predicting depression from smartphone behavioral markers

JMIR MHEALTH AND UHEALTH Opoku Asare et al
Original Paper
Predicting Depression From Smartphone Behavioral Markers

Using Machine Learning Methods, Hyperparameter Optimization,
and Feature Importance Analysis: Exploratory Study
Kennedy Opoku Asare1, MSc; Yannik Terhorst2, MSc; Julio Vega3, PhD; Ella Peltonen1, PhD; Eemil Lagerspetz4,
PhD; Denzil Ferreira1, PhD
1
Center for Ubiquitous Computing, University of Oulu, Oulu, Finland
2
Department of Clinical Psychology and Psychotherapy, Ulm University, Ulm, Germany
3
Department of Medicine, University of Pittsburgh, Pittsburgh, PA, United States
4
Department of Computer Science, University of Helsinki, Helsinki, Finland
Corresponding Author:
Kennedy Opoku Asare, MSc
Center for Ubiquitous Computing
University of Oulu
Erkki Koiso-Kanttilan katu 3
PO Box 4500
Oulu, FI-90014
Finland
Phone: 358 294 482807
Fax: 358 8 553 2612
Email: kennedy.opokuasare@oulu.fi
Abstract
Background: Depression is a prevalent mental health challenge. Current depression assessment methods using self-reported
and clinician-administered questionnaires have limitations. Instrumenting smartphones to passively and continuously collect
moment-by-moment data sets to quantify human behaviors has the potential to augment current depression assessment methods
for early diagnosis, scalable, and longitudinal monitoring of depression.
Objective: The objective of this study was to investigate the feasibility of predicting depression with human behaviors quantified
from smartphone data sets, and to identify behaviors that can influence depression.
Methods: Smartphone data sets and self-reported 8-item Patient Health Questionnaire (PHQ-8) depression assessments were
collected from 629 participants in an exploratory longitudinal study over an average of 22.1 days (SD 17.90; range 8-86). We
quantified 22 regularity, entropy, and SD behavioral markers from the smartphone data. We explored the relationship between
the behavioral features and depression using correlation and bivariate linear mixed models (LMMs). We leveraged 5 supervised
machine learning (ML) algorithms with hyperparameter optimization, nested cross-validation, and imbalanced data handling to
predict depression. Finally, with the permutation importance method, we identified influential behavioral markers in predicting
depression.
Results: Of the 629 participants from at least 56 countries, 69 (10.97%) were females, 546 (86.8%) were males, and 14 (2.2%)
were nonbinary. Participants’ age distribution is as follows: 73/629 (11.6%) were aged between 18 and 24, 204/629 (32.4%) were
aged between 25 and 34, 156/629 (24.8%) were aged between 35 and 44, 166/629 (26.4%) were aged between 45 and 64, and
30/629 (4.8%) were aged 65 years and over. Of the 1374 PHQ-8 assessments, 1143 (83.19%) responses were nondepressed scores
(PHQ-8 score <10), while 231 (16.81%) were depressed scores (PHQ-8 score ≥10), as identified based on PHQ-8 cut-off. A
significant positive Pearson correlation was found between screen status–normalized entropy and depression (r=0.14, P<.001).
LMM demonstrates an intraclass correlation of 0.7584 and a significant positive association between screen status–normalized
entropy and depression (β=.48, P=.03). The best ML algorithms achieved the following metrics: precision, 85.55%-92.51%;
recall, 92.19%-95.56%; F1, 88.73%-94.00%; area under the curve receiver operating characteristic, 94.69%-99.06%; Cohen κ,
86.61%-92.90%; and accuracy, 96.44%-98.14%. Including age group and gender as predictors improved the ML performances.
Screen and internet connectivity features were the most influential in predicting depression.
https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 1

(page number not for citation purposes)
XSL• FO
RenderX
Conclusions: Our findings demonstrate that behavioral markers indicative of depression can be unobtrusively identified from
smartphone sensors’ data. Traditional assessment of depression can be augmented with behavioral markers from smartphones
for depression diagnosis and monitoring.
(JMIR Mhealth Uhealth 2021;9(7):e26540) doi: 10.2196/26540
KEYWORDS
mHealth; mental health; mobile phone; digital biomarkers; digital phenotyping; smartphone; supervised machine learning;
depression
Related Work
Introduction
A growing body of research in smartphone and wearable
Background sensing, human behavior modeling has improved our
Depression is one of the most prevalent, complex, and understanding of the relationship between mental health and
heterogeneous mental health challenges of our time. In 2020, biomarkers [3,20-22,26,27,29-31]. In medicine, biomarkers are
the World Health Organization (WHO) estimated that depression pathological, anatomical, or physiological characteristics that
has impacted 264 million people worldwide [1], and it is are quantified and evaluated as indicators of a biological process
projected to be the leading contributing factor to global disease or a response to medical interventions [23]. Here, we define
burden by 2030 [2]. In these individuals, depression inflicts biomarkers (or digital biomarkers) of mental health as
recurrent episodes of guilt, sadness, cognitive impairments, quantifiable behaviors (or features) extracted from smartphone
suicidal ideation, and sleep disturbances [1,3-5]. Depression or wearable data that can be monitored and collected over time
increases the risk and medical costs of many medical disorders to objectively assess mental health and effectiveness of
such as stroke, Parkinson, or Alzheimer [6-11]. Depression is interventions. Monitoring these biomarkers’ fluctuations is
treatable with psychotherapy and medication. Yet, in many essential in the early detection and treatment of mental health
individuals with depression, it remains undiagnosed and disorders [3,32]. For example, in Alzheimer disease, biomarkers
untreated due to barriers such as social stigma and inaccurate such as cognitive, sensory, and motor degeneration precedes
assessment methods [1,3,12,13]. The ability to detect early clinical diagnosis for about 10 or 15 years [32].
warning signs of depression, continuously and as effortlessly In the StudentLife study [22], for example, geolocation, sleep,
as possible, by extending current assessment methods could and activity-based biomarkers were extracted from a data set
have a significant impact in mitigating or addressing depression collected from 48 students over 10 weeks. Significant
and its related negative consequences [3,10,11,14]. correlations were found between the following digital
For the past 30 years, clinician-administered and self-reported biomarkers: sleep duration (r=–0.360), activity duration
questionnaires remain the gold standard in the assessment and (r=–0.388), traveled distance (r=0.338), and various mental
diagnosis of depression [3,13]. However, the limitations of these health symptoms. Similarly, Wang et al [3] collected data sets
traditional depression assessment methods have been debated. with the StudentLife sensing app from 83 college students across
Such methods are applied sparingly (eg, a couple of times within two 9-week terms. Significant correlations were found between
a year), thus missing out on the moment-by-moment behavioral depression and accelerometer-based biomarkers (mean stationary
patterns of individuals between health assessments. Lastly, time, r=0.256) and screen usage–based biomarkers (mean unlock
self-reported appraisals are affected by memory and recall biases duration, r=0.283). With an analysis of variance, a significant
in reconstructing past events and may be prone to socially difference in unlock duration (F=5.733) was found between
desirable reporting from individuals [12,15-17]. depressed and nondepressed groups. Saeb et al [33] in their
study of 40 participants for 2 weeks also found a statistically
Today, smartphones and wearables offer a unique opportunity significant correlation between depression and GPS
to overcome limitations in traditional depression assessment location–based biomarkers (ie, variance in locations visited,
methods. Smartphones and wearables (eg, Fitbit, Oura Rings, [r=0.58] and regularity in 24-hour movement rhythm [r=0.63])
and smartwatches) have become ubiquitous in the global and phone usage–based biomarkers (ie, phone usage frequency
population, they are inherently personal, and people are [r=0.54]).
continuously monitored through their embedded sensors (eg,
camera, accelerometer, global positioning system [GPS], Another promising source of biomarkers is wearable devices
Bluetooth, and many more) [18,19]. Instrumenting smartphones [30].
and wearables to capture in situ, fine-grained, and Actigraphy-based biomarkers that quantify time sequences of
moment-by-moment data sets with sensing apps [20-24] has rest and active behaviors with accelerometer sensors are known
made it possible to passively collect data sets in naturalistic to be useful in predicting mood disorders such as depression
settings. Inherent in these data sets are behavioral patterns: and bipolar disorder [34,35]. In a 2-week study on 23
routines, rhythms, activities, and interactions that are useful in participants, Jacobson et al [34] extracted biomarkers from data
complementing traditional depression assessment methods, in sets collected with wrist-worn actigraph watches. A machine
studying the mental health of individuals, and in developing learning (ML) model trained with these digital biomarkers could
timely mental health interventions [14,22,25-28]. predict the depression status of participants with high accuracy

