Traditionally, clinicians focus on collecting one-time snapshots of patient information in a clinical setting and making decisions based upon them, thereby losing an opportunity to create a thorough understanding of the patient’s health status [20]. In these situations, PGHD are a useful tool for continuous monitoring, especially for patients with chronic conditions that require daily management and benefit from effective tracking [21-23]. PGHD can also improve disease surveillance by more accurately assigning patients to disease categories rather than solely using national and regional data [24].

PGHD can come from multiple data sources, such as family history, medication, and physiological data sensing [25]. PGHD, along with other health data, can ultimately form a repository of patient-centered personal health records, which can then be used to store and manage PGHD [26]. Clinicians see the potential of PGHD but raise some concerns about its use, such as difficulties in summarizing PGHD patterns across different clinical specialties and concerns regarding data management, patient privacy, and ethical challenges of PGHD generation [9,27,28].

Internet use has been growing steadily since 1990, with approximately 4 billion internet users worldwide, with 15- to 24-year-olds leading the frontier of internet adoption [29,30]. The internet has also become increasingly mobile. An unsurprising result of this trend has been increasingly sophisticated mobile technology and wearable devices, including the development of numerous mobile health (mHealth) apps, all of which have helped generate large amounts of PGHD.

Wearable devices and mHealth apps can facilitate health behavior improvements if designed with proper engagement strategies and data collection methods [31,32]. Effective mHealth app design has helped users monitor and promote positive health habits, such as physical activity and eating behaviors [33,34]. Furthermore, PGHD can help improve clinical trials’ efficiency and output [35]. However, these opportunities come with a number of challenges [2]. First, the management and interoperability of massive PGHD require standard vocabularies and data models. Consumer health vocabulary (CHV) is one such effort that has been developed since 2007 [36,37] and enhanced by several text mining–based projects [38-41]. In addition, frameworks and data models have been proposed to incorporate PGHD into EHRs [42,43]. However, very few projects use both CHV and common data models to facilitate PGHD comprehension. Indeed, to our knowledge, only one position paper proposes an interpretability-aware framework to systematically understand PGHD [44]. The second challenge is the lack of guidelines and best practices for integrating PGHD into the clinical workflow [15]. In addition, most PGHD are collected through mHealth apps, which should be carefully evaluated in a standardized manner for efficacy and health outcome improvement [45]. Finally, there are concerns about the quality and ownership of PGHD gathered by mHealth apps and wearable devices [28,46]. App developers should develop more transparent data ownership policies so that users can make informed decisions regarding their PGHD [47]. Guidelines should also be developed to ensure that high-quality data are gathered by users in their daily routines [48].

Psychiatry is a medical specialty focused on “the diagnosis, treatment, and prevention of mental, emotional, and behavioral disorders” [49]. Clinicians use a variety of data from patients to determine psychiatric diagnoses. However, these data are often collected solely in clinical settings. Since the 1960s, mental health clinicians have begun to pay more attention to patients’ personal perspectives, including health-related quality of life (HRQoL) outcomes. HRQoL is a type of PRO that includes “symptoms of disease or health condition, treatment side effects, and functional status across physical, social, and mental health life domains” [50]. PROs are derived from patients completing standardized questionnaires and cannot guarantee large-scale, continuous data collection [51,52]. However, with the help of mobile technology and wearable devices, PGHD can collect large amounts of patient health data unobtrusively and continuously. Furthermore, because psychiatry has historically depended on PROs and HRQoLs, clinicians in this field should be able to use PGHD without many conceptual barriers.

For instance, ecological momentary assessment (EMA) was proposed to assist clinical psychologists in monitoring HRQoL changes by tracking patient behaviors in real time and in their natural environment. EMA uses various data collection tools, such as written diaries and telephones, and PGHD gathered by mHealth apps and wearables can further support this approach [53]. PGHD can facilitate large-scale environmental psychiatric research in naturalistic settings and create digital footprints to measure patients’ health status (eg, mood and sleep) in an unobtrusive and longitudinal fashion [54].

