Longitudinal Assessment of Seasonal Impacts and Depression Associations on Circadian Rhythm Using Multimodal Wearable Sensing: Retrospective Analysis (Preprint) DOI
Yuezhou Zhang, Amos Folarin, Shaoxiong Sun

et al.

Published: Dec. 8, 2023

BACKGROUND Previous mobile health (mHealth) studies have revealed significant links between depression and circadian rhythm features measured via wearables. However, the comprehensive impact of seasonal variations was not fully considered in these studies, potentially biasing interpretations real-world settings. OBJECTIVE This study aims to explore associations severity wearable-measured rhythms while accounting for impacts. METHODS Data were sourced from a large longitudinal mHealth study, wherein participants’ assessed biweekly using 8-item Patient Health Questionnaire (PHQ-8), behaviors, including sleep, step count, heart rate (HR), tracked Fitbit devices up 2 years. We extracted 12 14-day data preceding each PHQ-8 assessment, cosinor variables, such as HR peak timing (HR acrophase), nonparametric features, onset most active continuous 10-hour period (M10 onset). To investigate association also assessing impacts, we used three nested linear mixed-effects models feature: (1) incorporating score an independent variable, (2) adding seasonality, (3) interaction term season score. RESULTS Analyzing 10,018 records alongside 543 participants (n=414, 76.2% female; median age 48, IQR 32-58 years), found that after adjusting effects, higher scores associated with reduced daily steps (β=–93.61, <i>P</i>&lt;.001), increased sleep variability (β=0.96, delayed (ie, onset: β=0.55, <i>P</i>=.001; offset: β=1.12, <i>P</i>&lt;.001; M10 β=0.73, <i>P</i>=.003; acrophase: β=0.71, <i>P</i>=.001). Notably, negative more pronounced spring (β × = –31.51, <i>P</i>=.002) summer –42.61, <i>P</i>&lt;.001) compared winter. Additionally, correlation observed solely 1.06, <i>P=</i>.008). Moreover, winter, experienced shorter duration by 16.6 minutes, increase 394.5, delay 20.5 time 67.9 minutes during summer. CONCLUSIONS Our findings highlight influences on human their depression, underscoring importance considering research applications. indicates potential digital biomarkers depression.

Language: Английский

Framework for Ranking Machine Learning Predictions of Limited, Multimodal, and Longitudinal Behavioral Passive Sensing Data: Combining User-Agnostic and Personalized Modeling DOI Creative Commons
Tahsin Mullick, Sam Shaaban, Ana Radović

et al.

JMIR AI, Journal Year: 2024, Volume and Issue: 3, P. e47805 - e47805

Published: May 20, 2024

Passive mobile sensing provides opportunities for measuring and monitoring health status in the wild outside of clinics. However, longitudinal, multimodal sensor data can be small, noisy, incomplete. This makes processing, modeling, prediction these challenging. The small size set restricts it from being modeled using complex deep learning networks. current state art (SOTA) tackles sets following a singular modeling paradigm based on traditional machine (ML) algorithms. These opt either user-agnostic approach, making model susceptible to larger degree noise, or personalized where training individual alludes more limited set, giving rise overfitting, therefore, ultimately, having seek trade-off by choosing 1 2 approaches reach predictions.

Language: Английский

Citations

1
Digital Phenotyping of Mental and Physical Conditions: Remote Monitoring of Patients Through RADAR-Base Platform DOI Creative Commons
Zulqarnain Rashid, Amos Folarin, Yuezhou Zhang

et al.