XSL• FO
RenderX
(ie, accuracy 89%, Cohen κ=0.773). In another 2-week study the institutional review board license from the University of
[35] on 40 geriatric participants, accelerometer-based digital California, Berkeley and the University of Helsinki [21,40].
biomarkers were extracted from wrist-worn actigraph watches.
With 4 ML models, the study found that these biomarkers could
Mobile Sensing Variables Collected by Carat
predict depression with a high accuracy (ie, accuracy 0.910, Besides battery consumption data, the Carat Android app
precision 0.929, and specificity 0.940). Other promising unobtrusively collected participants’ time zone and
biomarkers from wearable devices are heart rate variability, time-stamped data, including foreground app usage (ie, app the
which has been found to be consistently lower in patients with participant has interacted with), internet connectivity (ie,
psychiatric disorders, electrodermal activity, and skin connected and disconnected states), and screen lock and unlock
conductance [36,37]. logs. This data set was sampled at each 1% battery change (ie,
while charging or discharging) on the participants’ smartphone.
Taken together, previous research has shown the potential of The Carat app also collected participant’s demographic
quantifying human behavior from smartphones and wearables information, including age, gender, education, and occupation
data set as biomarkers. These biomarkers are insightful in via a self-report.
understanding depression.
Mental Health Assessment
Objectives
In addition to the mobile sensing and demographic variables,
In this study, we aim to investigate the feasibility of predicting depression severity was assessed by a self-report instrument.
depression using digital biomarkers quantified from a Participants answered the 8-item Patient Health Questionnaire
smartphone data set. To this end, we explore the relationship (PHQ-8) [41] at 2-week intervals as specified by the PHQ-8
between digital biomarkers and depression severity with protocol. Although the depression assessments were
statistical methods. We investigate whether depression can be self-reported, the PHQ-8 is clinically validated for the
predicted with digital biomarkers using supervised ML assessment of depression severity, has high internal consistency
algorithms. (Cronbach α=.82), and has been used in several previous studies
[3,7,41]. PHQ-8 measures depression severity for the past 2
Methods weeks with items such as “Little interest or pleasure in doing
things,” “Feeling down, depressed, or hopeless,” “Trouble
The Data Set
falling or staying asleep, or sleeping too much.” Each item of
We utilized an existing data set collected in a longitudinal the PHQ-8 is scored on a scale from 0 (Not at all) to 3 (Nearly
observational study with the Carat app [21], derived from a every day). The total PHQ-8 score ranges from 0 to 24, with a
cohort of anonymous Android participants worldwide [38]. The score of 10 or more indicating major depression or other severe
data set was collected from 843 participants between March forms of depression [41].
and August 2018 (~6 months).
Data Inclusion and Exclusion
The Carat app is a mobile sensing app, originally developed by
For each participant’s data set, we excluded days with at least
a team of researchers from the University of Helsinki and the
10 missing log intervals (ie, days where no data were logged
University of California, Berkeley, for smartphone energy
by the Carat app for at least 10% battery charging or discharging
consumption research [21,39]. The Carat app is freely available
periods). Next, we only included PHQ-8 responses from
on mobile app stores and gives users personalized smartphone
participants with at least 8 days of data within the preceding 2
battery consumption reports. Anonymous users worldwide who
weeks of PHQ-8 response. Consequently, the final data
install the Carat app may voluntarily be recruited to contribute
contained 629 participants, 1374 PHQ-8 responses with 13,898
their data set to research.
days of participants’ data set.
The data set used in this study was a subset of the large-scale
crowdsourced Carat app data set from anonymous volunteers. Feature Engineering
The study data set was collected for a multifaceted purpose, Characterization of Data Set
which includes studying the relationship between smartphone
Our data set is primarily categorized into screen status, internet
app usage and Big 5 personality traits [40]; studying the
connectivity, and foreground app usage logs. For data
similarities and differences in demographic, geographic, and
preprocessing, we converted the time stamps of the data set to
cultural factors of smartphone usage [38]; and mental health
local date and time, using the participants’ time zone. We
research. The advertisement for the recruitment of participants
computed digital biomarkers (herein features) by quantifying
was sent as push notifications through the Carat app to 25,323
the per-participant hourly and daily behavioral patterns (ie,
verified users (ie, users with matching time zone and mobile
routines, irregularity, variability) from these data sets with
country code) [38].
simple counts, SDs, entropy [6,14,25,42-45], and regularity
All participants in this data set are Android-based smartphone index [27,44] measures. All computed features were merged
users, who explicitly and voluntarily gave their consent from per participant at the day level.
their mobile devices after they were informed about the purpose
of the data collection, the data collection procedures, and
Entropy
management of the data set. The data set does not contain We computed entropy to capture the degree of variability,
personally identifiable information, and was collected under complexity, disorder, and randomness in the participant behavior

XSL• FO
RenderX
states from screen status (ie, on and off states), internet Correlation and Association Analysis
connectivity (ie, disconnected and connected states), and Before beginning the statistical analysis, we pooled (ie,
foreground app (ie, the frequency of use per app) over a 24-hour aggregated) the extracted features within the preceding 2 weeks
period of each day. Entropy was calculated using the Shannon (ie, assessment window) of each PHQ-8 response from a
entropy [45] formula: participant. The pooling is to ensure that the timelines of the
feature variables in the analysis are aligned with those of the
PHQ-8 assessment window. The pooling was done as follows:
for each PHQ-8 response, we pooled all entropy and regularity
index features by computing the average feature values for all
where N is the number of states and pi is the percentage of the days within the PHQ-8 assessment window. Instead of average
state i in the time series data. For example, a higher screen status values for SD, we took a different approach due to the additive
entropy reflects the fact that the participant’s screen on and off properties of SD measures. For SD features, we computed the
pattern is more distributed between on and off states, albeit with pooled SD [46,47].
a high degree of uncertainty and complexity in the transition
For correlation analysis, we used the pooled data to quantify
between the screen on and off states in a 24-hour period.
the linear relationship between the features and depression
Conversely, a lower screen status entropy reflects that fact that
severity (ie, PHQ-8 responses). The correlations were computed
the participant’s screen is much often in one state (on or off)
using the Pearson correlation coefficient. Full information
over a 24-hour period. In addition to entropy, we computed
maximum likelihood [48-50] was used in the correlation analysis
normalized entropy as the entropy divided by log(N).
to avoid biases introduced by missing data. We used the
Regularity Index Holm–Bonferroni [51] method to adjust the P values for multiple
Regularity index quantifies routines in participant behaviors by testing, with a false discovery rate of 0.05.
capturing the similarity (or difference) in participant behaviors For association analysis, we used the bivariate linear mixed
between the same hours across different days. For internet model (LMM) [8,44,52-55] to study the association between
connectivity, for instance, the regularity index quantifies the the pooled features and the PHQ-8 response. The data set in
routineness of the participant’s internet connectivity behavior this study is a longitudinal data set with repeated measures from
at the same hours (eg, every 9 am) for all days. We determined the same individuals. Given this nested structure, the assumption
the hourly values as follows: for screen status, the modal screen of normally and independently distributed residuals would be
status for each hour; for internet connectivity, the modal violated in linear regression models. Hence, we opted for LMM,
connectivity state for each hour; and for foreground app usage, which takes into account fixed and random variations in the
the number of distinct apps usage for each hour. data set in respect of a grouping variable, the participant in this
Following the regularity index computation method of Wang case. LMM also reduces the likelihood of Type I error [56]. To
et al [44], we computed the regularity index of the screen, verify our decision for LMM, we computed the intraclass
internet connectivity, and foreground app usage for days a and correlation (ICC). ICC > 0.05 necessitates LMM.
b using the formula In the LMM, we used multiple imputation to handle missing
data, taking into account the nested structure of the data set [57].
Using predictive mean matching for multilevel data [58], a total
of 20 imputed data sets were generated. We used Robin’s rule
[59] to pool the results of the LMM run with each imputed data
where a and b are 2-day pairs, T=24 hours, and is the rescaled set. In the LMM analysis, all features were normalized to have
(ie, between –0.5 and 0.5) value of hour t of day b. For each a 0 mean and a unit SD. To address multiple comparison
day, we computed the average regularity index for all problems, we adjusted the P values in the LMM to control the
combinations of that day and other days of the week. false discovery rate with the Benjamini–Hochberg procedure
[60,61]. Adjusted P value <.05 was considered to be significant.
Standard Deviation and Counts
The SD features capture the variance of daily behavior between Predictive Analysis With Machine Learning
4-day epochs based on the hour of the day. We defined morning Machine Learning Setup
as the 6-11th hour, afternoon as the 12-17th hour, evening as
We developed population-based supervised ML classifiers to
the 18-23rd hour, and night as the 0-5th hour of the day. We
explore how the digital biomarkers/features predict the
computed the count of each screen status, the count of each
depression state of an individual. We also explored whether
internet connectivity status, and the count of foreground app
including participants’ self-reported demographics as features
usage per day epoch. With these counts per day epoch, we
would improve the ML classifier performance.
computed the SD per day.
To this end, we used 5 supervised ML models: random forest
We also computed the day level count of each screen status, the
(RF), support vector machine (SVM) with radial basis function
count of each internet connectivity status, and the count of
(RBF) kernel, XGBoost (XGB) [62], K-nearest neighbor (KNN),
foreground app usage. Additionally, we computed the count of
and logistic regression (LR). With the RBF kernel, an SVM
minutes until the first and last use of foreground app per day.
classifier, a linear classifier, can classify nonlinear data sets

