Digital Phenotyping in Health Using Machine Learning Approaches: Scoping Review.

BackgroundGiven the growing interest in mobile data collection due to the proliferation of mobile phone ownership and network coverage in low- and middle-income countries (LMICs), we synthesized the evidence comparing estimates of health outcomes from multiple modes of data collection. In particular, we reviewed studies that compared a mode of remote data collection with at least one other mode of data collection to identify mode effects and areas for further research.ObjectiveThe study systematically reviewed and summarized the findings from articles and reports that compare a mode of remote data collection to at least one other mode. The aim of this synthesis was to assess the reliability and accuracy of results.MethodsSeven online databases were systematically searched for primary and grey literature pertaining to remote data collection in LMICs. Remote data collection included interactive voice response (IVR), computer-assisted telephone interviews (CATI), short message service (SMS), self-administered questionnaires (SAQ), and Web surveys. Two authors of this study reviewed the abstracts to identify articles which met the primary inclusion criteria. These criteria required that the survey collected the data from the respondent via mobile phone or landline. Articles that met the primary screening criteria were read in full and were screened using secondary inclusion criteria. The four secondary inclusion criteria were that two or more modes of data collection were compared, at least one mode of data collection in the study was a mobile phone survey, the study had to be conducted in a LMIC, and finally, the study should include a health component.ResultsOf the 11,568 articles screened, 10 articles were included in this study. Seven distinct modes of remote data collection were identified: CATI, SMS (singular sitting and modular design), IVR, SAQ, and Web surveys (mobile phone and personal computer). CATI was the most frequent remote mode (n=5 articles). Of the three in-person modes (face-to-face [FTF], in-person SAQ, and in-person IVR), FTF was the most common (n=11) mode. The 10 articles made 25 mode comparisons, of which 12 comparisons were from a single article. Six of the 10 articles included sensitive questions.ConclusionsThis literature review summarizes the existing research about remote data collection in LMICs. Due to both heterogeneity of outcomes and the limited number of comparisons, this literature review is best positioned to present the current evidence and knowledge gaps rather than attempt to draw conclusions. In order to advance the field of remote data collection, studies that employ standardized sampling methodologies and study designs are necessary to evaluate the potential for differences by survey modality.

As our society rapidly employs new forms of communication, new modes of data collection are challenging the best practices developed over years of polling. Preelection polling must simultaneously evolve, as new modes have emerged in the past few decades, including computer-mediated communication, mobile texting, and the use of touch tone keypads to communicate information. A tension exists between traditional and novel means of interpersonal communication, and researchers are struggling to determine which traditional methods of data collection still have a place in the modern industry. This study examined three relatively new modes of preelection poll data collection, online, mobile, and IVR (interactive voice recognition) to determine what relationships exist, if any, between the mode of data collection and the composition of a sample across eight demographic variables: age, education, gender, political affiliation, race, region, 2016 Vote History, and 2020 Vote Intention. Twenty-six preelection polls were used in the study, with each poll ranging in collection dates between August 30 and October 31, 2020. The total combined sample size for this study is n = 19,886; 49% were IVR respondents ( n = 9,795), 25% was collected from online panels ( n = 5,039), and 25% was collected from short message service (SMS)-to-web respondents ( n = 5,052). A χ2 (chi-square) test for association was conducted using a significance level of p < .05 and a 95% confidence interval (CI) and found a significant difference between each mode of data collection across the eight aforementioned variables. A significant difference between political party affiliation/registration and mode of data collection was attributed to the educational attainment of individuals participating in each preelection polls based on the mode of data collection. This study suggests that underlying variables within the sample composition of different modes of data collection can have an impact on the accuracy of preelection polls.

