Automatic Depression Prediction Using Internet Traffic Characteristics on Smartphones

Messenger Use and Video Calls as Correlates of Depressive and Anxiety Symptoms: Results From the Corona Health App Study of German Adults During the COVID-19 Pandemic

BACKGROUND

Specialized studies have shown that smartphone-based social interaction data are predictors of depressive and anxiety symptoms. Moreover, at times during the COVID-19 pandemic, social interaction took place primarily remotely. To appropriately test these objective data for their added value for epidemiological research during the pandemic, it is necessary to include established predictors.

OBJECTIVE

Using a comprehensive model, we investigated the extent to which smartphone-based social interaction data contribute to the prediction of depressive and anxiety symptoms, while also taking into account well-established predictors and relevant pandemic-specific factors.

METHODS

We developed the Corona Health App and obtained participation from 490 Android smartphone users who agreed to allow us to collect smartphone-based social interaction data between July 2020 and February 2021. Using a cross-sectional design, we automatically collected data concerning average app use in terms of the categories video calls and telephony, messenger use, social media use, and SMS text messaging use, as well as pandemic-specific predictors and sociodemographic covariates. We statistically predicted depressive and anxiety symptoms using elastic net regression. To exclude overfitting, we used 10-fold cross-validation.

RESULTS

The amount of variance explained (R2) was 0.61 for the prediction of depressive symptoms and 0.57 for the prediction of anxiety symptoms. Of the smartphone-based social interaction data included, only messenger use proved to be a significant negative predictor of depressive and anxiety symptoms. Video calls were negative predictors only for depressive symptoms, and SMS text messaging use was a negative predictor only for anxiety symptoms.

CONCLUSIONS

The results show the relevance of smartphone-based social interaction data in predicting depressive and anxiety symptoms. However, even taken together in the context of a comprehensive model with well-established predictors, the data only add a small amount of value.

Investigating Rhythmicity in App Usage to Predict Depressive Symptoms: Protocol for Personalized Framework Development and Validation Through a Countrywide Study

BACKGROUND

Understanding a student's depressive symptoms could facilitate significantly more precise diagnosis and treatment. However, few studies have focused on depressive symptom prediction through unobtrusive systems, and these studies are limited by small sample sizes, low performance, and the requirement for higher resources. In addition, research has not explored whether statistically significant rhythms based on different app usage behavioral markers (eg, app usage sessions) exist that could be useful in finding subtle differences to predict with higher accuracy like the models based on rhythms of physiological data.

OBJECTIVE

The main objective of this study is to explore whether there exist statistically significant rhythms in resource-insensitive app usage behavioral markers and predict depressive symptoms through these marker-based rhythmic features. Another objective of this study is to understand whether there is a potential link between rhythmic features and depressive symptoms.

METHODS

Through a countrywide study, we collected 2952 students' raw app usage behavioral data and responses to the 9 depressive symptoms in the 9-item Patient Health Questionnaire (PHQ-9). The behavioral data were retrieved through our developed app, which was previously used in our pilot studies in Bangladesh on different research problems. To explore whether there is a rhythm based on app usage data, we will conduct a zero-amplitude test. In addition, we will develop a cosinor model for each participant to extract rhythmic parameters (eg, acrophase). In addition, to obtain a comprehensive picture of the rhythms, we will explore nonparametric rhythmic features (eg, interdaily stability). Furthermore, we will conduct regression analysis to understand the association of rhythmic features with depressive symptoms. Finally, we will develop a personalized multitask learning (MTL) framework to predict symptoms through rhythmic features.

RESULTS

After applying inclusion criteria (eg, having app usage data of at least 2 days to explore rhythmicity), we kept the data of 2902 (98.31%) students for analysis, with 24.48 million app usage events, and 7 days' app usage of 2849 (98.17%) students. The students are from all 8 divisions of Bangladesh, both public and private universities (19 different universities and 52 different departments). We are analyzing the data and will publish the findings in a peer-reviewed publication.

CONCLUSIONS

Having an in-depth understanding of app usage rhythms and their connection with depressive symptoms through a countrywide study can significantly help health care professionals and researchers better understand depressed students and may create possibilities for using app usage-based rhythms for intervention. In addition, the MTL framework based on app usage rhythmic features may more accurately predict depressive symptoms due to the rhythms' capability to find subtle differences.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/51540.

A Fast and Minimal System to Identify Depression Using Smartphones: Explainable Machine Learning-Based Approach

BACKGROUND

Existing robust, pervasive device-based systems developed in recent years to detect depression require data collected over a long period and may not be effective in cases where early detection is crucial. Additionally, due to the requirement of running systems in the background for prolonged periods, existing systems can be resource inefficient. As a result, these systems can be infeasible in low-resource settings.