XSL• FO
RenderX
[63,64]. These algorithms have been used in previous work cross-validation. For each ML classifier, we ranked the features
[65-69] on mental health studies. We used the same pooled and by the average feature importance computed across all 10 folds
imputed data set from statistical analysis, but ensured that all of the outer cross-validation.
records are distinct. We labeled our data set with 2 classes,
based on PHQ-8 scoring guidelines, where PHQ-8 score ≥ 10
Model Evaluation
is depressed (label 1) and PHQ-8 score <10 is nondepressed We created 2 baseline classifiers to benchmark the performance
(label 0). We created 2 data sets for the ML modeling: (1) a of the ML classifiers. The first baseline is a random weighted
data set with PHQ-8 scores as labels, and digital classifier (RWC) with 10,000 randomly generated predictions
biomarkers/features as predictors, and (2) a second data set with based on a multinomial distribution of the nondepressed and
PHQ-8 scores as labels and digital biomarkers/features, age depressed classes. The second baseline is a decision tree (DT)
group, and gender as predictors. The age group and gender classifier trained using the same approach as the ML classifiers,
demographics were converted from categorical to numerical but with age group and gender as features. The performance of
data using one-hot encoding [63,70]. the ML classifiers and baseline classifiers was measured using
the following performance metrics: accuracy, precision, recall,
All the ML modeling was performed using stratified and nested AUC, F1 score, and Cohen κ. The precision, recall, and F1
cross-validation [71,72]. The nested cross-validation is a scores were computed with an emphasis on predicting the
state-of-the-art procedure to prevent overfitting and depressed score (ie, label 1).
overestimation of the hyperparameters of the ML classifiers.
Stratified 10 folds in the outer cross-validation and stratified 3 Software
folds in the inner cross-validation were used. With this approach, Data preprocessing and feature extraction pipeline were created
for each iteration of the outer cross-validation, 1 stratified fold with Python (version 3.7.6) and R (version 4.0.2) programming
was used as a testing data set. The remaining 9 stratified folds languages, using Snakemake [77] and RAPIDS [78] for
were used for hyperparameter optimization in the inner workflow management. All statistical analysis was performed
cross-validation. in R, with mice [79] package for multiple imputation, lmer4
The hyperparameter optimization in the inner cross-validation [80] and lmerTest [81] packages for LMM, and psych [50]
was done with grid search over a grid of parameters, where all package for computing correlation. All the ML was done in
combinations of hyperparameters are exhaustively considered. Python, with scikit-learn [63], imbalanced-learn [33], and XGB
We use the F1 (macro averaged) score to select the most library [62,82].
optimized hyperparameters.
Results
Imbalanced Data Handling
It is worth noting that in the ML setup, we used stratified Participants’ Demographics
sampling in the nested cross-validation. The stratified sampling Self-reported demographic data from the 629 participants
ensures the splitting of the data set into folds that have an equal included in our analyses show that 69/629 (10.97%) were
proportion of each class (ie, labels 1 and 0). However, the females, 546/629 (86.8%) were males, and 14/629 (2.2%) were
proportion of each class is still dependent on its availability in nonbinary or preferred not to disclose their gender.
the data set. We handled class imbalance in the training data
For the participants’ age distribution, 73/629 (11.6%) were aged
set with the synthetic minority over-sampling technique
between 18 and 24, 204/629 (32.4%) were aged between 25
(SMOTE) [65,73,74], which generates synthetic data for the
and 34, 156/629 (24.8%) were aged between 35 and 44, 166/629
minority class, resulting in a balanced training data set.
(26.4%) were aged between 45 and 64, and 30/629 (4.8%) were
Feature Analysis aged 65 years and over. The participants were distributed across
We used the permutation importance method [75,76] to compute at least 56 different countries, including 91/629 (14.5%) from
the importance of features. The permutation importance method unknown countries, 199/629 (31.6%) from the USA, 66/629
is model agnostic and computes the proportional decrease in a (10.5%) from Finland, 32/629 (5.1%) from Great Britain, 42/629
model score when features are randomly shuffled. We used the (6.7%) from Germany, and 29/629 (4.6%) from India. The data
area under the curve (AUC) receiver operating characteristic as set also has participants from varied educational and
the model in the permutation importance computation. We occupational backgrounds. Table 1 provides summary statistics
computed the feature importance using the test set of the outer of the 629 participants in this study.

XSL• FO
RenderX
Table 1. Summary statistics of participants who were included in the data analysis (N=629).
Variable Value, n (%)
Age (years)
18-24 73 (11.6)
25-34 204 (32.4)
35-44 156 (24.8)
45-64 166 (26.4)
≥65 30 (4.8)
Gender
Female 69 (11.0)
Male 546 (86.8)
Other or Rather not tell 14 (2.2)
Education
Elementary school/basic education 9 (1.4)
High school/sixth form/other upper secondary level 98 (15.6)
No education or rather not to tell 5 (0.8)
Professional graduate degree/higher university degree (master’s or equivalent) 193 (30.7)
Research graduate degree (PhD or equivalent) 34 (5.4)
Undergraduate degree/lower university degree (bachelor’s or equivalent) 228 (36.2)
Vocational school/trade school/other education leading to a profession 62 (9.9)
Occupation
Agricultural forestry or fishery 1 (0.2)
Clerical support 14 (2.2)
Craft and trade or plant and machine operations 8 (1.3)
Entrepreneur or freelancer 30 (4.8)
Manager 59 (9.4)
No suitable option or rather not to tell 34 (5.4)
Professional 227 (36.1)
Retired 39 (6.2)
Sales or services 29 (4.6)
Staying at home (eg, with kids) 5 (0.8)
Student 74 (11.8)
Technician or associate professional 90 (14.3)
Unemployed or between jobs 19 (3.0)
Country
Unknown 91 (14.5)
USA 199 (31.6)
Finland 66 (10.5)
Great Britain 32 (5.1)
Germany 42 (6.7)
Canada 16 (2.5)
India 29 (4.6)
Othera 154 (24.5)

XSL• FO
RenderX
a
Comprising 49 different countries with less than 15 participants, including South Africa, Morocco, Brazil, Philippines, Qatar, Japan, Russia, and
Denmark.
For the distribution of the PHQ-8 scores, 1143/1374 (83.19%)

Smartphone Data and PHQ-8 Distribution responses were nondepressed scores (PHQ-8 score <10), while
We had 1374 PHQ-8 responses. Table 2 presents the distribution 231/1374 (16.81%) were depressed scores (PHQ-8 score ≥10).
of participants and their corresponding number of responses in The mean PHQ-8 score is 5.19 (SD 5.22; range 0-24).
the PHQ-8 data set. All PHQ-8 responses were collected every
2 weeks. At the minimum, 316/629 (50.2%) participants The number of smartphone data set days was 13,898 for all 629
responded 1 time to the PHQ-8 depression assessment, and at participants. Table 3 shows the distribution of participants and
the maximum, 1/629 (0.2%) participants responded 7 times. their corresponding number of days in the smartphone data set.
The mean number of responses per participant is 2.18 (SD 1.57). The mean number of days per participant is 22.1 (SD 17.90;
range 8-86).
Table 2. Distribution of participants’ contribution to the PHQ-8a responses (N=629).

Participants, n (%) PHQ-8 responses, n
316 (50.2) 1
129 (20.5) 2
57 (9.1) 3
47 (7.5) 4
40 (6.4) 5
39 (6.2) 6
1 (0.2) 7
a
PHQ-8: 8-Item Patient Health Questionnaire.
Table 3. Distribution of participants’ smartphone data set days (N=629).

Days, n Participants, n (%)
8-14 364 (57.9)
15-28 126 (20.0)
29-42 53 (8.4)
43-56 34 (5.4)
57-70 29 (4.6)
71-84 22 (3.5)
85-98 1 (0.2)
Regarding the association analysis, we found an ICC of 0.7584;

Features Engineered From Smartphone Data Set thus, 75.84% of the variations in the features are explainable
In all, we computed 22 features from the smartphone data set. by the interindividual differences. We found a significant
All features were aggregated at the day level. For example, the positive association between screen status–normalized entropy
screen_offCount feature is the count of all screen off states and depression (β=.48, P=.03). We found no significant
during the day. We summarize the engineered features in association between other screen, app, and internet connectivity
Multimedia Appendix 1. features and depression. Multimedia Appendix 2 presents the
results of the LMM analysis showing the estimates (β) and
Correlation and Association Between Features and
adjusted P values calculated using the Benjamini–Hochberg
Depression
method.
We found a significant positive correlation between screen
status–normalized entropy and depression (r=0.14, P<.001). Predicting Depression From Features
We found no significant correlation between other screen, app, The overall performance of the ML classifiers trained with
and internet connectivity features and PHQ-8 depression score. features only (ie, with no demographics data set) is listed in
Multimedia Appendix 2 presents the full Pearson correlation Table 4. Table 5 shows the ML classifier performance with the
coefficients and adjusted P values, with the Holm–Bonferroni features plus age group and gender data set as predictors. The
method, between features and PHQ-8 depression score. tuned hyperparameters of all ML classifiers are detailed in
Multimedia Appendix 3. The performance of all 10