IntroductionSmartphones, central to modern life, offer a cost-effective tool for gaining patient insights outside the consultation room. Through passive data collection (e.g., sensor data) and active questioning, smartphones enable ecological assessments of psychiatric symptoms and self-reported experiences. This “moment-by-moment quantification of the individual-level human phenotype in situ using data from personal digital devices” via digital phenotyping (DP) has garnered significant research attention, showing potential for early detection and intervention in mental health.ObjectivesWe explore recent DP developments in mental health, highlighting its potential to transform clinical practice while acknowledging challenges and risks.MethodsNarrative literature review resorting to PubMed and Google Scholar using keywords such as “digital phenotyping”, “digital phenotype”, “digital biomarker” and “mobile sensing”.ResultsDP studies, particularly in Mood Disorders and Schizophrenia, mostly rely machine learning for data analysis. Biomarkers from passive data (e.g., GPS, social connectivity, physical activity) correlate with self-reports and clinical measures of depression, anxiety, mania, and psychosis. Speech and text analysis through Natural Language Processing (NLP) offers new research avenues. DP promises early detection, relapse prevention, and treatment monitoring but faces challenges, including privacy concerns, and low user engagement - that could be solved by closing the loop by returning individual research results or a tailor-made intervention. Nevertheless, regulation and good practice standards are still lacking, posing the threat of diagnostic inaccuracy and undeniable iatrogenic risk.ConclusionsFor DP to fully realize its potential, integration with standard care and existing systems is essential. While risks exist, when comparing DP with other medical interventions currently under research, perils are minor. Mental health care urgently needs disruptive innovation to improve access and quality.Disclosure of InterestNone Declared

There is a well-documented gap between the need for mental health treatment and the availability of these resources. Ecologically valid data collection is becoming increasingly more feasible through the increased everyday use of smartphones, wearable devices, and social media. These modes of data collection allow access to data collected outside of a clinical site. These data can give clinicians new insight into a patient’s everyday life. Digital phenotyping—the quantification of an individual’s phenotype through the use of data taken in situ—relies on machine learning models to potentially detect, diagnose, and treat a variety of different mental health disorders. Digital phenotyping thus has the potential to increase both the scalability and ecological validity of mental health treatment. Although digital phenotyping research is still in early stages, preliminary research shows promise for these data to help address the gap between mental health need and mental health care.

Passive non-invasive sensing signals from wearable devices and smartphones are typically collected continuously without user input. This passive and continuous data collection makes these signals suitable for moment-by-moment monitoring of health-related outcomes, disease diagnosis, and prediction modeling. A growing number of studies have utilized machine learning (ML) approaches to predict and analyze health indicators and diseases using passive non-invasive signals collected via wearable devices and smartphones. This systematic review identified peer-reviewed journal articles utilizing ML approaches for digital phenotyping and measuring digital biomarkers to analyze, screen, identify, and/or predict health-related outcomes using passive non-invasive signals collected from wearable devices or smartphones. PubMed, PubMed with Mesh, Web of Science, Scopus, and IEEE Xplore were searched for peer-reviewed journal articles published up to June 2024, identifying 66 papers. We reviewed the study populations used for data collection, data acquisition details, signal types, data preparation steps, ML approaches used, digital phenotypes and digital biomarkers, and health outcomes and diseases predicted using these ML techniques. Our findings highlight the promising potential for objective tracking of health outcomes and diseases using passive non-invasive signals collected from wearable devices and smartphones with ML approaches for characterization and prediction of a range of health outcomes and diseases, such as stress, seizure, fatigue, depression, and Parkinson’s disease. Future studies should focus on improving the quality of collected data, addressing missing data challenges, providing better documentation on study participants, and sharing the source code of the implemented methods and algorithms, along with their datasets and methods, for reproducibility purposes.

Qualitative research is known as an exploratory research, in which the common modes of data collection are interviews, focus group discussions and observational methods. Though there is a large amount of data available on how to conduct qualitative research through the above-mentioned three modes, little do we know about photographic documentation as a rather new and under researched mode of data collection in qualitative research. This paper attempts to explore photographic documentation as a primary mode of data collection in linguistic landscape. Linguistic landscape is a very popular and thriving area of research in applied linguistics. It is the visibility and salience of languages in public spaces, most commonly conducted on public and private signboards. The usual norm of conducting research in linguistic landscape is to click pictures of signboards and conduct an analysis of the language used on those signboards. This paper discusses the pilot testing of this mode of data collection, i.e. photographic documentation. After clicking pictures of public signboards from two public parks in Islamabad, the capital of Pakistan (following a very systematic approach), mediated discourse analysis was applied on it. Mediated discourse analysis is a specialized form of discourse analysis, having its roots in critical discourse analysis. Mediated discourse analysis observes the relationship between discourse and action. According to Scollon & Scollon (2003), we can only fully interpret the meaning of public texts by considering the social and physical world that surrounds them. The photographic data collected was analyzed at three levels; the text, the physicality of signboards and the social world in which they are displayed. This study suggests ways in improving this mode and procedure of data collection as to get notable findings in a qualitative research on linguistic landscape.