OBJECTIVE

Our main objective was to develop a minimalistic system to identify depression using data retrieved in the fastest possible time. Another objective was to explain the machine learning (ML) models that were best for identifying depression.

METHODS

We developed a fast tool that retrieves the past 7 days' app usage data in 1 second (mean 0.31, SD 1.10 seconds). A total of 100 students from Bangladesh participated in our study, and our tool collected their app usage data and responses to the Patient Health Questionnaire-9. To identify depressed and nondepressed students, we developed a diverse set of ML models: linear, tree-based, and neural network-based models. We selected important features using the stable approach, along with 3 main types of feature selection (FS) approaches: filter, wrapper, and embedded methods. We developed and validated the models using the nested cross-validation method. Additionally, we explained the best ML models through the Shapley additive explanations (SHAP) method.

RESULTS

Leveraging only the app usage data retrieved in 1 second, our light gradient boosting machine model used the important features selected by the stable FS approach and correctly identified 82.4% (n=42) of depressed students (precision=75%, F1-score=78.5%). Moreover, after comprehensive exploration, we presented a parsimonious stacking model where around 5 features selected by the all-relevant FS approach Boruta were used in each iteration of validation and showed a maximum precision of 77.4% (balanced accuracy=77.9%). Feature importance analysis suggested app usage behavioral markers containing diurnal usage patterns as being more important than aggregated data-based markers. In addition, a SHAP analysis of our best models presented behavioral markers that were related to depression. For instance, students who were not depressed spent more time on education apps on weekdays, whereas those who were depressed used a higher number of photo and video apps and also had a higher deviation in using photo and video apps over the morning, afternoon, evening, and night time periods of the weekend.

CONCLUSIONS

Due to our system's fast and minimalistic nature, it may make a worthwhile contribution to identifying depression in underdeveloped and developing regions. In addition, our detailed discussion about the implication of our findings can facilitate the development of less resource-intensive systems to better understand students who are depressed and take steps for intervention.

Using Digital phenotyping to understand health-related outcomes: A scoping review

BACKGROUND

Digital phenotyping may detect changes in health outcomes and potentially lead to proactive measures to mitigate health declines and avoid major medical events. While health-related outcomes have traditionally been acquired through self-report measures, those approaches have numerous limitations, such as recall bias, and social desirability bias. Digital phenotyping may offer a potential solution to these limitations.

OBJECTIVES

The purpose of this scoping review was to identify and summarize how passive smartphone data are processed and evaluated analytically, including the relationship between these data and health-related outcomes.

METHODS

A search of PubMed, Scopus, Compendex, and HTA databases was conducted for all articles in April 2021 using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Review (PRISMA-ScR) guidelines.

RESULTS

A total of 40 articles were included and went through an analysis based on data collection approaches, feature extraction, data analytics, behavioral markers, and health-related outcomes. This review demonstrated a layer of features derived from raw sensor data that can then be integrated to estimate and predict behaviors, emotions, and health-related outcomes. Most studies collected data from a combination of sensors. GPS was the most used digital phenotyping data. Feature types included physical activity, location, mobility, social activity, sleep, and in-phone activity. Studies involved a broad range of the features used: data preprocessing, analysis approaches, analytic techniques, and algorithms tested. 55% of the studies (n = 22) focused on mental health-related outcomes.

CONCLUSION

This scoping review catalogued in detail the research to date regarding the approaches to using passive smartphone sensor data to derive behavioral markers to correlate with or predict health-related outcomes. Findings will serve as a central resource for researchers to survey the field of research designs and approaches performed to date and move this emerging domain of research forward towards ultimately providing clinical utility in patient care.

Opportunities for Smartphone Sensing in E-Health Research: A Narrative Review

Recent years have seen significant advances in the sensing capabilities of smartphones, enabling them to collect rich contextual information such as location, device usage, and human activity at a given point in time. Combined with widespread user adoption and the ability to gather user data remotely, smartphone-based sensing has become an appealing choice for health research. Numerous studies over the years have demonstrated the promise of using smartphone-based sensing to monitor a range of health conditions, particularly mental health conditions. However, as research is progressing to develop the predictive capabilities of smartphones, it becomes even more crucial to fully understand the capabilities and limitations of using this technology, given its potential impact on human health. To this end, this paper presents a narrative review of smartphone-sensing literature from the past 5 years, to highlight the opportunities and challenges of this approach in healthcare. It provides an overview of the type of health conditions studied, the types of data collected, tools used, and the challenges encountered in using smartphones for healthcare studies, which aims to serve as a guide for researchers wishing to embark on similar research in the future. Our findings highlight the predominance of mental health studies, discuss the opportunities of using standardized sensing approaches and machine-learning advancements, and present the trends of smartphone sensing in healthcare over the years.