XSL• FO
RenderX
cross-validation folds of the features-only data set and that of As shown in Table 4, the worst performing classifier is LR.
the feature plus demographics data set are detailed in Multimedia Compared with the baseline classifiers in Table 6, the RWC
Appendix 3. The performance of the DT and RWC baselines is and DT baselines classifiers outperform the LR classifier in
shown in Table 6. terms of recall. The LR could predict the PHQ-8 depression
score barely better than the RWC and DT baselines in all other
As shown in Table 4, it is evident that nonlinear classifiers such
performance metrics. The RWC baseline classifier also
as XGB, RF, and KNN had superior performance in all metrics
outperformed the SVM classifier on the recall metrics.
than LR. SVM (ie, SVM with RBF kernel) also performed better
than LR. When age group and gender were included with features as
predictors, we observed a general improvement in all
In terms of precision, recall, and F1 scores, which were
performance metrics for all classifiers, as shown in Table 5.
computed with an emphasis on the predictive performance of
The SVM classifier had the most substantial improvement with
the positive label (ie, depressed score, PHQ-8 ≥ 10), XGB was
precision increasing by 18.48% and Cohen κ increasing by
the best performing classifier, followed by RF and KNN. XGB,
19.3%. KNN had a 6.58% improvement in precision and a 5.1%
RF, and KNN performed better than the RWC and DT baselines,
improvement in F1 score. RF and XGB classifiers had marginal
as shown in Table 6.
improvements in all performance metrics. The worst performing
Likewise, with AUC and Cohen κ performance metrics, which classifier (ie, LR) had some gains in performance, but it could
take into consideration both positive and negative labels (ie, still barely outperform the DT baseline classifier in Table 6 in
nondepressed score, PHQ-8 <10), XGB, RF, and KNN all performance metrics.
classifiers had the best performance, as shown in Table 4. The
AUC and Cohen κ are not biased by imbalance labels.
Table 4. Average and SDs of accuracy, precision, recall, F1, area under the curve, and Cohen κ metrics for 10-fold cross-validation, with features-only
data set as predictors.
Metric RFa, mean (SD) XGBb, mean (SD) SVMc, mean (SD) LRd, mean (SD) KNNe, mean (SD)
Accuracy 97.97 (0.37) 98.14 (0.37) 85.68 (1.16) 59.27 (1.45) 96.44 (0.52)
Precision 92.50 (1.78) 92.51 (1.25) 51.98 (2.58) 20.29 (1.25) 85.55 (1.97)
Recall 94.38 (1.86) 95.56 (1.99) 80.67 (2.36) 57.25 (4.14) 92.19 (2.24)
F1 93.41 (1.19) 94.00 (1.21) 63.20 (2.29) 29.95 (1.87) 88.73 (1.63)
Area under the curve 98.83 (0.67) 99.06 (0.54) 89.47 (1.06) 62.43 (2.22) 94.69 (1.15)
Cohen κ 92.21 (1.41) 92.90 (1.43) 54.83 (2.92) 9.66 (2.38) 86.61 (1.93)
a
RF: random forest.
b
XGB: XGBoost.
c
SVM: support vector machine.
d
LR: logistic regression.
e
KNN: K-nearest neighbor.
Table 5. Average and SDs of accuracy, precision, recall, F1, area under the curve, and Cohen κ metrics for 10-fold cross-validation, with features, age
group, and gender data set as predictors.
Metric RFa, mean (SD) XGBb, mean (SD) SVMc, mean (SD) LRd, mean (SD) KNNe, mean (SD)
Accuracy 98.55 (0.40) 98.56 (0.31) 92.61 (0.46) 60.37 (1.39) 98.09 (0.26)
Precision 95.65 (1.59) 94.93 (1.08) 70.46 (1.63) 21.40 (1.20) 92.13 (1.41)
Recall 94.78 (1.59) 95.62 (1.52) 88.76 (3.19) 60.00 (3.94) 95.62 (1.56)
F1 95.20 (1.31) 95.27 (1.03) 78.52 (1.41) 31.54 (1.78) 93.83 (0.85)
Area under the curve 99.01 (0.51) 99.36 (0.33) 95.45 (1.00) 66.62 (3.06) 97.07 (0.73)
Cohen κ 94.34 (1.55) 94.42 (1.21) 74.13 (1.66) 11.74 (2.28) 92.69 (1.00)
a
RF: random forest.
b
XGB: XGBoost.
c
SVM: support vector machine.
d
LR: logistic regression.
e
KNN: K-nearest neighbor.

XSL• FO
RenderX
Table 6. Average and SDs of accuracy, precision, recall, F1, area under the curve, and Cohen κ metrics for the RWC and DT baselines.
Metric RWCa, mean (SD) DTb, mean (SD)
Accuracy 25.80 (0.33) 46.80 (3.77)
Precision 15.21 (0.36) 18.70 (0.66)
Recall 84.79 (0.86) 74.33 (4.71)
F1 25.80 (0.24) 29.85 (0.75)
Area under the curve 50.00 (0.47) 62.94 (1.38)
Cohen κ 0.00 (0.32) 7.33 (1.26)
a
RWC: random weighted classifier; RWC metrics is the average and SD of 10,000 random predictions.
b
DT: decision tree; DT metrics is the average for 10-fold cross-validation, with age group and gender only as features.
Likewise, for the KNN classifier in Figure 3, the top 5 most

Feature Importance Analysis important features are the internet regularity index, screen
We present the mean permutation feature importance in status–normalized entropy, screen regularity index, internet
predicting PHQ-8 depression score across the 10-fold status–normalized entropy, and the internet status entropy. As
cross-validation with the top 3 performing ML classifiers (ie, shown in Figure 3 app entropy, count, distinct count, regularity
XGB, RF, and KNN) in Figures 1-3. index, and count SD were less important to the KNN classifier.
For the XGB classifier in Figure 1, the top 5 most important Removing these less important features could further improve
features were the internet regularity index, screen on count, the performance of the KNN classifier.
screen regularity index, screen status entropy, and the screen By ranking all important features for KNN, XGB, and RF
off count. classifiers, the top 5 most were the screen regularity index,
For the RF classifier in Figure 2, the top 5 most important screen status entropy, internet regularity index, screen
features were screen status–normalized entropy, screen status–normalized entropy, and the screen off count SD. App
regularity index, screen off count SD, screen off count, and count SD is the least important feature for all classifiers (Figures
internet regularity index. 1-3), and could be removed to improve ML classifier
performance in the case of RF and KNN.
Figure 1. Mean permutation feature importance across 10-fold cross-validation with the XGBoost machine learning classifier.

XSL• FO
RenderX
Figure 2. Mean permutation feature importance across 10-fold cross-validation with the random forest machine learning classifier.
Figure 3. Mean permutation feature importance across 10-fold cross-validation with the K-nearest neighbor machine learning classifier.