Although studies have shown that wearable devices and mHealth apps help in the treatment of mental health by increasing awareness and giving reinforcement, such as in the study by Ng et al [55], research regarding the implementation of mobile technology into the care process is lacking. Studies have only just started to design and develop apps that focus on interface usability and workflow integration. Notably, Bauer et al [56] applied Principles for Digital Development to develop a highly usable mHealth app to support collaborative care for patients and generated 4 more principles based on user feedback. Recently, mental health studies have focused on PGHD, showing a transition to participatory and personalized medicine [57]. However, other stakeholders, such as consultants, policy makers, and vendors, should also be considered when integrating PGHD into mental health [58].

Clearly, more PGHD will be generated in the coming future through mobile technology and hold the potential to improve mental health practice, which has long relied on patient-reported data. Therefore, we follow this trend and aim to provide empirical evidence regarding the current use of PGHD in mental health practices, clinicians’ attitudes toward PGHD, and the mHealth app features and selection criteria considered by mental health practitioners.

Figure 1 shows a simplified version of the consolidated workflow diagram, which includes step-by-step movement from a patient taking an initial assessment, development of diagnosis and treatment plan, meeting with a physician, the influence of PGHD, and subsequent adjustments in the diagnosis and treatment plan. The detailed workflow diagram is included in Multimedia Appendix 2.

As shown in Figure 1, a typical clinician workflow in mental health started from gathering patient information and PROs using an intake assessment. Mental health clinicians then reviewed the form along with EHR information, if any, and conducted a clinical interview to establish psychiatric diagnoses and a corresponding treatment plan. Next, patients participated in the treatment (eg, practice assignments and collection of PGHD), following their clinicians’ requirements. The ways to collect PGHD varied depending on the clinicians’ preferences and patient situations. Next, in subsequent clinical encounters, clinicians reviewed the PGHD and assessed the changes in their patients’ symptoms and functioning since the previous session. Clinicians may revise the treatment plan based on the updated patient health status.

Many participants do not consider themselves as the sole decision maker in developing the treatment plan. Instead, they worked with patients to align their goals with the treatment. Homework is a common term used by clinicians to discuss patients’ efforts to improve their mental health status between clinic visits. PROs and PGHD can assist patients in their homework, demonstrate their achievement, and provide information to clinicians to make evidence-based decisions on the treatment plan and assess treatment progress. However, participants reported no standardized way to collect and manage PGHD.

Although collecting PROs and PGHD has been part of mental health practices, its use is not optimized. Taking the intake forms as an example, one participant talked about troubles using survey scales when patients wait in the waiting room. This suboptimal data collection may slow down the clinics and reduce the efficiency and quality of care:

Sometimes, PROs can be ambiguous and confusing and require more data to understand patients’ health status change and nuanced differences. In this case, PGHD can be complementary and provide more detailed behavioral data to inform the shared clinical decisions on the treatment plans. One participant explained how PROs may be confusing:

All participants followed the simplified consolidated workflow in their practices. However, there were some noticeable differences between psychiatrists and clinical psychologists. Psychiatric appointments, in general, were shorter (30 min), followed by appointments with clinical psychologists (45-60 min). As psychiatrists can prescribe medication, they check medication use and effects in every patient visit. Psychiatrists also conduct psychotherapy and value the therapeutic relationship between patients and themselves. On the other hand, clinical psychologists cannot prescribe medication, so they focus solely on conducting psychological assessments and delivering psychotherapy. Therefore, the workflow for clinical psychologists can be very dynamic and conversational. Clinical psychologists pay significant attention to patients’ narratives of experience to understand their unique health status and changes. The workflow would be more standardized, however, if a clinical psychologist only focuses on psychological assessments, for example, a cognitive evaluation, as the assessments have a validated procedure to follow and tools to use.

