Principal Findings
In general, we infer a growing interest in utilizing smartphone apps for delivering psychological treatments, with research increasing from only a few (7) articles in 2013 to an order of magnitude more (60) in 2018. This is a promising trend, as these apps can complement therapist-led psychological treatments and, hence, increase their efficacy and availability. When delegating (part of) psychological treatment to smartphone-based interventions, the need for face-to-face sessions and manual follow-up is decreased, which, in turn, lowers costs and reduces waiting lists in the public health system. According to the mental health workforce breakdown (by the World Health Organization region), there are only 4.6 psychologists per 100,000 inhabitants in Europe [188]. For Spain, studies have reported wait times of more than 45 days before the first psychological assistance by a clinical psychologist or psychiatrist [188], and a frequency of face-to-face sessions of around once a month [189]. Moreover, leveraging smartphones’ capabilities enables ecological momentary interventions (EMI), whereby patients are able to access psychological care when and where they need it most, in their natural environment and daily routines [190-192]. Below, we discuss our concrete observations on the assessment types of included studies, coverage of mental disorders, and technical features.
Evolution of Research and Types of Assessment
It is a promising sign that overall, the number of articles with some sort of assessment is slowly increasing. Furthermore, we observe that proportionally, there is a much higher number of studies with an evaluation of only usability/UX, compared with only effect. Usability factors have been widely recognized as key factors to enhance the acceptance of information and communication technologies (ICT) tools; on the basis of the technology acceptance model, authors have suggested that the intention to use a product in the future is strongly correlated with its ease of use [193,194]. Hence, initial efforts to research and ensure the usability of new ICT tools are essential. At the same time, we observe an overall much lower number (less than half compared with usability/UX) of studies that explicitly assesses the effect of smartphone interventions on clinical symptomatology (despite a peak in 2017). Yet, it is specifically this type of studies, focusing on the (long-term) clinical effects of the intervention, that are needed to demonstrate efficacy, and increase therapists’ and patients’ trust in smartphone-based interventions.
Moreover, RCTs, which are considered the gold standard of experiment design in mental health (and medicine in general), are only minimally represented in the literature (22/158, 15.8% of articles overall). Among them, we see an even spread of combined effect and usability/UX evaluations over time; for studies specifically focusing on effect, however, the majority took place in the last 2 years (with a peak in 2017). This is a promising sign, although there are still relatively few effect studies (see Multimedia Appendix 3). Furthermore, the most commonly addressed disorders using the RCT methodology are depressive disorders, followed by anxiety disorders. It is, thus, important to carry out more RCTs to prove mental health apps’ efficacy in treating other mental disorders, and to study the satisfaction and experience of the patients using these apps. Moreover, to draw rigorous and trustworthy conclusions on the clinical efficacy of smartphone apps, more long-term RCT studies will be needed (eg, to better measure the effects of attrition). Similarly, we observe a distinct lack of cross-validation studies, with only a few apps having been studied in multiple articles (19/138, 13.8%). One could note that this phenomenon is correlated with the lack of rigorous long-term studies on smartphone interventions—a single, multiyear study would warrant multiple articles for a single smartphone intervention on study protocol, usability evaluation, and effect studies at multiple intervals. Particularly when utilizing novel technological features, rigorous assessment studies are needed to validate their potential for psychological interventions and encourage further research in the field. A stronger cooperation between research groups could increase the resources needed for such long-term psychological intervention studies.