XSL• FO
RenderX
biomarker quantified the frequency and distribution (ie,

Discussion complexity and uncertainty) in the transition of the participants’
Overview of Data Set Employed phone screen on and off states. All other indicators were
nonsignificant. Moreover, correlation coefficients only measure
Our objective was to investigate the feasibility of predicting the extent of the linear relationship between variables [87,88].
depression using multivariate digital biomarkers quantified from Instead of correlations, previous studies on mental health relied
smartphone data sets collected in a real-world study. In this on the mutual information (MI) method from Information theory
study, we used 13,898 days of smartphone data set, and 1374 [87-89]. The advantage of using the MI method is that the MI
PHQ-8 depression assessments from 629 participants to explore measures both linear and nonlinear statistical dependencies
the feasibility of detecting depression from participants’ between variables.
behavioral markers (ie, digital biomarkers) quantified from their
smartphones. We focused on finding the relationship between Given the high ICC, we tested whether LMM, a much robust
repeated measures of depression scores and participant’s digital method for finding linear relationships, can identify additional
biomarkers and developing predictive models to classify linear relationships. However, the results from the association
depressed and nondepressed symptom severity scores. analyses further showed that a unit increase in the screen
status–normalized entropy positively increases the average
Principal Results depression score (β=.48, P=.03), but all other variables remained
This data set was collected from a heterogeneous geographic nonsignificant. This suggests that conventional statistical
(ie, from at least 56 different countries), occupational, and methods (eg, correlation or LMM) may not depict the complex
educational population, with high interindividual differences nonlinear relationship between digital biomarkers and
(ie, 75.84% interclass correlation). depression, and indeed more powerful methods such as ML
models (eg, XGB) are needed to make better predictions [90].
Despite this heterogeneity, digital biomarkers extracted from
participants’ smartphone data set were able to predict The heterogeneity and inconsistency in correlation findings (ie,
participants’ depression state (ie, depressed or nondepressed) linear relationships) in the field are common issues [26]. Until
with high predictive performance using ML models. The ML now, it is unclear whether this is due to differences in
models achieved the following: precision, 85.55%-92.51%; sociodemographic characteristics in samples, in used sensors,
recall, 92.19%-95.56%; F1, 88.73%-94.00%; AUC, in the method to calculate features, small sample sizes and lack
94.69%-99.06%; Cohen κ, 86.61%-92.90%; and accuracy, of power, or even due to other between-study factors.
96.44%-98.14%. These findings show that predictive modeling Meta-analysis on digital markers and health outcomes (eg,
of mental health using digital biomarkers is not only possible depression) would be highly valuable to clearly show whether
in small homogenous populations [83,84], but also in a more and to which extent linear relationships exist. Using
general population, which further supports the scalability of meta-regression, the factors causing the differences in
this approach and its potential positive impact on health care if correlation findings may also be identified.
implemented (eg, early detection of mental disorders, RED-flag
Nevertheless, both the correlation findings and the feature
systems after treatment).
importance analysis in the prediction models clearly showed
Moreover, we found that the predictive performances of ML that participants’ phone screen (lock and unlock) behaviors,
classifiers improved when demographic characteristics were such as routinely and randomly locking and unlocking phone
included among predictors, indicating that such variables should screen, and internet connectivity behaviors played the most
also be included in clinical applications. Previous studies suggest important role in predicting their depression state. The findings
a relationship between demographic factors, smartphone usage in this study are also supported by prior research that
behavior, and depression [38,85,86]. Thus, encoded in the investigated the relationship between screen interactions and
demographic data of this study’s population is additional mental health [3,27,28,33]. Passively sensed participants’
information that is useful in predicting the depression state of smartphone screen interaction (ie, on and off states) behavior
the participants. Therefore, the inclusion of additional data from was demonstrated to be an important predictor of mental health
clinical information systems (eg, blood parameters, previous [27]. Similar findings have been reported previously [3,33],
clinical diagnosis) might be a valuable way to further increase where the number of times a participant interacts with their
the performance of prediction models. phone, including screen lock and unlocks, was found to correlate
with participants’ mental health state. In a neuroscience study
Interestingly, tree-based, nearest neighbor–based classifiers had
[91], screen unlocks were found to be important behavioral
superior performance over linear classifiers, including SVM
markers that correlate and predict resting state brain functional
with RBF kernel, corroborating the existence of nonlinear
connectivity, which is known to be associated with depression
relationships between digital biomarkers and depression. This
[92]. On internet usage behaviors, research has demonstrated
finding further supported the correlation finding, which failed
an association between internet usage patterns and depression
to replicate previous results reported in Saeb et al [33]. We
[93,94], which was also a key feature in our analysis. Thus,
could only identify that participants with depression symptoms
including these features in future studies is highly recommended.
were more likely to lock and unlock their phone’s screen in a
random and uncertain manner (ie, significant positive correlation Limitations and Future work
between screen status–normalized entropy and depression, Given the crowdsourced nature of the deployment of the Carat
r=0.14, P<.001). The screen status–normalized entropy app, the sample size in the data set is small (N=629) and may
XSL• FO
RenderX
not be representative of the general population. Despite the data digital biomarkers of participants using Android, iOS, and other
set having a fair distribution of age groups, with a spread over mobile platforms).
several countries, it is biased toward highly educated and
Replicating the findings from this study with additional
professional occupations. The data set is also biased toward
biomarkers from GPS and wearable sensor data sets and
males in gender distribution. Future research with a larger
comparing their correlation and biomarker predictive importance
sample size and a balanced gender distribution could explore
will be interesting in future work. There would be a major design
correlations, associations, and prediction performance for
implication for depression intervention development if the
population subgroups.
behavioral markers from screen and internet connectivity
Clinical diagnosis of depression was not an inclusion criterion achieve similar promising results as biomarkers from GPS and
for our sample population. The data set also does not contain a wearable devices. We hypothesize that screen interaction and
clinical or self-reported baseline assessment of depression and internet connectivity data sets alone are less privacy intrusive,
has scarce high depression scores. Because of the crowdsourced could better capture behaviors of immobile persons, and people
rolling recruitment nature of participants, the data set contained may be more willing to donate such data sets to science. For
an unequal number of repeated depression assessments for all example, Apple’s Screen Time and Google’s Digital Wellbeing
participants. Future research should benefit from replicating the app are processing and presenting such data to users to inform
experiment in a clinical population and a more controlled where and how users spent their time on smartphones.
experimental design. With a clinical baseline data set, future
research could study the differences in features between
Conclusions
depressed and nondepressed groups. In summary, this study sought to find whether we can detect
changes in human behavior that would be indicative of
The correlation and association between behavioral patterns depression using smartphones. In addition, we sought to find
extracted from the data set in this study and depression do not what objective measures of human behavior from smartphones
necessarily imply causal relationships. For example, the are insightful in understanding depression. Our results
correlation between screen status–normalized entropy and established a positive statistically significant linear correlation
depression may be caused by other confounding variables. In and association between depression and screen
addition, the correlation and association between screen status–normalized entropy behavior quantified from smartphone
status–normalized entropy and depression are not strong and data sets. Our findings also establish that behavioral markers
may not generalize in other populations. Further research is extracted from smartphone data sets can predict whether or not
needed to establish the extent to which such behaviors cause or a participant is depressed based on the PHQ-8 depression score,
are a consequence of depression. and that phone screen and internet connectivity behaviors were
Lastly, the data set was collected from Android participants the most insightful behaviors that influence depression in
only, and the long-term use of the Carat app could influence participants. The findings in this study are supported by previous
participants’ behavior [38,39]. Research has shown that research findings and contribute to compelling evidence on the
participant’s behavior and sociodemographics may differ utility of digital biomarkers in augmenting traditional assessment
between Android platforms and other mobile platforms such as of depression, thus enabling continuous and passive monitoring
iOS [44,95]. Future research could replicate this study to explore of the complex vectors of depression.
the extent of the differences in the participants’ behaviors (ie,
Acknowledgments
This research is supported by the Academy of Finland 6Genesis Flagship (Grant No. 318927), SENSATE (Grant Nos 316253,
320089), Infotech Institute University of Oulu Emerging Project, and Nokia Foundation (Jorma Ollila Grant for EP). We thank
the Carat Project for making their data set available for this study, and all participants who contributed to the Carat Project.
Authors' Contributions
KOA, DF, and EP contributed to the conceptualization of the study and bulk data transfers from Carat project archive servers.
EP and EL were principally involved in the Carat Project. KOA and JV contributed to data preparation, feature extraction, and
machine learning analysis. KOA and YT contributed to the statistical analysis. KOA and DF prepared the original draft. All
authors critically reviewed and edited the draft. All authors read and approved the final manuscript.
Conflicts of Interest
None declared.
Multimedia Appendix 1
Description of Features Extracted from the Smartphone Dataset.
[PDF File (Adobe PDF File), 36 KB-Multimedia Appendix 1]

XSL• FO
RenderX
Extended Results for Statistical Analysis.
Extended Results for Machine Learning Analysis.
References
1. Depression. World Health Organisation (2020). URL: https://www.who.int/news-room/fact-sheets/detail/depression [accessed
2021-06-19]
2. Lépine J, Briley M. The increasing burden of depression. Neuropsychiatr Dis Treat 2011 May;7(Suppl 1):3-7 [FREE Full
text] [doi: 10.2147/NDT.S19617] [Medline: 21750622]
3. Wang R, Wang W, daSilva A, Huckins JF, Kelley WM, Heatherton TF, et al. Tracking Depression Dynamics in College
Students Using Mobile Phone and Wearable Sensing. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol 2018 Mar
26;2(1):1-26 [FREE Full text] [doi: 10.1145/3191775]
4. Perlis ML, Giles DE, Buysse DJ, Tu X, Kupfer DJ. Self-reported sleep disturbance as a prodromal symptom in recurrent
depression. J Affect Disord 1997 Feb;42(2-3):209-212. [doi: 10.1016/s0165-0327(96)01411-5] [Medline: 9105962]
5. Hamilton J, Ladouceur C, Silk J, Franzen P, Bylsma L. Higher Rates of Sleep Disturbance Among Offspring of Parents
With Recurrent Depression Compared to Offspring of Nondepressed Parents. J Pediatr Psychol 2020 Jan 01;45(1):1-11
[FREE Full text] [doi: 10.1093/jpepsy/jsz079] [Medline: 31633781]
6. Saeb S, Zhang M, Kwasny M, Karr C, Kording K, Mohr D. The relationship between clinical, momentary, and sensor-based
assessment of depression. 2015 Presented at: 9th Int Conf Pervasive Comput Technol Healthc PervasiveHealth IEEE; .?232;
2015; Istanbul p. 229. [doi: 10.4108/icst.pervasivehealth.2015.259034]
7. Pressler SJ, Subramanian U, Perkins SM, Gradus-Pizlo I, Kareken D, Kim J, et al. Measuring depressive symptoms in heart
failure: validity and reliability of the patient health questionnaire-8. Am J Crit Care 2011 Mar;20(2):146-152 [FREE Full
text] [doi: 10.4037/ajcc2010931] [Medline: 20378777]
8. Burns MN, Begale M, Duffecy J, Gergle D, Karr CJ, Giangrande E, et al. Harnessing context sensing to develop a mobile
intervention for depression. J Med Internet Res 2011;13(3):e55 [FREE Full text] [doi: 10.2196/jmir.1838] [Medline:
21840837]
9. Hemmerle A, Herman J, Seroogy K. Stress, depression and Parkinson's disease. Exp Neurol 2012 Jan;233(1):79-86 [FREE
Full text] [doi: 10.1016/j.expneurol.2011.09.035] [Medline: 22001159]
10. Macchia A, Monte S, Pellegrini F, Romero M, D'Ettorre A, Tavazzi L, et al. Depression worsens outcomes in elderly
patients with heart failure: an analysis of 48,117 patients in a community setting. Eur J Heart Fail 2008 Jul 13;10(7):714-721
[FREE Full text] [doi: 10.1016/j.ejheart.2008.05.011] [Medline: 18565789]
11. Depression costs employers $210 billion Internet. Behavioral Health Executive. 2015. URL: https://www.psychcongress.com/
news-item/depression-costs-employers-210-billion [accessed 2021-06-28]
12. Fried EI, Nesse RM. Depression is not a consistent syndrome: An investigation of unique symptom patterns in the STAR*D
study. J Affect Disord 2015 Mar 01;172:96-102 [FREE Full text] [doi: 10.1016/j.jad.2014.10.010] [Medline: 25451401]
13. Fried EI, Nesse RM. Depression sum-scores don't add up: why analyzing specific depression symptoms is essential. BMC
Med 2015 Apr 06;13(1):72 [FREE Full text] [doi: 10.1186/s12916-015-0325-4] [Medline: 25879936]
14. Opoku Asare K, Visuri A, Ferreira D. Towards Early Detection of Depression Through Smartphone Sensing. In: Adjunct
Proceedings of the 2019 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings
of the 2019 ACM International Symposium on Wearable Computers. New York, NY: ACM; 2019 Presented at: ACM
International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2019 ACM International
Symposium on Wearable Computers; September 09, 2019; London p. 1158-1161. [doi: 10.1145/3341162.3347075]
15. van Berkel N, Ferreira D, Kostakos V. The Experience Sampling Method on Mobile Devices. ACM Comput. Surv 2018
Jan 12;50(6):1-40. [doi: 10.1145/3123988]
16. Paulhus D. Socially Desirable Responding on Self-Reports. In: Encyclopedia of Personality and Individual Differences.
Cham, Switzerland: Springer International Publishing; 2017:1-5.
17. Gorzelitz J, Peppard PE, Malecki K, Gennuso K, Nieto FJ, Cadmus-Bertram L. Predictors of discordance in self-report
versus device-measured physical activity measurement. Ann Epidemiol 2018 Jul;28(7):427-431 [FREE Full text] [doi:
10.1016/j.annepidem.2018.03.016] [Medline: 29681429]
18. Spathis D, Servia-Rodriguez S, Farrahi K, Mascolo C, Rentfrow J. Passive mobile sensing and psychological traits for large
scale mood prediction. In: Proceedings of the 13th EAI International Conference on Pervasive Computing Technologies
for Healthcare. New York, NY: ACM; 2019 Presented at: 13th EAI International Conference on Pervasive Computing
Technologies for Healthcare; May 20, 2019; Trento, Italy p. 272-281. [doi: 10.1145/3329189.3329213]