JMIR Mental Health, Journal Year: 2024, Volume and Issue: 11, P. e51259 - e51259

Published: May 8, 2024

Abstract Background The use of digital biomarkers through remote patient monitoring offers valuable and timely insights into a patient’s condition, including aspects such as disease progression treatment response. This serves complementary resource to traditional health care settings leveraging mobile technology improve scale lower latency, cost, burden. Objective Smartphones with embedded connected sensors have immense potential for improving various apps (mHealth) platforms. capability could enable the development reliable from long-term longitudinal data collected remotely patients. Methods We built an open-source platform, RADAR-base, support large-scale collection in studies. RADAR-base is modern platform around Confluent’s Apache Kafka scalability, extensibility, security, privacy, quality data. It provides study design setup active (eg, patient-reported outcome measures) passive phone sensors, wearable devices, Internet Things) capabilities feature generation behavioral, environmental, physiological markers). back end enables secure transmission scalable solutions storage, management, access. Results has been used successfully collect cohorts number areas multiple sclerosis, depression, epilepsy, attention-deficit/hyperactivity disorder, Alzheimer disease, autism, lung diseases. Digital developed are providing useful different Conclusions contemporary, solution driven by community monitoring, collecting data, digitally characterizing both physical mental conditions. Clinicians ability enhance their insight biomarkers, enabling improved prevention, personalization, early intervention context management.

Language: Английский

Citations

1
A template and tutorial for preregistering studies using passive smartphone measures DOI Creative Commons
Anna M. Langener,

Björn S. Siepe,

Mahmoud Medhat Elsherif

et al.

Behavior Research Methods, Journal Year: 2024, Volume and Issue: 56(8), P. 8289 - 8307

Published: Aug. 7, 2024

Abstract Passive smartphone measures hold significant potential and are increasingly employed in psychological biomedical research to capture an individual's behavior. These involve the near-continuous unobtrusive collection of data from smartphones without requiring active input participants. For example, GPS sensors used determine (social) context a person, accelerometers measure movement. However, utilizing passive presents methodological challenges during analysis. Researchers must make multiple decisions when working with such measures, which can result different conclusions. Unfortunately, transparency these decision-making processes is often lacking. The implementation open science practices only beginning emerge digital phenotyping studies varies widely across studies. Well-intentioned researchers may fail report on some due variety choices that be made. To address this issue enhance reproducibility studies, we propose adoption preregistration as way forward. Although there have been attempts preregister template for registering currently missing. This could problematic high level complexity requires well-structured template. Therefore, our objective was develop easy use understandable researchers. Additionally, explain provide resources assist making informed regarding collection, cleaning, Overall, aim researchers' explicit, transparency, elevate standards measures.

Language: Английский

Citations

0
Overcoming technical hurdles in mobile health: insights from the Fit2ThriveMB breast cancer study DOI

Qirui Guo,

Mohan Liu, Yan Li

et al.

Breast Cancer Research and Treatment, Journal Year: 2024, Volume and Issue: 208(2), P. 467 - 468

Published: Sept. 5, 2024

Language: Английский

Citations

0
Biomarker-Guided Tailored Therapy in Major Depression DOI
Giampaolo Perna, Alessandro Spiti,

Tatiana Torti

et al.

Advances in experimental medicine and biology, Journal Year: 2024, Volume and Issue: unknown, P. 379 - 400

Published: Jan. 1, 2024

Language: Английский

Citations

0
From smartphone data to clinically relevant predictions: A systematic review of digital phenotyping methods in depression DOI Creative Commons

Imogen E. Leaning,

Nessa Ikani, Hannah S. Savage

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Oct. 5, 2023

Abstract Background Smartphone-based digital phenotyping enables potentially clinically relevant information to be collected as individuals go about their day. This could improve monitoring and interventions for people with Major Depressive Disorder (MDD). The aim of this systematic review was investigate current features methods used in MDD. Methods We searched PubMed, PsycINFO, Embase, Scopus Web Science (10/11/2023) articles including: (1) MDD population, (2) smartphone-based features, (3) validated ratings. Risk bias assessed using several sources. Studies were compared within analysis goals (correlating depression, predicting symptom severity, diagnosis, mood state/episode, other). Twenty-four studies (9801 participants) included. Results achieved moderate performance. Common themes included challenges from complex missing data (leading a risk bias), lack external validation. Discussion made progress towards relating phenotypes clinical variables, often focusing on time-averaged features. investigating temporal dynamics more directly may beneficial patient monitoring. European Research Council consolidator grant: 101001118, Prospero: CRD42022346264, Open Framework: https://osf.io/s7ay4

Language: Английский

Citations

1
Uncovering social states in healthy and clinical populations using digital phenotyping and Hidden Markov Models DOI Creative Commons
Imogen E Leaning, Andrea Costanzo, Raj R. Jagesar

et al.

