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 the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Electronic mental health interventions for mood disorders have increased rapidly over the past decade, most recently in the form of various systems and apps that are delivered via smartphones.
We aim to provide an overview of studies on smartphone-based systems that combine subjective ratings with objectively measured data for longitudinal monitoring of patients with affective disorders. Specifically, we aim to examine current knowledge on: (1) the feasibility of, and adherence to, such systems; (2) the association of monitored data with mood status; and (3) the effects of monitoring on clinical outcomes.
We systematically searched PubMed, Web of Science, PsycINFO, and the Cochrane Central Register of Controlled Trials for relevant articles published in the last ten years (2007-2017) by applying Boolean search operators with an iterative combination of search terms, which was conducted in February 2017. Additional articles were identified via pearling, author correspondence, selected reference lists, and trial protocols.
A total of 3463 unique records were identified. Twenty-nine studies met the inclusion criteria and were included in the review. The majority of articles represented feasibility studies (n=27); two articles reported results from one randomized controlled trial (RCT). In total, six different self-monitoring systems for affective disorders that used subjective mood ratings and objective measurements were included. These objective parameters included physiological data (heart rate variability), behavioral data (phone usage, physical activity, voice features), and context/environmental information (light exposure and location). The included articles contained results regarding feasibility of such systems in affective disorders, showed reasonable accuracy in predicting mood status and mood fluctuations based on the objectively monitored data, and reported observations about the impact of monitoring on clinical state and adherence of patients to the system usage.
The included observational studies and RCT substantiate the value of smartphone-based approaches for gathering long-term objective data (aside from self-ratings to monitor clinical symptoms) to predict changes in clinical states, and to investigate causal inferences about state changes in patients with affective disorders. Although promising, a much larger evidence-base is necessary to fully assess the potential and the risks of these approaches. Methodological limitations of the available studies (eg, small sample sizes, variations in the number of observations or monitoring duration, lack of RCT, and heterogeneity of methods) restrict the interpretability of the results. However, a number of study protocols stated ambitions to expand and intensify research in this emerging and promising field.
Unipolar depressive disorder is one of the leading chronic medical conditions in Europe, accounting for 13.7% of the disability burden [
Electronic mental health (e-mental health) interventions represent a promising means of overcoming such barriers and increasing the capacity for patients’ self-management of depression [
Capitalizing on these technological advances, a novel approach in e-mental health has emerged with the development and testing of several technology-enhanced systems for affective disorders that rely on combinations of subjective (ie, self-reported) and objective data, measured by sensors embedded within (or connected to) smartphones. When considering subjective self-ratings, smartphone apps result in better adherence and less missing data compared with paper-pencil or personal computer-based questionnaires or diaries [
Using objective data for symptom monitoring is especially relevant for affective disorders, in which reliable diagnostic and predictive biomarkers are still lacking and self-reported symptoms are not only influenced by recall bias, but also by a mood-state associated bias. Therefore, it is an important question whether long-term continuous monitoring of subjective and objective parameters combined with a time-series analysis could provide earlier and more reliable predictors and indicators of mood changes. Such findings could be helpful for the individual patient to improve their self-management and prevent relapses [
The aim of this systematic review is to examine the literature on smartphone-based systems that combine subjective ratings with objectively measured data for the self-monitoring of depressive symptoms. The following aspects will be covered: (1) feasibility and adherence, (2) the association of monitoring data with mood status, and (3) effects on clinical outcome.
Literature was systematically reviewed in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [
In February 2017, the search for relevant articles published in the last ten years (2007-2017) was conducted in the following bibliographic databases: PubMed, Web of Science, PsycINFO, and Cochrane Central Register of Controlled Trials. Additional articles were identified via pearling, author correspondence, selected reference lists, and trial protocols.