Digital phenotyping—the moment-by-moment quantification of human phenotypes in situ using data related to activity, behavior, and communications, from personal digital devices, such as smart phones and wearables—has been gaining interest. Personalized health information captured within free-living settings using such technologies may better enable the application of patient-generated health data (PGHD) to provide patient-centered care. The primary objective of this scoping review is to characterize the application of digital phenotyping and digitally captured active and passive PGHD for outcome measurement in surgical care. Secondarily, we synthesize the body of evidence to define specific areas for further work. We performed a systematic search of four bibliographic databases using terms related to “digital phenotyping and PGHD,” “outcome measurement,” and “surgical care” with no date limits. We registered the study (Open Science Framework), followed strict inclusion/exclusion criteria, performed screening, extraction, and synthesis of results in line with the PRISMA Extension for Scoping Reviews. A total of 224 studies were included. Published studies have accelerated in the last 5 years, originating in 29 countries (mostly from the USA, n = 74, 33%), featuring original prospective work (n = 149, 66%). Studies spanned 14 specialties, most commonly orthopedic surgery (n = 129, 58%), and had a postoperative focus (n = 210, 94%). Most of the work involved research-grade wearables (n = 130, 58%), prioritizing the capture of activity (n = 165, 74%) and biometric data (n = 100, 45%), with a view to providing a tracking/monitoring function (n = 115, 51%) for the management of surgical patients. Opportunities exist for further work across surgical specialties involving smartphones, communications data, comparison with patient-reported outcome measures (PROMs), applications focusing on prediction of outcomes, monitoring, risk profiling, shared decision making, and surgical optimization. The rapidly evolving state of the art in digital phenotyping and capture of PGHD offers exciting prospects for outcome measurement in surgical care pending further work and consideration related to clinical care, technology, and implementation.

BackgroundThe digital health care community has been urged to enhance engagement and clinical outcomes by analyzing multidimensional digital phenotypes.ObjectiveThis study aims to use a machine learning approach to investigate the performance of multivariate phenotypes in predicting the engagement rate and health outcomes of digital cognitive behavioral therapy.MethodsWe leveraged both conventional phenotypes assessed by validated psychological questionnaires and multidimensional digital phenotypes within time-series data from a mobile app of 45 participants undergoing digital cognitive behavioral therapy for 8 weeks. We conducted a machine learning analysis to discriminate the important characteristics.ResultsA higher engagement rate was associated with higher weight loss at 8 weeks (r=−0.59; P<.001) and 24 weeks (r=−0.52; P=.001). Applying the machine learning approach, lower self-esteem on the conventional phenotype and higher in-app motivational measures on digital phenotypes commonly accounted for both engagement and health outcomes. In addition, 16 types of digital phenotypes (ie, lower intake of high-calorie food and evening snacks and higher interaction frequency with mentors) predicted engagement rates (mean R2 0.416, SD 0.006). The prediction of short-term weight change (mean R2 0.382, SD 0.015) was associated with 13 different digital phenotypes (ie, lower intake of high-calorie food and carbohydrate and higher intake of low-calorie food). Finally, 8 measures of digital phenotypes (ie, lower intake of carbohydrate and evening snacks and higher motivation) were associated with a long-term weight change (mean R2 0.590, SD 0.011).ConclusionsOur findings successfully demonstrated how multiple psychological constructs, such as emotional, cognitive, behavioral, and motivational phenotypes, elucidate the mechanisms and clinical efficacy of a digital intervention using the machine learning method. Accordingly, our study designed an interpretable digital phenotype model, including multiple aspects of motivation before and during the intervention, predicting both engagement and clinical efficacy. This line of research may shed light on the development of advanced prevention and personalized digital therapeutics.Trial RegistrationClinicalTrials.gov NCT03465306; https://clinicaltrials.gov/ct2/show/NCT03465306

The aim of this systematic review and meta-analysis is to assess the prevalence of loneliness in many countries worldwide which have different ways of assessing it. Systematic review and meta-analysis. We searched seven electronic databases for English peer-reviewed studies published between 1992 and 2021. We selected English-language peer-reviewed articles, with data from non-clinical populations of community-dwelling older adults (>60 years), and with "loneliness" or "lonely" in the title. A multilevel random-effects meta-analysis was used to estimate the prevalence of loneliness across studies and to pool prevalence rates for different measurement instruments, data collection methods, and countries. Our initial search identified 2,021 studies of which 45 (k = 101 prevalence rates) were included in the final meta-analysis. The estimated pooled prevalence rate was 31.6% (n = 168,473). Measurement instrument was a statistically significant moderator of the overall prevalence of loneliness. Loneliness prevalence was lowest for single-item questions and highest for the 20-item University of California-Los Angeles Loneliness Scale. Also, differences between modes of data collection were significant: the loneliness prevalence was significantly the highest for face-to-face data collection and the lowest for telephone and CATI data collection. Our moderator analysis to look at the country effect indicated that four of the six dimensions of Hofstede also caused a significant increase (Power Distance Index, Uncertainty Avoidance Index, Indulgence) or decrease (Individualism) in loneliness prevalence. This study suggests that there is high variability in loneliness prevalence rates among community-dwelling older adults, influenced by measurement instrument used, mode of data collection, and country.