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: June 8, 2024

Abstract Background Brain related disorders are characterized by observable behavioral symptoms, for example social withdrawal. Smartphones can passively collect data reflecting digital activities, such as communication app usage and calls. This is collected objectively in real time, avoiding recall bias, may therefore be a useful tool measuring behaviors to functioning. Despite promising clinical utility, analyzing smartphone challenging datasets often include range of missingness-prone temporal features. Objective Hidden Markov Models (HMMs) provide interpretable, lower-dimensional representations data, allowing missingness. We aimed investigate the HMM method modeling time series data. Methods applied an aggregate dataset measures designed assess phone-related functioning healthy controls (HCs), participants with schizophrenia, Alzheimer’s disease (AD) memory complaints. trained on subset HCs (n=91) selected model socially “active” “inactive” states, then generated hidden state sequences per participant calculated their “total dwell time”, i.e. percentage spent active state. Linear regression models were used compare total available measures, logistic was times between diagnostic groups HCs. primarily report results from two-state but also verified HMMs more whole dataset. Results identified lower AD (n=26) versus withheld (n=156) (odds ratio=0.95, FDR corrected P <.001), well complaints (n=57) ratio=0.97, =0.004). The result very robust across variations, whilst less robust. observed interaction group when predicting (FDR =0.02). No significant relationships regarding schizophrenia (n=18). Conclusions found practical, interpretable phenotyping analysis, providing objective phenotype that possible indicator

Language: Английский

Citations

0
From Data Streams to Mental Health Predictions: Improving the use of Passive Measures from Digital Devices DOI Creative Commons

Anna Langener

Published: Aug. 13, 2024

Poor mental health is a global concern, with the World Health Organization reporting that one in eight people suffer from disorder. Identification and treatment are hampered by limited access to care inadequate insurance coverage. Digital technologies, such as smartphones, offer promising tools for improving through continuous monitoring timely intervention. These devices can collect rich data on various factors, social context behavior, active (e.g., questionnaires) passive GPS tracking) methods. Researchers often aim use this passively collected predict outcomes. Despite its potential, collection still evolving, current predictive accuracy remains low moderate. The overall goal of thesis therefore optimize measures digital predicting outcomes.The first part focuses Results show combining methods outperforms alone, but performance Advanced machine learning models also only moderate success variability depressive symptoms. second transparency reproducibility studies using measures. It highlights key challenges researchers face provides guidance working measures, example proposing preregistration template. Preregistration involves publicly outlining study plan before research begins, which increase prevent bias.

Language: Английский

Citations

0
Comparison of self-reported and physiological sleep quality from consumer devices to depression and neurocognitive performance DOI Creative Commons
Samir Akre, Zachary D. Cohen,

Amelia Welborn

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 15, 2024

Abstract This study examines the relationship between self-reported and physiologically measured sleep quality in individuals with depression its impact on neurocognitive performance. Using data from 249 participants medium to high monitored over 13 weeks, was assessed via retrospective self-report physiological measures consumer smartphones smartwatches. Correlations were generally weak. Machine learning models revealed that could detect all symptoms Patient Health Questionnaire-14, whereas only detected “sleeping too much” low libido. Notably, disturbances correlated significantly Physiological able changes domains of medication use latency. These findings emphasize are not measuring same construct, both important monitor when studying relation depression.

Language: Английский

Citations

0
Window-Based Quaternion Principal Component Analysis of Eye Gaze Dynamics for Depression Severity Prediction DOI
Kevin Hung, G.M.T. Man,

John Kwok-Tai Chui

et al.

TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON), Journal Year: 2024, Volume and Issue: unknown, P. 1084 - 1087

Published: Dec. 1, 2024

Language: Английский

Citations

0