The search strategy was developed by three of the authors and included combinations of the search terms, such as
Titles and abstracts were screened for eligibility and were included if: (1) abstracts were published in English or German, (2) any form of smartphone-assessed self-monitoring of depressive symptoms was mentioned, and (3) adult study participants with depressive symptoms (but not postpartum/postnatal depression or pregnant women) were investigated. In cases of insufficient information to determine eligibility, papers were nonetheless subjected to further screening. Considering the novelty of this research area, the review was planned to be as inclusive as possible. Therefore, experimental and observational studies were included.
We obtained full reports of all potentially relevant papers. Articles were discarded if they met at least one of the following exclusion criteria: (1) review papers only; (2) study protocols or descriptive planned studies or future study intents; (3) descriptive studies that did not include participants diagnosed with affective disorders (apart from postpartum/postnatal depression); (4) studies that did not monitor objective data involving a smartphone/mobile device (eg, computerized/Internet-delivered or Internet-based Cognitive Behavioral Therapy interventions or Ecological Momentary Assessment [EMA] studies in which only subjective ratings were obtained); or (5) other reasons (eg, comments, letters, poster abstracts, articles not written in English). Additional information from study authors was obtained when necessary to resolve questions about eligibility. Four authors (ED, EK, CS, and XW) screened the full text reports and decided whether these papers met the inclusion criteria. Records were originally split among the four reviewers, each of whom made independent decisions regarding exclusion. Uncertainties and disagreement were resolved through discussion (ED, EK, CS, and UH). Reasons for excluding articles were recorded (including multiple assignments to exclusion criteria, if applicable). Finally, all exclusion assignments were reviewed and each record was allocated to one exclusion criteria (CS, XW, and EK).
The Downs and Black Instrument [
Due to the novelty of the approach/research field and the heterogeneous nature of the included studies, a quantitative data synthesis (meta-analysis) has not been performed, since there was an insufficient amount of data to calculate effect sizes. Standardized forms were developed for data collection purposes and two authors (ED and EK) were trained to use and apply them. Piloting of the extraction forms with a subsample of randomly selected included studies also took place. To ensure consistency between reviewers, calibration exercises took place before starting the review. Data extraction was performed in duplicate by two authors (ED and EK), disagreement was resolved by discussion, and another author (CS) verified the information to reduce bias and errors in data extraction. We contacted study authors to resolve any uncertainties or for clarification, if needed. A systematic narrative synthesis will be provided with information presented in the text and tables to summarize and explain the characteristics and findings of the included studies.
After the description of the selection process and characteristics of included systems and studies are detailed, the results section is structured according to the objectives of the review. The
The electronic database search resulted in a total of 4122 records (PubMed, n=650; PsycINFO, n=378; Web of Science, n=1534; Cochrane Central Register of Controlled Trial, n=1491; and n=69 identifications by cross-referencing), of which 3463 were unique articles. In the first selection step (abstract screening), 3173 records were excluded after it was determined that they did not meet inclusion criteria based on information in the abstract. The majority of these studies were excluded because they only contained descriptions of online/computerized Cognitive Behavioral Therapy systems, descriptions of pure online interventions, or self-monitoring or screening tools (using only subjective assessment and self-monitoring or smartphone-based ecological momentary assessments). A total of 290 full text papers were retrieved for further consideration. After reviewing these full texts, an additional 261 records were excluded.
PRISMA Flow chart.
Twenty-nine studies were included for qualitative synthesis in this review. Although the search range spanned from 2007 to 2017, the included studies were all published after 2010, attesting to the novelty of the research field.
A detailed description of the included studies is provided in
Overview of smartphone-based systems that combine subjective mood ratings and objective assessment of physiological and behavioral parameters for the monitoring of affective symptoms.