Sleep disorders, a critical issue in global health, affect millions worldwide.Disorders ranging from insomnia to sleep apnea profoundly impact individual well-being and societal productivity [1].While traditional diagnostic and therapeutic methods like polysomnography and cognitive-behavioral therapy for insomnia are effective, they are also labor-intensive, less patient-centered, and expensive.The emergence of digital phenotyping, using data from personal digital devices such as smartphones and wearables, heralds a promising new direction in sleep medicine [2].Digital phenotyping offers several advantages over traditional methods.It allows continuous, active, and passive data collection in a patient's natural environment, capturing a nuanced and comprehensive image of daily sleep patterns.These insights illuminate the interplay between sleep, lifestyle, behavior, health, and overall well-being [2].Digital phenotyping is also cost-effective, negating the need for expensive equipment or hospitalization, facilitating early identification of high-risk individuals for testing, and reducing unnecessary healthcare expenditure.Recent studies have validated the use of digital phenotyping in sleep medicine, revealing that sleep patterns derived from smartphones or wearable devices closely correlate with actigraphy, a noninvasive method for monitoring rest/activity cycles [3,4].Techniques introduced to measure aspects such as sleep stages and sleep apnea events using only smartphone data demonstrate that digital phenotyping may facilitate screening for sleep disorders [5].Additionally, conditions like mood disorders, closely linked to sleep-wake rhythms, can be assessed or predicted based on digital

Digital health interventions incorporating self-management strategies are increasingly used to support individuals in managing depression. These interventions often leverage self-monitoring and passive sensor-based data collection to provide personalized feedback, guiding behavioral change. With the proliferation of smartphones and wearable devices, there is growing potential to continuously collect behavioral and physiological data. However, a major limitation in the field is the lack of consolidated evidence on which specific parameters are most useful for monitoring and predicting depression-related outcomes. This scoping review aims to identify and synthesize common digital phenotyping and self-report parameters for monitoring and predicting depression. Specifically, it addresses the methodological and knowledge gap concerning which types of sensor-based and self-reported data are most frequently used and which demonstrate predictive value in tracking changes in depressive symptoms across digital platforms. A literature search was conducted across 4 databases, including PubMed, Embase, Cochrane Library, and the Web of Science Core Collection. Articles published between January 1, 2021, and November 26, 2025, were included. Eligible studies included adults (≥18 years) with depression confirmed through validated clinical measures and using digital approaches that collected passive sensor data, self-reports, or both. Studies focusing on comorbid disorders, nondigital interventions, or not reporting depression-related outcomes were excluded. The PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines were followed, and a 5-stage methodological framework for scoping reviews was implemented. Quality assessment was performed using the Downs and Black Instrument and the Mixed Methods Appraisal Tool (MMAT). Nineteen studies were included, comprising a total of 85,193 participants. Most studies used smartphone- or wearable-based tools, with passive sensing as the predominant data source and Patient Health Questionnaire-9 (PHQ-9) as the most commonly used depression measure. Five overarching parameter categories were identified: (1) physical activity and location, (2) behavioral patterns, (3) physiological signals, (4) sleep indicators, and (5) sociability and self-reported assessments. Within these categories, 11 metrics, including step count, heart rate variability, sleep duration and mood self-ratings, were most frequently reported. Most studies used a multimodal digital phenotyping approach, integrating passive sensor-derived data with active user-reported input, enabling more individualized symptom monitoring over time. This scoping review provides a novel synthesis of common digital parameters used across diverse tools for monitoring and predicting depression, moving beyond tool- or modality-specific perspectives adopted in prior reviews. Unlike existing reviews focusing on individual sensing modalities, prediction methods, or intervention effectiveness, this review maps shared parameters across observational, predictive, and interventional studies. By identifying convergent digital markers, the review supports comparability across studies and informs future model development. These findings have practical implications for the design of scalable digital mental health tools and for translating digital phenotyping into real-world clinical and self-management contexts.