All participants had worked with their patients to track their sleep and/or mood behaviors through an mHealth app and/or a wearable device. Overall, 6 participants held a dual attitude toward PGHD use. On the one hand, clinicians had seen the potential of PGHD and looked forward to taking advantage of them, especially the ability to track their patients’ activities between consultations better:

On the other hand, some concerns were raised about data validity and reliability. The participants were being cautious because they identified the need for reliable PGHD to inform evidence-based treatment implementation and evaluation:

Clinicians’ concerns about integrating PGHD into workflow and EHR systems were a recurring theme in the interviews. One participant indicated that there was no app that seamlessly integrates its data into EHR:

A total of 7 participants preferred using paper-based PGHD in their practice because of the patients’ preference or the lack of data sharing mechanisms in the apps. In this case, the data flow is deemed indirect. Clinicians would review the PGHD in the session and put the interpretations in the clinical notes, which can slow down the clinical workflow. Generating data visualization of PGHD to facilitate the interpretation of the data seems to be a preferable method for both clinicians and patients:

Manually transcribing data into electronic medical records is not only time consuming but also interferes with the patient-clinician interaction because the clinicians are distracted by the data entry tasks on their computer. Common concerns found included the difficulties of having to face the computer and type in data during sessions:

Lack of Information for App Selection

In addition to the challenges in using data from mHealth apps, it can be challenging to identify what apps to use in the first place. In all, 3 participants found that it was difficult to find an app that met all the requirements. Others sometimes found useful apps through patient recommendations. The following 2 quotes exemplify this situation:

When asked about using apps for collecting data, some clinicians responded that they did not know that mobile apps could perform certain data tracking tasks:

In contrast, the app review (see details given in the section Sleep and Mood Tracking App Reviews) showed that mood apps in the current market can offer certain functions, which means there is a lack of information to increase app awareness and support app selection process, especially for clinicians. This suggests that an app recommendation system for the clinician to use would be very helpful.

One factor contributing to the difficulty of finding the right app is the variety in patients’ health statuses and conditions. Clinicians would prefer to personalize treatment and collect data in a personalized manner. However, current apps were not designed with features to support personalization in data collection. One participant elaborated that:

Another type of feature that current apps may be missing is to support collaborative care. It is beneficial to support collaborative care so that patients can be more engaged in their health and take control of their care, and clinicians can create a treatment plan that is most suitable and effective for the patient’s situation. One clinician further talked about the difficulty in finding an app in current app markets to address his needs of data collection, although they do have data collection features:

Table 2 shows the statistical summary of the selected apps. The data were retrieved on September 8, 2019. The average number of reviews was around 400, and the average user rating was around 4.0 for both sleep and mood tracking apps.

The data interpretation, which includes collection and visualization, and sharing features in the selected apps varied. All 31 sleep tracking apps offered at least one feature for data interpretation, including sleep quantity statistics (n=27) and sleep quality analysis (n=5). Of the 31 sleep tracking apps, 18 (58%) support data sharing in various means, including direct sharing with other people (n=8) and integration with other apps (n=7).

The major source of data interpretation is through collecting statistics, through recording several days' worth of accumulated data points. These data points are either manually inputted by users or automatically recorded by the app. Overall, 5 apps also offer qualitative assessments of sleep patterns by presenting the sleep cycles that users experience, recording sound files, and/or providing descriptive sleep analyses.

Many sleep tracking apps also support sharing data. This is usually done by exporting and downloading the data as a comma-separated values (CSV), PDF, etc file (N=8) or via integration with an alternate app, such as Apple Health (n=7). Other sources include social media (n=3) and email (n=4). Many apps also support multiple forms of sharing.

Many users find accurate sleep tracking to be helpful in improving their sleep quality and daily life:

On the other hand, common complaints of sleep apps included inaccurate tracking, failures in data collection, and difficulty in use due to technical issues. For example, one user expressed frustration in inaccurate tracking:

The analysis of mood tracking apps shows similar results. Of the 25 mood tracking apps, 21 (84%) offered either a qualitative (n=4) and/or quantitative (n=22) data interpretation feature. Eight mood apps had more than one mode of quantitative data interpretation. Of the 25 mood tracking apps, 16 (64%) had data sharing capabilities, mainly through direct sharing with other users (n=13) or social media (n=5). Three mood apps had more than one mode of data sharing.

Many apps offer multiple means of data interpretation. The forms of qualitative data interpretation include monthly reports of the moods logged. Forms of quantitative data interpretation include daily, weekly, or monthly graphs or charts of the frequency of moods logged by the user. The modes of sharing data are exporting the data as a spreadsheet, CSV, PDF, etc file (n=6); via email (n=3); or via cloud sharing (n=3).