Self-Report Data | Physiological Data | Behavioral Data | Context and Environmental Data | Type of clinical trials | ||||
Phone Usage | Physical Activity | Voice features | Light Exposure | Location | ||||
Empath [ |
+ | + | - | + | + | - | - | Case Study |
FINE [ |
+ | - | + | + | - | - | + | Feasibility Study |
MedLink [ |
+ | - | - | - | - | - | - | Pilot Study |
Mobilyze! [ |
+ | - | + | + | - | + | + | Feasibility Study |
MONARCA [ |
+ | - | + | + | + | + | + | Feasibility Study, Pilot Study, RCT |
MoodRhythm [ |
+ | - | + | + | - | + | - | Pilot Study |
PSYCHE [ |
+ | + | - | + | + | - | - | Pilot Study |
SIMBA [ |
+ | - | + | + | - | - | + | Pilot Study |
Unnamed System [ |
+ | + | + | + | - | - | - | Feasibility Study |
Unnamed System [ |
+ | +c | - | + | - | - | - | Pilot Study |
aObjective monitoring of medication adherence using a
bSystem was utilized in an Automated Monitoring of Symptom Severity (AMoSS) study.
cAn ingestion sensor was used for gathering the physiological data.
According the Downs and Black instrument, the methodological quality scores of the studies ranged from 5 [
Among the included studies, large variations in drop-out rates were reported, ranging from 0% [
In a field trial of MONARCA 1.0 [
Beiwinkel et al [
The overall satisfaction of users with smartphone-based systems is reported to be rather positive (eg, the satisfaction with the Mobilyze! intervention was reported to be 5.71 [SD=1.38] on a scale ranging from 1 [strong disagreement] to 7 [strong agreement]) [
Patient compliance was reported as a limitation in several studies [
Reporting experiences from a first field trial of their FINE app prototype, Dang et al [
At present, most empirical data exist on the association between objectively recorded data and clinical state or symptom changes (self-reported or rated by clinicians). Therefore, the following section summarizes results obtained with the different (bio-) parameters. If applicable and relevant, additional evidence from nonincluded publications is given to provide a comprehensive overview. Based on our results, no cumulative statements about the direction and/or magnitude of effect sizes can be drawn, since all studies differed concerning selected features, statistical approaches, and handling of ground truth (classification of affective states) for mood state prediction. Overall, the prediction accuracy depended largely on the methodological approach and the respective affective states as outcome measures (eg, depression versus euthymic state compared to depression versus [hypo]mania).
Many of the included studies reported results on a small number of subjects. However, when individual time-series are to be analyzed, a sufficient recording duration for the individual patients seems to be more desirable than a larger number of subjects. Twenty-five studies monitored data for several months or weeks, three of which still had to exclude subjects from the final analysis, as they did not experience any changes in mood states, and thus associations between mood states and objective parameters could not be analyzed [
Self-assessment or subjective rating of mood status via different assessment methods is an integral part of nearly all included studies and systems currently being tested. The superiority of EMAs in mental health versus standard paper-pencil questionnaires has been proved by many studies [
The content of self-monitored items varies from system to system. Within the MONARCA system, self-assessment of mood, sleep length, medication adherence, irritability, activity, mixed mood, cognitive problems, stress, alcohol consumption, menstruation for women, and individualized early warning signs are requested [
Self-assessments may serve different purposes: they provide a continuous support to individuals (eg, supportive monitoring) and also allow for a better estimation of individuals’ need for support, and may serve as an underlying information system to guide and tailor care accordingly (eg, by detecting symptom deterioration or early warning signs for change in moods states) [
According to Valenza et al [
Using the PSYCHE platform, ECG/HRV and respiration activity were gathered using a sensorized t-shirt [
During a depressive phase, the desire and ability for social interaction is mostly reduced, whereas during the (hypo)manic phase it is increased [
According to Grunerbl et al [
Within the MONARCA project, significant positive correlations between the number and duration of incoming and outgoing calls per day and scores on YMRS, and between scores on the HDRS-17 and duration of incoming and outgoing calls per day, were found; additionally, a significant positive correlation between scores on the YMRS and the number of outgoing text messages per day was shown [
Beiwinkel et al [
Phone usage patterns including screen state and usage of apps might provide information about patients’ current mood state. Alvarez-Lorenzo et al [
Patients with affective disorder tend to have lower acceleration and activity energy expenditure compared to healthy individuals [
Eight of the selected studies reported on the relationship between affective states and physical activity [