State-of-the-art digital phenotyping methods for cardiometabolic risk prevention and management: a scoping review.

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.

Computational psychiatry has the potential to advance the diagnosis, mechanistic understanding, and treatment of mental health conditions. Promising results from clinical samples have led to calls to extend these methods to mental health risk assessment in the general public; however, data typically used with clinical samples are neither available nor scalable for research in the general population. Digital phenotyping addresses this by capitalizing on the multimodal and widely available data created by sensors embedded in personal digital devices (eg, smartphones) and is a promising approach to extending computational psychiatry methods to improve mental health risk assessment in the general population. Building on recommendations from existing computational psychiatry and digital phenotyping work, we aim to create the first computational psychiatry data set that is tailored to studying mental health risk in the general population; includes multimodal, sensor-based behavioral features; and is designed to be widely shared across academia, industry, and government using gold standard methods for privacy, confidentiality, and data integrity. We are using a stratified, random sampling design with 2 crossed factors (difficulties with emotion regulation and perceived life stress) to recruit a sample of 400 community-dwelling adults balanced across high- and low-risk for episodic mental health conditions. Participants first complete self-report questionnaires assessing current and lifetime psychiatric and medical diagnoses and treatment, and current psychosocial functioning. Participants then complete a 7-day in situ data collection phase that includes providing daily audio recordings, passive sensor data collected from smartphones, self-reports of daily mood and significant events, and a verbal description of the significant daily events during a nightly phone call. Participants complete the same baseline questionnaires 6 and 12 months after this phase. Self-report questionnaires will be scored using standard methods. Raw audio and passive sensor data will be processed to create a suite of daily summary features (eg, time spent at home). Data collection began in June 2022 and is expected to conclude by July 2024. To date, 310 participants have consented to the study; 149 have completed the baseline questionnaire and 7-day intensive data collection phase; and 61 and 31 have completed the 6- and 12-month follow-up questionnaires, respectively. Once completed, the proposed data set will be made available to academic researchers, industry, and the government using a stepped approach to maximize data privacy. This data set is designed as a complementary approach to current computational psychiatry and digital phenotyping research, with the goal of advancing mental health risk assessment within the general population. This data set aims to support the field's move away from siloed research laboratories collecting proprietary data and toward interdisciplinary collaborations that incorporate clinical, technical, and quantitative expertise at all stages of the research process. DERR1-10.2196/53857.

Digital phenotyping is a promising methodology for capturing moment-to-moment data that can inform individually adapted and timely interventions for youths with chronic pain. This study aimed to investigate adolescent and parent endorsement, perceived utility, and concerns related to passive data stream collection through smartphones for digital phenotyping for clinical and research purposes in youths with chronic pain. Through multiple-choice and open-response survey questions, we assessed the perspectives of patient-parent dyads (103 adolescents receiving treatment for chronic pain at a pediatric hospital with an average age of 15.6, SD 1.6 years, and 99 parents with an average age of 47.8, SD 6.3 years) on passive data collection from the following 9 smartphone-embedded passive data streams: accelerometer, apps, Bluetooth, SMS text message and call logs, keyboard, microphone, light, screen, and GPS. Quantitative and qualitative analyses indicated that adolescents and parent endorsement and perceived utility of digital phenotyping varied by stream, though participants generally endorsed the use of data collected by passive stream (35%-75.7% adolescent endorsement for clinical use and 37.9%-74.8% for research purposes; 53.5%-81.8% parent endorsement for clinical and 52.5%-82.8% for research purposes) if a certain level of utility could be provided. For adolescents and parents, adjusted logistic regression results indicated that the perceived utility of each stream significantly predicted the likelihood of endorsement of its use in both clinical practice and research (Ps<.05). Adolescents and parents alike identified accelerometer, light, screen, and GPS as the passive data streams with the highest utility (36.9%-47.5% identifying streams as useful). Similarly, adolescents and parents alike identified apps, Bluetooth, SMS text message and call logs, keyboard, and microphone as the passive data streams with the least utility (18.5%-34.3% identifying streams as useful). All participants reported primary concerns related to privacy, accuracy, and validity of the collected data. Passive data streams with the greatest number of total concerns were apps, Bluetooth, call and SMS text message logs, keyboard, and microphone. Findings support the tailored use of digital phenotyping for this population and can help refine this methodology toward an acceptable, feasible, and ethical implementation of real-time symptom monitoring for assessment and intervention in youths with chronic pain.