XSL• FO
RenderX
19. Wagner DT, Rice A, Beresford AR. Device analyzer. SIGMETRICS Perform. Eval. Rev 2014 Apr 17;41(4):53-56. [doi:
10.1145/2627534.2627553]
20. Ferreira D, Kostakos V, Dey AK. AWARE: Mobile Context Instrumentation Framework. Front. ICT 2015 Apr 20;2:6.
[doi: 10.3389/fict.2015.00006]
21. Oliner A, Iyer A, Stoica I, Lagerspetz E, Tarkoma S. Carat: Collaborative energy diagnosis for mobile devices. In: Proceedings
of the 11th ACM Conference on Embedded Networked Sensor Systems. 2013 Nov 11 Presented at: 11th ACM Conference
on Embedded Networked Sensor Systems; November 11-15, 2013; Roma, Italy p. 1-14. [doi: 10.1145/2517351.2517354]
22. Wang R, Chen F, Chen Z, Li T, Harari G, Tignor S, et al. StudentLife: assessing mental health, academic performance and
behavioral trends of college students using smartphones. In: Proceedings of the 2014 ACM International Joint Conference
on Pervasive and Ubiquitous Computing. New York, NY: Association for Computing Machinery; 2014 Presented at: ACM
International Joint Conference on Pervasive and Ubiquitous Computing; September 13, 2014; Seattle, WA p. 3-14. [doi:
10.1145/2632048.2632054]
23. Coravos A, Khozin S, Mandl KD. Erratum: Author Correction: Developing and adopting safe and effective digital biomarkers
to improve patient outcomes. NPJ Digit Med 2019 May 10;2(1):40 [FREE Full text] [doi: 10.1038/s41746-019-0119-8]
[Medline: 31304386]
24. Torous J, Kiang MV, Lorme J, Onnela J. New Tools for New Research in Psychiatry: A Scalable and Customizable Platform
to Empower Data Driven Smartphone Research. JMIR Ment Health 2016 May 05;3(2):e16 [FREE Full text] [doi:
10.2196/mental.5165] [Medline: 27150677]
25. Rohani D, Tuxen N, Lopategui A, Faurholt-Jepsen M, Kessing L, Bardram J. Personalizing Mental Health: A Feasibility
Study of a Mobile Behavioral Activation Tool for Depressed Patients. 2019 Presented at: PervasiveHealth'19: The 13th
International Conference on Pervasive Computing Technologies for Healthcare; May 20-23, 2019; Trento, Italy p. 282-291.
[doi: 10.1145/3329189.3329214]
26. Rohani DA, Faurholt-Jepsen M, Kessing LV, Bardram JE. Correlations Between Objective Behavioral Features Collected
From Mobile and Wearable Devices and Depressive Mood Symptoms in Patients With Affective Disorders: Systematic
Review. JMIR Mhealth Uhealth 2018 Aug 13;6(8):e165 [FREE Full text] [doi: 10.2196/mhealth.9691] [Medline: 30104184]
27. Tseng VW, Sano A, Ben-Zeev D, Brian R, Campbell AT, Hauser M, et al. Using behavioral rhythms and multi-task learning
to predict fine-grained symptoms of schizophrenia. Sci Rep 2020 Sep 15;10(1):15100 [FREE Full text] [doi:
10.1038/s41598-020-71689-1] [Medline: 32934246]
28. Moshe I, Terhorst Y, Opoku Asare K, Sander LB, Ferreira D, Baumeister H, et al. Predicting Symptoms of Depression and
Anxiety Using Smartphone and Wearable Data. Front Psychiatry 2021 Jan 28;12:625247 [FREE Full text] [doi:
10.3389/fpsyt.2021.625247] [Medline: 33584388]
29. Dogan E, Sander C, Wagner X, Hegerl U, Kohls E. Smartphone-Based Monitoring of Objective and Subjective Data in
Affective Disorders: Where Are We and Where Are We Going? Systematic Review. J Med Internet Res 2017 Jul
24;19(7):e262 [FREE Full text] [doi: 10.2196/jmir.7006] [Medline: 28739561]
30. Mohr DC, Zhang M, Schueller SM. Personal Sensing: Understanding Mental Health Using Ubiquitous Sensors and Machine
Learning. Annu Rev Clin Psychol 2017 May 08;13:23-47. [doi: 10.1146/annurev-clinpsy-032816-044949] [Medline:
28375728]
31. Seppälä J, De Vita I, Jämsä T, Miettunen J, Isohanni M, Rubinstein K, M-RESIST Group, et al. Mobile Phone and Wearable
Sensor-Based mHealth Approaches for Psychiatric Disorders and Symptoms: Systematic Review. JMIR Ment Health 2019
Mar 20;6(2):e9819 [FREE Full text] [doi: 10.2196/mental.9819] [Medline: 30785404]
32. Kourtis LC, Regele OB, Wright JM, Jones GB. Digital biomarkers for Alzheimer's disease: the mobile/ wearable devices
opportunity. NPJ Digit Med 2019;2(1):1-9 [FREE Full text] [doi: 10.1038/s41746-019-0084-2] [Medline: 31119198]
33. Saeb S, Zhang M, Kwasny MM, Karr CJ, Kording K, Mohr DC. The Relationship between Clinical, Momentary, and
Sensor-based Assessment of Depression. Int Conf Pervasive Comput Technol Healthc 2015 Aug;17(7):e175 [FREE Full
text] [doi: 10.4108/icst.pervasivehealth.2015.259034] [Medline: 26640739]
34. Jacobson NC, Weingarden H, Wilhelm S. Digital biomarkers of mood disorders and symptom change. NPJ Digit Med 2019
Feb 1;2(1):3 [FREE Full text] [doi: 10.1038/s41746-019-0078-0] [Medline: 31304353]
35. Kim H, Lee S, Lee S, Hong S, Kang H, Kim N. Depression Prediction by Using Ecological Momentary Assessment,
Actiwatch Data, and Machine Learning: Observational Study on Older Adults Living Alone. JMIR Mhealth Uhealth 2019
Oct 16;7(10):e14149 [FREE Full text] [doi: 10.2196/14149] [Medline: 31621642]
36. Malhi GS, Hamilton A, Morris G, Mannie Z, Das P, Outhred T. The promise of digital mood tracking technologies: are
we heading on the right track? Evid Based Ment Health 2017 Nov;20(4):102-107. [doi: 10.1136/eb-2017-102757] [Medline:
28855245]
37. Ward NG, Doerr HO, Storrie MC. Skin conductance: a potentially sensitive test for depression. Psychiatry Res 1983
Dec;10(4):295-302. [doi: 10.1016/0165-1781(83)90076-8] [Medline: 6583718]
38. Peltonen E, Lagerspetz E, Hamberg J, Mehrotra A, Musolesi M, Nurmi P, et al. The hidden image of mobile apps: geographic,
demographic, and cultural factors in mobile usage. Proc 20th Int Conf Hum-Comput Interact Mob Devices Serv 2018 Sep
03:1-12. [doi: 10.1145/3229434.3229474]