Many users find that the ability of the apps to track and reflect on one’s mood trends was very helpful:

The most common complaint was the limited options of mood offered, which caused difficulties for users in tracking their real mood, as indicated in the following review comments:

With both mood and sleep apps, the primary method of data visualization was through graphs and charts (n=37). Sleep apps, which recorded quantitative data more frequently than qualitative data, used statistics (such as the length of time slept) to present graphical summaries of sleep over time. These graphs were then analyzed to show the quality of sleep over time, track trends, inconsistencies, and improvements. Similarly, mood apps also used graphical representations to present data summaries. However, these apps typically used qualitative data entries (such as recorded moods on certain days) to provide graphs and charts that showed the number of times certain moods were recorded and how moods fluctuated over the course of weeks or months. The popular usage of graphs across both categories suggests that regardless of the nature of data collection (either quantitative or qualitative), users prefer apps that visualize data through digestible and succinct representations that make trends identifiable and trackable over time.

Current mood or sleep tracking apps are focused on collecting PGHD in a patient-centered manner, which is a critical first step. However, to maximize the value of PGHD, these apps should improve their mechanisms in data interpretation and sharing. Specifically, data visualization can be a viable way to help both mental health clinicians and patients interpret much PGHD and identify patterns and trends regardless of their integration into EHR. On the other hand, mHealth apps should enable data sharing mechanisms with different parties, including, but not limited to, clinicians, families, friends, and other practitioners, as this was a concern noted by both clinicians during interviews and users in the app review [58]. It is worth emphasizing the importance of information confidentiality when designing a sharing mechanism. Psychology clinics are considered as a safe bed for patients to discuss their mental health status and conditions with clinicians. Hence, data sharing mechanisms should not be one-size-fit-all; they should be designed to allow patients to select which part of PGHD to share and how to share to keep highest data confidentiality based on their psychiatric conditions.

Technology-enabled clinical data capture and documentation should consider the clinical workflow [65]. Similarly, we suggest that mHealth apps designed to gather PGHD should consider clinical workflow to improve the quality of patient experience. Although PROs and PGHD have been used in mental health clinics, there are no guidelines for data collection and use. It would be beneficial to conduct observational workflow analysis, such as time and motion studies, to understand when and where PGHD are used and identified bottlenecks. In fact, it is common to conduct such workflow analysis to design and develop any health information technology that will be used in clinics (eg, clinical decision support tools) [66].

On the other hand, mHealth apps should also consider EHR integration. We have discussed the importance of sharing PGHD with various parties in a way to protect information confidentiality. PGHD may be shared with clinicians directly through EHR integration or indirectly through data summarization and electronic reports. We have also seen a working example in our interviews to integrate PGHD with EHR using a Research Electronic Data Capture database [67]. As clinicians may have concerns on data reliability and accuracy, having an indirect EHR integration (eg, a dashboard or PDF report demonstrating patterns in PGHD for review) may be a viable way to reduce clinicians’ concerns and increase their PGHD adoption in clinics. As many apps are coming into the market and are also going obsolete at a fast pace, a standardized data management and export system could be proposed to better integrate PGHD into clinical practice regardless of specific kinds of apps.

Our results showed that mHealth apps must support both personalization in data collection and collaboration between patients and clinicians during clinics. In terms of personalization, as each patient has a unique mental health status and conditions in various social contexts, it may be difficult to find an app that covers all kinds of needs of PGHD collection. There are 2 ways to approach this issue from a design perspective. First, mHealth apps should maximize their ability to personalize data collection methods to fit different patient needs. Participatory design methods may be helpful in identifying such needs and incorporating them into app features. Second, as clinicians do not always have the information of some existing apps that are potentially useful, an app recommendation system may be developed to assist clinicians in choosing which apps to use to collect PGHD. Currently, some clinicians rely on patients’ recommendations. This app recommendation system may be maintained by both clinicians and patients.

In addition, mHealth apps should be designed to support collaborative care. It is critical to ensure that PGHD are used effectively and efficiently during clinics. Ethnographic observations may be needed to systematically document the behaviors and interactions between patients and clinicians. The results can help researchers better understand the role that PGHD and mHealth apps play during clinic visits and generate guidelines to redesign mHealth apps and improve the use of PGHD to support shared decision-making and collaborative care.