Covered Mental Disorders
To an extent, the coverage of mental disorders in the relevant literature seems to be in line with their real-world prevalence. This holds, in particular, for depressive and anxiety disorders, commonly called emotional disorders [195]; they (1) represent the first and third most covered disorders in the literature (we point out that various disorders include apps addressing multiples disorders, where depressive and anxiety disorders, in addition to serious mental illness, are most represented), with the research on depressive disorders being doubled in 2018; and (2) they are known to affect the most people worldwide. For mental disorders with highest prevalence among people [196,197], lifetime prevalence has been estimated at 28.8% for anxiety disorders, 20.8% for mood disorders (including 16.6% for depressive disorders, which are a mood disorder), 24.8% for impulse-control disorders, and 14.6% for substance use disorders. Estimated 12-month prevalence follows a similar trend: anxiety disorders are the most prevalent class with 18.1%, followed by mood disorders with 9.5% (including 6.7% for depressive disorders), impulse-control disorders (8.9%), and substance disorders (3.8%) [196,197]. Hence, according to the psychological literature [198,199], the three most prevalent mental disorders include anxiety, mood (including depressive disorders), and substance disorders. Indeed, these similarly make up our top four of most covered mental disorders in smartphone intervention studies. Depressive and anxiety disorders reduce a patient’s psychosocial functioning and quality of life [198,200], and are associated with important personal, social, and economic repercussions [199,201]. Other ICT technologies for delivering psychological treatments, such as internet and Web-based programs, are also mostly focused on depressive and anxiety disorders [202,203]; this might also have had an influence on the proliferation of smartphone-based interventions. Although determining the underlying factor(s) behind the distribution of addressed mental disorders in the literature is certainly an interesting exercise, we consider this beyond the scope of this paper.
Beyond depressive and anxiety disorders, the literature is heavily focused on only a small number of disorders; six mental disorders account for approximately 73.4% (116/158) of research. On the other hand, more than half of the categories of mental disorders listed in DSM-5 (15) are fully excluded or very scarcely represented (<3%). Clearly, there is an opportunity, as well as an acute need, to pay more attention to the whole breadth of mental disorders—that is, including those that are less prevalent—to help as many people as possible. Some of these less prevalent disorders, such as personality disorders, often have a higher severity that may lead to extreme consequences. For instance, borderline personality disorder affects only 2% to 6% of the population [204,205], but its mortality rate by suicide is one of the highest in the world among people with psychiatric disorders [206].
Coverage of Technical Dimensions
When looking at technical dimensions, more traditional software features (see 7×12 vertical rectangle in Figure 5) are much more utilized than the novel sensing or analytical capabilities of smartphones. One may argue that these top seven features, which involve intervention-specific features (eg, learning) and communication features (eg, prompting), do not offer a significant advancement over the prior state of the art. Indeed, many previous studies that leveraged (nonsmartphone) mobile phones supported learning by displaying psychoeducational content [56,58,140], receiving tips/reminders via SMS [47,149], using (bidirectional) SMS communication to perform (in situ) assessment [149], or telephone calls to health care providers [75,81,145]. Notwithstanding these observations, even this rather conservative transition to smartphones has enabled interventions that are out of reach for classic mobile phones. Research leveraging smartphones have exploited larger screen resolutions and multimedia capabilities to provide multimodal learning materials, using audio and video guides [122,184], pictures [71,75-77,87,154], audio [71,76,87], music [75-77], and video [71,77,174]. Some authors have leveraged the improved connectivity and ubiquity of smartphones to offer access to entire Web-based libraries of learning materials [152,184]; others utilize in-app prompting as intervention techniques, for example, sending reminders to use the app [76,85,116,178], motivational messages [47], or messages from the therapist [80,103]. We found studies that exploit the improved interactivity of smartphones to provide interactive quizzes for training skills and improved learning [62,152], assessments for panic attacks [133], suicidal intentions [143], symptoms of various disorders [54,120,121], and communication with therapists [37,58,92,133] or other users [79,104] through message/chat. Furthermore, aside from being better supported by smartphone capabilities, many of these psychological smartphone interventions are available at the touch of a button, instead of relying on receipt of SMS or phone calls.