Changes in travel patterns (eg, distance travelled per day, location changes), which can be gathered through GPS, Bluetooth (via detection of other wireless devices), cell tower identification (ID) and WiFi, may provide information about patients’ mood states. For example, it is expected that people in a depressive phase travel and stay outside less often [
Light density in the environment can be used to provide information about patients’ sleep-wake cycles, which is disturbed in unipolar depression as well as in bipolar disorder [
Analysis of speech and voice features appears to be a useful approach in predicting mood state in unipolar depression and in bipolar disorder [
Eight articles [
Analyses of voice and speech features are also performed within the frame of the MONARCA project. Muaremi et al [
Most of the reviewed papers described a combination of self-reported data with behavioral and/or physiological parameters. Grünerbl et al [
Most of the reviewed studies did not report any results regarding the effects of smartphone-based monitoring on depressive symptoms or clinical outcomes. However, Burns et al [
Saunders et al [
Results from only one RCT were available. Faurholt-Jepsen et al [
To our knowledge, this is the first systematic review of the current scientific literature on smartphone-based systems for the subjective and objective monitoring of mood disorders and studies testing these systems. A considerable number of projects have started developing smartphone-based tools for the management and monitoring of mood disorders and have conducted promising tests of feasibility. Nonetheless, it became obvious during the review process that only a small number of systems that are currently available (or in preparation) have been described in scientific papers, let alone been subjected to empirical studies.
The results showed great potential in monitoring parameters relevant to the course of affective disorders, although most of the reviewed studies were feasibility tests or pilot trials without control groups, conducted with only a small number of participants and with short assessment periods, and did not investigating any medium- or long-term effects or possible risks of the respective applications. It is important to highlight that there are several systems described in publications that could not be included in this review due to: not meeting the inclusion criteria (eg, no monitoring of objective data [
Given the lack of reliable biomarkers in affective disorders, the combination of different physiological and behavioral parameters seems to be a highly promising approach [
The acceptance of using smartphones for continuous monitoring (of subjective and objective data) seems encouraging based on the results of this review and reflects patients’ interests. However, regarding the decrease in adherence rates in several studies [
Technical issues that might occur in this context (ie, energy management) could serve as a major obstacle. Using external sensors that do not rely on the smartphone’s power source was discussed as one possible solution, but Prociow et al [
Only five of the reviewed articles reported on clinical outcomes [
Combining smartphone-based self-monitoring with digital intervention strategies has a high potential for improving the current treatment and management of affective disorders. By investigating the efficacy of smartphone-based interventions, the notion of the “digital placebo” effect [
Additionally, there are several ethical issues arising from continuous monitoring and the use of objective smartphone data. While some users/patients may find that remote monitoring allows them to feel more control over their illness, also opening new possibilities for patient-empowerment and self-management, others may feel overwhelmed by the added responsibility of self-monitoring, or perceive it as a constant reminder that they are ill [
Sensitive data concerning mental health require an especially high standard of protection from security breaches due to a high risk of stigmatization and discrimination in the case of data disclosure. Patients’ concerns about the security of the data monitored and recorded by the system have significant impacts on the acceptance of the apps and the readiness to use them for continuous monitoring [
Smartphone-based systems for managing and monitoring mood disorders present a highly promising field of innovation in health care. Simply using and carrying a smartphone on a day-to-day basis generates a larger amount of data than is typically collected in questionnaire-based studies or online interventions. Analyses of covariations of recorded bio-parameters within individual time-series and identification of individual patterns that predict state changes must be employed, in addition to group statistics. However, to identify such individual patterns, long monitoring durations with a certain amount of state changes are required. Most of the available studies used short or moderately sized recording durations and three of them [
