi-task learning, Proc ACM Interact Mob Wearable Ubiquitous Technol source

      Not checked in the free run.

    • cites [44] Cao, 2020, Tracking and predicting depressive symptoms of adolescents using smartphone-based self-reports, parental evaluations, and passive phone sensor data: development and usability study, JMIR Ment Health source

      Not checked in the free run.

    • cites [26] Jacobson, 2020, Passive sensing of prediction of moment-to-moment depressed mood among undergraduates with clinical levels of depression sample using smartphones, Sensors. source

      Not checked in the free run.

49 of 49

  • The paper relies on this without support.

    Passive, real-time symptom monitoring may facilitate early detection of response or non-response to treatment, or of depression relapse, and allow expedited delivery of relief to patients.

    “Methods enabling passive, real-time symptom monitoring may facilitate early detection of response or non-response to treatment, or of depression relapse and allow expedited delivery of relief to patients.”

  • The paper relies on this without support.

    Residualized HDRS scores account for the heterogeneous presentation of depressive symptoms.

    “Residualized HDRS scores allowed to account for the heterogeneous presentation of depressive symptoms.”

  • The paper relies on this without support.

    Most studies evaluating passive monitoring have examined depressive symptoms among patients with bipolar disorders.

    “Most studies evaluating passive monitoring have examined depressive symptoms among patients with bipolar disorders.”

  • The paper relies on this without support.

    Smartphones are ubiquitous.

    “while the ubiquity of smartphones makes their use in monitoring symptoms highly scalable”

  • The paper relies on this without support.

    Technological problems are the main cause of non-adherence, so fixing them would raise adherence above 90%.

    “a finding that also suggested that fixing the reliability of network access, connectivity, and sensor, laptop, and phone hardware would lead to more than 90% adherence.”
5 of 5

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?

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • ConsistentF(2, 12) = 19.04reported p < .002 · recomputed p < .001

    “F (2,12) = 19.04, p < .002”

  • ConsistentF(2, 12) = 0.55reported p < .59 · recomputed p = .591

    “F (2,12) = 0.55, p < .59”

2 of 2

Self-check

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

Each contradiction between two passages listed here was confirmed by an independent check.

  • Time-split training includes a screening score excluded by the stated outcome definition

    Minor

    The outcome definition says training and test scores are residuals from visits 2-6, but the time-split description also places the first HDRS score-the screening score-in the training set.

  • Hispanic/Latino percentage does not match the participant count

    Minor

    Four of the 31 participants is approximately 12.9%, not 23%.

    Reported 23 · recomputed 12.903226

  • Abstract gives a different confidence interval for the 0.46 correlation

    Minor

    The abstract reports a lower confidence bound of 0.42 for the 0.46 correlation; Table 1 reports 0.18.

3 of 3

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. 1994, Diagnostic and Statistical Manual of Mental Disorders

Cited by

Sorted by citations

Related

Authors
Pedrelli, P., Fedor, S., Ghandeharioun, A., Howe, E., Ionescu, D. F., Bhathena, D., Fisher, L. B., Cusin, C., Nyer, M., Yeung, A., Sangermano, L., Mischoulon, D., Alpert, J. E., & Picard, R. W.
Affiliations
  • Massachusetts General Hospital (US)
  • MIT Media Lab (US)
  • University of California, Berkeley (US)
  • Janssen Research & Development (United States) (US)
  • Albert Einstein College of Medicine (US)
Corresponding author
Pedrelli, Paola
ORCID
Venue
Frontiers in Psychiatry
Publisher
Frontiers Media SA
Published
2020-12-18
Volume
11
Pages
584711-584711
ISSN
1664-0640 (ISSN-L)
DOI
10.3389/fpsyt.2020.584711
PMC
PMC7775362
OpenAlex
W3111758161
Semantic Scholar
976b22a260e13b73c9677342af3b13a8be04ae31
Cited by
158
References
47
Citations by year
2026: 20 · 2025: 35 · 2024: 44 · 2023: 26 · 2022: 25 · 2021: 8
FWCI
13.34
Type
journal article
Language
en
Open access
Licence
Locations
  • Frontiers in Psychiatry (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • European Commission (327702)
  • National Institute of Mental Health (R01MH118274)
Topics
  • Social Sciences › Psychology › Applied Psychology › Digital Mental Health Interventions
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Mental Health Research Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Emotion and Mood Recognition
Keywords
major depressive disorder, wearable sensors, wrist-worn sensors, smartphone sensors, physiological traits, skin conductance, heart rate variability, activity level, Hamilton Depression Rating Scale, machine learning prediction, remote monitoring

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

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