XSL• FO
RenderX
39. Athukorala K, Lagerspetz E, von KM, Jylhä A, Oliner A, Tarkoma S, et al. How carat affects user behavior: implications
for mobile battery awareness applications. In: Proceedings of the SIGCHI Conference on Human Factors in Computing
Systems. New York, NY: Association for Computing Machinery; 2014 Presented at: CHI '14: CHI Conference on Human
Factors in Computing; 26 April, 2014-1 May, 2014; Toronto, ON p. 1029-1038. [doi: 10.1145/2556288.2557271]
40. Peltonen E, Sharmila P, Opoku Asare K, Visuri A, Lagerspetz E, Ferreira D. When phones get personal: Predicting Big
Five personality traits from application usage. Pervasive and Mobile Computing 2020 Nov;69:101269. [doi:
10.1016/j.pmcj.2020.101269]
41. Kroenke K, Strine TW, Spitzer RL, Williams JBW, Berry JT, Mokdad AH. The PHQ-8 as a measure of current depression
in the general population. J Affect Disord 2009 Apr;114(1-3):163-173. [doi: 10.1016/j.jad.2008.06.026] [Medline: 18752852]
42. Hausser J, Strimmer K. entropy: Estimation of Entropy, Mutual Information and Related Quantities. CRAN. 2021 Apr 25.
URL: https://CRAN.R-project.org/package=entropy [accessed 2021-05-28]
43. Stachl C, Au Q, Schoedel R, Gosling SD, Harari GM, Buschek D, et al. Predicting personality from patterns of behavior
collected with smartphones. Proc Natl Acad Sci U S A 2020 Jul 28;117(30):17680-17687 [FREE Full text] [doi:
10.1073/pnas.1920484117] [Medline: 32665436]
44. Wang W, Harari GM, Wang R, Müller SR, Mirjafari S, Masaba K, et al. Sensing Behavioral Change over Time. Proc.
ACM Interact. Mob. Wearable Ubiquitous Technol 2018 Sep 18;2(3):1-21. [doi: 10.1145/3264951]
45. Shannon C. A mathematical theory of communication. Bell System Technical Journal 1948;27(3):379-423. [doi:
10.1002/j.1538-7305.1948.tb01338.x]
46. Gaskin CJ, Happell B. Power of mental health nursing research: a statistical analysis of studies in the International Journal
of Mental Health Nursing. Int J Ment Health Nurs 2013 Mar 27;22(1):69-75. [doi: 10.1111/j.1447-0349.2012.00845.x]
[Medline: 22738423]
47. Cohen J. Statistical Power Analysis for the Behavioral Sciences (Second Edition). New York, NY: Academic Press; 1988.
48. Enders C. Applied Missing Data Analysis. New York, NY: The Guilford Press; 2010.
49. Larsen R. Missing Data Imputation versus Full Information Maximum Likelihood with Second-Level Dependencies.
Structural Equation Modeling: A Multidisciplinary Journal 2011 Oct 05;18(4):649-662. [doi: 10.1080/10705511.2011.607721]
50. Revelle W. psych: Procedures for Psychological, Psychometric, and Personality Research. 2020 Jun 18. URL: https://CRAN.
R-project.org/package=psych [accessed 2021-06-28]
51. Holm S. A Simple Sequentially Rejective Multiple Test Procedure. Scandinavian Journal of Statistics 1979;6(2):65-70.
52. Goldstein H. Hierarchical Data Modeling in the Social Sciences. Journal of Educational and Behavioral Statistics
1995;20(2):201. [doi: 10.2307/1165357]
53. Nezlek J. Multilevel modeling for psychologists. In: APA Handbook of Research Methods in Psychology. Washington,
D.C: American Psychological Association; 2012:219-241.
54. Kendzor DE, Shuval K, Gabriel KP, Businelle MS, Ma P, High RR, et al. Impact of a Mobile Phone Intervention to Reduce
Sedentary Behavior in a Community Sample of Adults: A Quasi-Experimental Evaluation. J Med Internet Res 2016 Jan
25;18(1):e19 [FREE Full text] [doi: 10.2196/jmir.5137] [Medline: 26810027]
55. Zulueta J, Piscitello A, Rasic M, Easter R, Babu P, Langenecker SA, et al. Predicting Mood Disturbance Severity with
Mobile Phone Keystroke Metadata: A BiAffect Digital Phenotyping Study. J Med Internet Res 2018 Jul 20;20(7):e241
[FREE Full text] [doi: 10.2196/jmir.9775] [Medline: 30030209]
56. Musca SC, Kamiejski R, Nugier A, Méot A, Er-Rafiy A, Brauer M. Data with hierarchical structure: impact of intraclass
correlation and sample size on type-I error. Front Psychol 2011;2:74 [FREE Full text] [doi: 10.3389/fpsyg.2011.00074]
[Medline: 21687445]
57. Buuren SV, Groothuis-Oudshoorn K. : Multivariate Imputation by Chained Equations in. J. Stat. Soft 2011;45(3):1-67.
[doi: 10.18637/jss.v045.i03]
58. Snijders T, Bosker R. An Introduction to Basic and Advanced Multilevel Modeling. London, UK: SAGE; 2011.
59. Rubin D. Multiple Imputation after 18+ Years. Journal of the American Statistical Association 1996 Jun;91(434):473-489
[FREE Full text] [doi: 10.1080/01621459.1996.10476908]
60. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing.
Journal of the Royal Statistical Society: Series B (Methodological) 2018 Dec 05;57(1):289-300. [doi:
10.1111/j.2517-6161.1995.tb02031.x]
61. Benjamini Y, Yekutieli D. The control of the false discovery rate in multiple testing under dependency. Ann. Statist 2001
Aug 1;29(4):1165-1188. [doi: 10.1214/aos/1013699998]
62. Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. In: Proceedings of The 22nd ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining. New York, NY: Association for Computing Machinery; 2016
Presented at: The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; 13-17 August,
2016; San Francisco, CA p. 785-794. [doi: 10.1145/2939672.2939785]
63. Varoquaux G, Buitinck L, Louppe G, Grisel O, Pedregosa F, Mueller A. Scikit-learn. GetMobile: Mobile Comp. and Comm
2015 Jun;19(1):29-33. [doi: 10.1145/2786984.2786995]
64. Géron A. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to
Build Intelligent Systems. Newton, MA: O'Reilly Media, Inc; 2019.