That said, most studies still only scratch the surface of advanced smartphone capabilities. This is particularly apparent in the relatively low coverage of context awareness, that is, leveraging sensors to detect and react upon the current state of patients and their environments. We argue that such context awareness is a key ingredient of true EMI. Indeed, although EMI are meant to proactively issue suitable therapeutic interventions at the right time and place, most EMI studies consider smartphones merely as a tool for manually accessing interventions, or receiving predefined interventions at set time intervals, at any moment and place. We found a very limited number of smartphone-based studies leveraging external sensors for recording physiological parameters: measuring heart rate for detecting physiological arousal in the context of anger management [78], for instance. Similarly, we found very low coverage of analytics-based studies that could support advanced context awareness, for example, learning and assessing mental states based on physiological, environment, activity, and/or behavioral contexts. In our opinion, studies that progressively use internal and/or external smartphone sensors, possibly combined with advanced analytics, are a useful step toward realizing the full potential of EMI, where relevant events are detected through analysis of sensor readings, and acted upon by triggering suitable, personalized interventions when they are needed. The general hesitance toward context-aware EMI could be explained by the lack of validated computerized psychological models for assessing mental states based on patient context (including physiological, environment, and activity factors), as well as the need to combine technologically advanced solutions (ie, use of sensors, context awareness, and analytics).
Through their improved modalities (screen and sound), interactivity, and computational resources, smartphones enable novel intervention delivery paradigms, including virtual and augmented realities. However, these have found very limited coverage in the literature. We found a few individual studies, for example, utilizing a mobile VR system to help patients coping with agoraphobia by guiding an avatar through real-life simulations in a game-based setting [122]. Beyond exposure-type therapies, Repetto et al [137] utilized VR techniques to cope with generalized anxiety disorders, leveraging biofeedback to regulate features of the virtual world (eg, current heart rate). As mentioned before, the seeming lack of nontraditional intervention methods may be because of the lack of validated psychological models for supplying evidence-based VR or augmented reality, and/or the technical difficulty of novel delivery paradigms.
Given the lack of studies on these topics, we believe that there lies a huge potential for future research in utilizing technologically advanced solutions (ie, sensors, context awareness, and alternative delivery paradigms) to deliver smartphone-based psychological interventions tailored to the patient’s current health context.
Barriers to Implementation and Patient Risks
Despite our advocation for further research in the utilization of smartphone features to advance treatments for mental disorders, we note that technological innovation should not constitute a goal in itself. It must provide a distinct advantage toward patient care, such as improved mental health care access, that is, a broader reach and lower barriers; increased assessment frequency and accuracy; lower cost; improved efficacy; and immediate access to care, when and where the patient needs it most. These intended benefits must be balanced with possible adverse effects, and the use of treatment modalities, including advanced technical features, needs to be carefully contemplated.
Here, we point out possible risks and barriers to implementation of smartphone-based interventions. The lack of research evidence on the effectiveness of mobile health apps is likely the most important issue to consider [207,208]. As shown in Table 1, only a limited number of smartphone apps are validated using an RCT, and we observed a distinct lack of cross-validation studies, with only a few apps having been assessed in multiple studies. We further point out that this analysis only covers apps presented in the literature and not the many thousands of other, nonvalidated apps in popular app stores. There is a clear risk in using nonvalidated apps, as there is no evidence that they have any therapeutic effect, and they may even worsen the patient’s condition. For example, Baron et al [209] noted that low-accuracy sleep trackers to self-diagnose sleep disturbances caused patients to be overly concerned on getting the perfect sleep (orthosomnia), which may have exacerbated their insomnia.
The ability to protect the privacy and confidentiality of patient information is pivotal as well [207,210-212] and may form a barrier to adoption. Many health-related apps collect a large amount of demographic, medical, and lifestyle information, as well as data on users’ daily routines and practices [213]. Some apps have even been found to collect user data unrelated to the app’s purpose [211]. Despite the importance of privacy with respect to (mental) health data, several researchers have found that a significant portion (31%-49%) of studied mental health apps does not include a privacy policy [210,214]. Collected information may be distributed to (third-party) services for storage and analysis. Moreover, unregulated apps may even pass on this information to unidentified parties, for example, for advertising purposes [210,211]. In a study of depression and smoking cessation mobile apps, it was observed that the majority (81%) of the studied apps transmitted data for advertising and marketing purposes, but only 59% of the apps disclosed this in their privacy policy [210]. In all the aforementioned cases, there is a risk that personal data are being transmitted over insecure network connections [215,216]. Furthermore, there is the possibility of loss, theft, or malfunction of the mobile device [5], or cybercriminals specifically targeting health information [207], all of which could result in the loss of sensitive data.