Digital phenotyping in depression diagnostics: Integrating psychiatric and engineering perspectives

Depression is a serious medical condition and is a leading cause of disability worldwide. Current depression diagnostics and assessment has significant limitations due to heterogeneity of clinical presentations, lack of objective assessments, and assessments that rely on patients' perceptions, memory, and recall. Digital phenotyping (DP), especially assessments conducted using mobile health technologies, has the potential to greatly improve accuracy of depression diagnostics by generating objectively measurable endophenotypes. DP includes two primary sources of digital data generated using ecological momentary assessments (EMA), assessments conducted in real-time, in subjects' natural environment. This includes active EMA, data that require active input by the subject, and passive EMA or passive sensing, data passively and automatically collected from subjects' personal digital devices. The raw data is then analyzed using machine learning algorithms to identify behavioral patterns that correlate with patients' clinical status. Preliminary investigations have also shown that linguistic and behavioral clues from social media data and data extracted from the electronic medical records can be used to predict depression status. These other sources of data and recent advances in telepsychiatry can further enhance DP of the depressed patients. Success of DP endeavors depends on critical contributions from both psychiatric and engineering disciplines. The current review integrates important perspectives from both disciplines and discusses parameters for successful interdisciplinary collaborations. A clinically-relevant model for incorporating DP in clinical setting is presented. This model, based on investigations conducted by our group, delineates development of a depression prediction system and its integration in clinical setting to enhance depression diagnostics and inform the clinical decision making process. Benefits, challenges, and opportunities pertaining to clinical integration of DP of depression diagnostics are discussed from interdisciplinary perspectives.

Digital phenotype of mood disorders: A conceptual and critical review

BACKGROUND

Mood disorders are commonly diagnosed and staged using clinical features that rely merely on subjective data. The concept of digital phenotyping is based on the idea that collecting real-time markers of human behavior allows us to determine the digital signature of a pathology. This strategy assumes that behaviors are quantifiable from data extracted and analyzed through digital sensors, wearable devices, or smartphones. That concept could bring a shift in the diagnosis of mood disorders, introducing for the first time additional examinations on psychiatric routine care.

OBJECTIVE

The main objective of this review was to propose a conceptual and critical review of the literature regarding the theoretical and technical principles of the digital phenotypes applied to mood disorders.

METHODS

We conducted a review of the literature by updating a previous article and querying the PubMed database between February 2017 and November 2021 on titles with relevant keywords regarding digital phenotyping, mood disorders and artificial intelligence.

RESULTS

Out of 884 articles included for evaluation, 45 articles were taken into account and classified by data source (multimodal, actigraphy, ECG, smartphone use, voice analysis, or body temperature). For depressive episodes, the main finding is a decrease in terms of functional and biological parameters [decrease in activities and walking, decrease in the number of calls and SMS messages, decrease in temperature and heart rate variability (HRV)], while the manic phase produces the reverse phenomenon (increase in activities, number of calls and HRV).

CONCLUSION

The various studies presented support the potential interest in digital phenotyping to computerize the clinical characteristics of mood disorders.

Digital Phenotyping: an Epistemic and Methodological Analysis

Some claim that digital phenotyping will revolutionize understanding of human psychology and experience and significantly promote human wellbeing. This paper investigates the nature of digital phenotyping in relation to its alleged promise. Unlike most of the literature to date on philosophy and digital phenotyping, which has focused on its ethical aspects, this paper focuses on its epistemic and methodological aspects. The paper advances a tetra-taxonomy involving four scenario types in which knowledge may be acquired from human "digitypes" by digital phenotyping. These scenarios comprise two causal relations and a correlative and constitutive relation that can exist between information generated by digital systems/devices on the one hand and psychological or behavioral phenomena on the other. The paper describes several modes of inference involved in deriving knowledge within these scenarios. After this epistemic mapping, the paper analyzes the possible knowledge potential and limitations of digital phenotyping. It finds that digital phenotyping holds promise of delivering insight into conditions and states as well producing potentially new psychological categories. It also argues that care must be taken that digital phenotyping does not make unwarranted conclusions and is aware of potentially distorting effects in digital sensing and measurement. If digital phenotyping is to truly revolutionize knowledge of human life, it must deliver on a range of fronts, including making accurate forecasts and diagnoses of states and behaviors, providing causal explanations of these phenomena, and revealing important constituents of human conditions, psychology, and experience.