XSL• FO
RenderX
65. Sultana M, Al-Jefri M, Lee J. Using Machine Learning and Smartphone and Smartwatch Data to Detect Emotional States
and Transitions: Exploratory Study. JMIR Mhealth Uhealth 2020 Sep 29;8(9):e17818 [FREE Full text] [doi: 10.2196/17818]
[Medline: 32990638]
66. Birnbaum ML, Kulkarni P, Van Meter A, Chen V, Rizvi AF, Arenare E, et al. Utilizing Machine Learning on Internet
Search Activity to Support the Diagnostic Process and Relapse Detection in Young Individuals With Early Psychosis:
Feasibility Study. JMIR Ment Health 2020 Sep 01;7(9):e19348 [FREE Full text] [doi: 10.2196/19348] [Medline: 32870161]
67. Sano A, Taylor S, McHill AW, Phillips AJ, Barger LK, Klerman E, et al. 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 2018 Jun 08;20(6):e210 [FREE Full text] [doi: 10.2196/jmir.9410] [Medline:
29884610]
68. Wshah S, Skalka C, Price M. Predicting Posttraumatic Stress Disorder Risk: A Machine Learning Approach. JMIR Ment
Health 2019 Jul 22;6(7):e13946 [FREE Full text] [doi: 10.2196/13946] [Medline: 31333201]
69. Benoit J, Onyeaka H, Keshavan M, Torous J. Systematic Review of Digital Phenotyping and Machine Learning in Psychosis
Spectrum Illnesses. Harv Rev Psychiatry 2020;28(5):296-304. [doi: 10.1097/HRP.0000000000000268] [Medline: 32796192]
70. One-Hot Encoding - An Overview. URL: https://www.sciencedirect.com/topics/computer-science/one-hot-encoding
[accessed 2021-06-28]
71. Dallora AL, Kvist O, Berglund JS, Ruiz SD, Boldt M, Flodmark C, et al. Chronological Age Assessment in Young Individuals
Using Bone Age Assessment Staging and Nonradiological Aspects: Machine Learning Multifactorial Approach. JMIR
Med Inform 2020 Sep 21;8(9):e18846 [FREE Full text] [doi: 10.2196/18846] [Medline: 32955457]
72. Krstajic D, Buturovic LJ, Leahy DE, Thomas S. Cross-validation pitfalls when selecting and assessing regression and
classification models. J Cheminform 2014 Mar 29;6(1):10 [FREE Full text] [doi: 10.1186/1758-2946-6-10] [Medline:
24678909]
73. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer WP. SMOTE: Synthetic Minority Over-sampling Technique. jair 2002
Jun 01;16:321-357. [doi: 10.1613/jair.953]
74. Lemaître G, Nogueira F, Aridas C. Imbalanced-learn: a python toolbox to tackle the curse of imbalanced data sets in machine
learning. Journal of Machine Learning Research 2017;18(1):559-563. [doi: 10.5555/3122009.3122026]
75. Altmann A, Toloşi L, Sander O, Lengauer T. Permutation importance: a corrected feature importance measure. Bioinformatics
2010 May 15;26(10):1340-1347. [doi: 10.1093/bioinformatics/btq134] [Medline: 20385727]
76. Beware Default Random Forest Importances. URL: https://explained.ai/rf-importance/#6.2 [accessed 2021-06-28]
77. Köster J, Rahmann S. Snakemake--a scalable bioinformatics workflow engine. Bioinformatics 2012 Oct 01;28(19):2520-2522.
[doi: 10.1093/bioinformatics/bts480] [Medline: 22908215]
78. Vega J, Li M, Aguillera K, Goel N, Joshl E, Durica K, et al. Reproducible Analysis Pipeline for Data Streams (RAPIDS):
Open Source Software to Process Data Collected with Mobile Devices Internet. JMIR Preprints. 2021. URL: https://preprints.
jmir.org/preprint/23246 [accessed 2021-06-28]
79. Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med 2016
Jan;4(2):30 [FREE Full text] [doi: 10.3978/j.issn.2305-5839.2015.12.63] [Medline: 26889483]
80. Bates D, Maechler M, Bolker AB, cre, Walker S, Christensen R, et al. lme4: Linear Mixed-Effects Models using 'Eigen'
and S4. URL: https://CRAN.R-project.org/package=lme4 [accessed 2021-06-28]
81. Bates D, Mächler M, Bolker B, Walker S. Fitting Linear Mixed-Effects Models Using. J. Stat. Soft 2015;67(1):e1. [doi:
10.18637/jss.v067.i01]
82. XGBoost Documentation. URL: https://xgboost.readthedocs.io/en/latest/ [accessed 2021-06-28]
83. Gruenerbl A, Osmani V, Bahle G, Carrasco J, Oehler S, Mayora O, et al. Using smart phone mobility traces for the diagnosis
of depressive and manic episodes in bipolar patients. In: Proceedings of the 5th Augmented Human International Conference.
USA: Association for Computing Machinery; 2014 Presented at: 5th Augmented Human International Conference; 7 March,
2014- 8 March, 2014; Kobe, Japan. [doi: 10.1145/2582051.2582089]
84. Farhan A, Lu J, Bi J, Russell A, Wang B, Bamis A. Multi-view Bi-clustering to Identify Smartphone Sensing Features
Indicative of Depression. In: Proceedings of the First IEEE International Conference on Connected Health: Applications,
Systems and Engineering Technologies. New York, NY: IEEE; 2016 Presented at: First IEEE International Conference on
Connected Health: Applications, Systems and Engineering Technologies; June 27-29, 2016; Washington, DC. [doi:
10.1109/chase.2016.27]
85. Hovy D. Demographic factors improve classification performance. 2015 Jul Presented at: Proceedings of the 53rd Annual
Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language
Processing (V; July 2015; Beijing, China p. 752-762. [doi: 10.3115/v1/p15-1073]
86. Bonful HA, Anum A. Sociodemographic correlates of depressive symptoms: a cross-sectional analytic study among healthy
urban Ghanaian women. BMC Public Health 2019 Jan 10;19(1):50 [FREE Full text] [doi: 10.1186/s12889-018-6322-8]
[Medline: 30630448]
87. Gershon A, Thompson WK, Eidelman P, McGlinchey EL, Kaplan KA, Harvey AG. Restless pillow, ruffled mind: sleep
and affect coupling in interepisode bipolar disorder. J Abnorm Psychol 2012 Nov;121(4):863-873 [FREE Full text] [doi:
10.1037/a0028233] [Medline: 22845651]

XSL• FO
RenderX
88. Dionisio A, Menezes R, Mendes DA. Mutual information: a measure of dependency for nonlinear time series. Physica A:
Statistical Mechanics and its Applications 2004 Dec;344(1-2):326-329. [doi: 10.1016/j.physa.2004.06.144]
89. Jeong J, Gore JC, Peterson BS. Mutual information analysis of the EEG in patients with Alzheimer's disease. Clin
Neurophysiol 2001 May;112(5):827-835. [doi: 10.1016/s1388-2457(01)00513-2] [Medline: 11336898]
90. Rajula HSR, Verlato G, Manchia M, Antonucci N, Fanos V. Comparison of Conventional Statistical Methods with Machine
Learning in Medicine: Diagnosis, Drug Development, and Treatment. Medicina (Kaunas) 2020 Sep 08;56(9):455 [FREE
Full text] [doi: 10.3390/medicina56090455] [Medline: 32911665]
91. Obuchi M, Huckins JF, Wang W, daSilva A, Rogers C, Murphy E, et al. Predicting Brain Functional Connectivity Using
Mobile Sensing. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol 2020 Mar 18;4(1):1-22. [doi: 10.1145/3381001]
92. Connolly CG, Ho TC, Blom EH, LeWinn KZ, Sacchet MD, Tymofiyeva O, et al. Resting-state functional connectivity of
the amygdala and longitudinal changes in depression severity in adolescent depression. J Affect Disord 2017 Jan 01;207:86-94
[FREE Full text] [doi: 10.1016/j.jad.2016.09.026] [Medline: 27716542]
93. Alavi SS, Maracy MR, Jannatifard F, Eslami M. The effect of psychiatric symptoms on the internet addiction disorder in
Isfahan's University students. J Res Med Sci 2011 Jun;16(6):793-800 [FREE Full text] [Medline: 22091309]
94. Yang S, Fu S, Chen K, Hsieh P, Lin P. Relationships between depression, health-related behaviors, and internet addiction
in female junior college students. PLoS One 2019 Aug 9;14(8):e0220784 [FREE Full text] [doi:
10.1371/journal.pone.0220784] [Medline: 31398212]
95. Pryss R, Schobel J, Hoppenstedt B, Spiliopoulou M, Langguth B, Probst T, et al. Ecological Momentary Assessment based
Differences between Android and iOS Users of the TrackYourHearing mHealth Crowdsensing Platform. Annu Int Conf
IEEE Eng Med Biol Soc 2019 Jul;2019:3951-3955. [doi: 10.1109/EMBC.2019.8857854] [Medline: 31946737]
Abbreviations
AUC: area under the curve
DT: decision tree
ICC: intraclass correlation
KNN: K-nearest neighbor
LMM: linear mixed model
LR: logistic regression
ML: machine learning
PHQ-8: 8-Item Patient Health Questionnaire
RF: random forest
RWC: random weighted classifier
SVM: support vector machine
XGB: XGBoost
Edited by L Buis; submitted 16.12.20; peer-reviewed by A McLean, Y Mao; comments to author 06.02.21; revised version received
15.03.21; accepted 14.05.21; published 12.07.21
Please cite as:
Opoku Asare K, Terhorst Y, Vega J, Peltonen E, Lagerspetz E, Ferreira D
Predicting Depression From Smartphone Behavioral Markers Using Machine Learning Methods, Hyperparameter Optimization, and
Feature Importance Analysis: Exploratory Study
JMIR Mhealth Uhealth 2021;9(7):e26540
URL: https://mhealth.jmir.org/2021/7/e26540
doi: 10.2196/26540
PMID: 34255713
©Kennedy Opoku Asare, Yannik Terhorst, Julio Vega, Ella Peltonen, Eemil Lagerspetz, Denzil Ferreira. Originally published
in JMIR mHealth and uHealth (https://mhealth.jmir.org), 12.07.2021. This is an open-access article distributed under the terms
of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is
properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as
this copyright and license information must be included.

XSL• FO
RenderX

predicting depression from smartphone behavioral markers

Uploaded by

Copyright:

Available Formats

predicting depression from smartphone behavioral markers

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

predicting depression from smartphone behavioral markers

Uploaded by

Copyright:

Available Formats

JMIR MHEALTH AND UHEALTH Opoku Asare et al

Predicting Depression From Smartphone Behavioral Markers

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 1

(JMIR Mhealth Uhealth 2021;9(7):e26540) doi: 10.2196/26540

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 2

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 3

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 4

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 5

Othera 154 (24.5)

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 6

For the distribution of the PHQ-8 scores, 1143/1374 (83.19%)

Table 2. Distribution of participants’ contribution to the PHQ-8a responses (N=629).

Table 3. Distribution of participants’ smartphone data set days (N=629).

Regarding the association analysis, we found an ICC of 0.7584;

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 7

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 8

Likewise, for the KNN classifier in Figure 3, the top 5 most

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 9

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 10

biomarker quantified the frequency and distribution (ie,

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 12

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 13

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 14

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 15

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 16

https://mhealth.jmir.org/2021/7/e26540 JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 7 | e26540 | p. 17

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.