Toward addressing these two issues, some authors suggest that mental health professionals should screen the apps they recommend for privacy issues [217]; however, this does not seem like a feasible or robust solution. Olff [218] proposed a disclaimer for apps that have not been validated. Taking it one step further, certification processes for mobile (mental) health apps have recently been instigated in the European Union (EU) and the United States, even though they do not apply to all apps. Under the 2017 EU Medical Device Regulation [219], mobile apps with a medical intended purpose require a CE marking. Devices (apps) may be differently classified depending on the purpose and risk they pose, yielding more or less stringent regulations regarding quality, control, or development process. The EU General Data Protection Regulation [220] has furthermore motivated a number of app developers to improve transparency on privacy policy, although confusion remains on its applicability outside the EU [211]. Similarly, in the United States, the Food and Drug Administration policy for mobile medical apps [221] stipulates that any software that is utilized for the diagnosis of disease or other conditions, or the cure, mitigation, treatment, or prevention of disease, constitutes a medical device, regardless of the platform on which it is run (eg, desktop or mobile). The applicable regulations depend on the particular functionality. Given the recency of these (updated) regulations, their rate of adoption and effects on efficacy, safety, privacy, and usability of mHealth apps remain to be seen, and we expect the regulations to be further clarified, refined, and extended in scope in the future.
Further barriers to implementation include a lack of sufficient general, digital, and health literacy levels [207] and a digital divide [222], the view and attitude of practitioners and patients toward the use of mobile mental health interventions [223,224], availability and awareness of evidence-based apps [225], economic and other associated costs [226], and user acceptance and usability [226]. For further elaboration on barriers and facilitators to implementation of mental health care apps, we refer the readers to Lipschitz et al [225] and Simblett et al [226].
Recommendations for the Research Community
In line with our findings, we propose four recommendations for the research community to further develop and advance the field of smartphone-based psychological interventions:
Attention for less covered disorders: The majority of research (approximately 73%) pertains to the top six covered mental disorders, four of which coincide with highly prevalent mental disorders. On the other hand, more than half of the DSM-5 recognized mental disorders are not or very scarcely covered. We call upon the research community to invest into covering the full breadth of mental disorders.
Attention for advanced technical and software-based solutions: Many smartphone-based psychological interventions merely translate traditional and electronic health (eHealth) solutions to smartphones; that is, these interventions do not fully exploit their capabilities. Concretely, the use of sensors and corresponding context awareness, particularly to promote EMI, the exploration of alternative delivery paradigms such as virtual or augmented reality, and more advanced analytical methods are scarcely investigated. We call upon the research community to explore beyond traditional strategies, toward leveraging advanced technological features to improve mHealth interventions.
Multidisciplinary approaches: To fully exploit the smartphone’s capability as a pervasive, ubiquitously connected, sensor-packed computing platform to deliver innovative, real-time, and in situ psychological interventions, both the domain knowledge of psychologists and the technical expertise of computer scientists are needed. Hence, we call for multidisciplinary collaborations as to not let technical difficulties, or lack of psychological knowledge on mental disorders, hinder advances and novelties in the field.
Validation toward effect: Although we uncovered, at least in absolute numbers, a slight increase in effect validations and RCT-based effect assessments during the last 2 years combined, they are still underrepresented (particularly RCTs). Hence, there exists a need to rigorously validate smartphone-based psychological treatments for effect. Especially when utilizing advanced technical features (eg, context awareness, analytics, and alternative delivery paradigms), effect validation may increase trust and spark further research in such novel types of interventions. We call upon the research community to augment efforts in rigorous effect assessment, to allow transfer of research into practice.
Despite our call for research in technical innovation and its broader applicability and validation, we note that the eventual use of such advanced technical features—or any technological aid in psychological interventions for that matter—needs to be carefully balanced with the characteristics and needs of the individual patient.