Apart from the continuous assessment of bio-parameters themselves, smartphone-based monitoring also allows researchers to gather information on context and environment, which could prove valuable for the interpretation of the monitored biomedical data (eg, information about weather conditions) and allow for a better interpretation of changes in body temperature or skin conductance levels (as well as locomotor activity) that might be restricted on rainy versus sunny days. Furthermore, these large amounts of data allow for the employment of data mining techniques to potentially generate and explore new hypotheses about causes and predictors of affective episodes, define individual disorder phenotypes, or analyze the consequences of changes in medication or treatment regimens, or potentially even health policy changes [
Continuous measurement of physiological and behavioral parameters is a valuable addition to self-reported symptoms to help complete the clinical picture, especially since there is often significant variation between measured parameters and subjective perspective of the self-reporting patient [
Smartphone-based monitoring systems can also be combined with electronically delivered psychotherapeutic interventions. By linking behavioral monitoring and intervention feedback loops, interventions can be delivered right-on-time at the most useful moment, in contrast to face-to-face psychotherapy mainly based on retrospective data and information [
In this review, we have presented the current state of research on smartphone-based mental-health technologies for monitoring and managing mood disorders. A limitation of this publication is that it only covers those smartphone-based systems in which studies have been conducted, and results of these studies have been published in scientific literature. We also focused on articles listed in scientific literature databases; thus, grey literature (eg, conference proceedings and abstracts) was not included. There are a great number of apps for the monitoring and management of mood disorders available to the public that have either not been subjected to a scientific study, or the results have not been published (yet). Furthermore, some publications had to be excluded due to not reporting results on patients with affective disorders, although they were also introducing smartphone-based systems for affective symptoms. Another limitation is that a number of the included studies were feasibility studies, allowing conclusions to be drawn about the practicality of a certain monitoring technology, yet not about its clinical efficacy or effectiveness. However, the results of this systematic review on the current state of smartphone-based monitoring of depressive symptoms demonstrates that using smartphone apps for EMA and management of mood disorders is a promising field, in which more and larger efficacy studies and longitudinal research regarding placebo effects, possible risks, and adverse effects are required.
Smartphone-based monitoring of objective and subjective data in mood disorders is a rapidly growing approach and research field. The current body of literature consists mainly of observational studies and substantiates the value of smartphone-based approaches for gathering long-term objective data (aside from self-ratings) to monitor clinical symptoms, predict changes in clinical states, and investigate causal inferences about state changes in patients with affective disorders. Although promising, a much larger evidence-base is necessary to fully assess the potential, as well as the risks, of these approaches. Fortunately, a number of study protocols stated ambitions to expand and intensify research in this emerging and promising field in the coming years.
PRISMA checklist.
Search strategy.
Reference list of excluded full text articles with reason for exclusion.
Overview of selected studies.
Methodological Quality Scores.
Automated Monitoring of Symptom Severity
electrocardiogram
Ecological Momentary Assessment
electronic mental health
global positioning system
Hamilton Depression Rating Scale (17 item version)
heart rate variability
identification
Patient Health Questionnaire
Preferred Reporting Items for Systematic Review and Meta-Analysis
randomized controlled trial
standard deviation
short messaging system
Social Rhythm Metric
Young Mania Rating Scale
Preparation of this work was supported by the STEADY-project sponsored by the German Federal Ministry of Education and Research within the funding program
Authors UH, ED, and EK developed the concept of this paper and the search strategy. Authors ED, EK, and CS completed the data extraction of relevant articles. Authors ED, XW, CS, and EK wrote the first draft of the manuscript. After the critical revision of the manuscript by UH, the final draft of the manuscript was written by the authors, and all authors gave final approval of the submitted version of the manuscript.
The author UH declares several potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Within the last three years, UH was: an advisory board member for Lundbeck, Takeda Pharmaceuticals, Servier, and Otsuka Pharma; a consultant for Bayer Pharma; and a speaker for Medice Arzneimittel, Novartis, and Roche Pharma. ED, XW, CS and EK declare no competing interests.