Fatigue and particularly chronic fatigue syndrome (CFS) are conditions with a strong impact on quality of life. Crook et al [1] define fatigue as “more profound than being overtired; it is unrelenting exhaustion and a constant state of weariness that reduces a person’s energy, motivation, and concentration.” Within human-computer interaction (HCI) research on physical health, fatigue has been less explored, although recent work, focused on lived experiences of long COVID, has started to emerge [2]. In parallel to academic research, there has been a growing number of mobile apps targeting fatigue. Mobile apps for physical or mental health, and in particular, those highly rated by their users, are likely to reflect valuable design principles which benefit from being articulated as implications for design. Similar efforts have been made for apps for depression [3], digital well-being [4,5], and anxiety [6], but less so for fatigue. To address this gap, we report a functionality review of 17 top-rated apps for fatigue from the Apple App Store and Google Play Store, together with an analysis of their descriptions on marketplaces. Findings indicate that the main functionalities of these apps are tracking fatigue and its symptoms, visualizing tracked content, assessing symptom impact on everyday life, and providing interventions. We conclude with 5 design implications for better support of people living with fatigue, which include supporting hybrid tracking to balance the validity of self-reported fatigue with low-burden automatic tracking of symptoms, contextualizing fatigue data with symptoms and associated factors through combined visualizations for sense-making, and supporting energy estimates and pacing interventions.

To support our exploration and understanding of the functionalities, we draw on the limited but emerging area of HCI work on fatigue, comprising the tracking of symptoms, the design of technologies supporting their management, and the emerging area of research that has engaged in the analysis of apps created for this purpose.

A key area that we first needed to understand was the behaviors of individuals with CFS surrounding their use of technology to have a better idea of any ramifications that may arise from the condition, and consequently how they can be designed around. Paymal and Homewood [7] conducted a study to identify how CFS impacts technology usage in day-to-day life, in which they make use of their findings to present design recommendations. Their findings include the reduction of technology use on days where participants feel better, the increased usage of technology to help with tasks when symptoms were more noticeable, the avoidance of technology by some participants when their symptoms were more severe, and the usage of apps for tracking and self-management purposes. While the researchers identified how self-management technologies were used by this group, their reasons for turning to these technologies, and their criticisms regarding their usage experience, the researchers’ findings could be further supported by detailing the different functionalities offered.

Virtual reality (VR) is also an area that has been the focus of research; Best and Butler [8] investigated the use of VR as a support space for individuals with CFS. They discovered that CFS symptoms would prevent users from engaging with the technology for extended periods of time; however, this led to users improving their energy management to use their limited reserves more efficiently, in addition to optimizing their physical spaces and the technology they use to minimize fatigue triggers. This highlights the interplay between their energy levels and their self-tracking efforts, and is a key area of concern when designing for this group of users.

Davies et al [9] investigated the experiences of patients with chronic fatigue regarding their usage of technology for disease management, and a key area of interest was the participants’ usage of apps for self-management. In their study, they set forth design guidelines that are centered around the symptoms that patients want to manage the most, including features that assist users in identifying maximum and minimum energy thresholds, adding activity pacing functionality coupled with routine establishment, providing support for fatigue tracking by exploring automated logging methods to reduce effort expended, and finding ways to make self-management apps more engaging and motivating such as by gamification. Their findings can better inform our evaluation as the guidelines provide a frame of reference to examine the apps against.

Research has also delved into understanding how to support users with energy-limiting conditions (ELCs) through technology. For instance, Børsting and Culén [10] applied a sense-making process for complex domains in order to explore the design of technologies for supporting young individuals with CFS. The approach is a combination of a literature review, mapping techniques, and user research. The results of the user research, which was conducted with CFS experts, led to suggestions targeting self-management, namely, providing interventions discovered by other individuals with CFS, offering reminders for both resting and general purposes, providing functionalities to track and warn users of noise levels, providing social support technologies, and finally, providing the logging of symptoms coupled with tracking technologies suited for CFS.

Another strand of work has focused on technologies for supporting the management of chronic conditions. For instance, Sas et al [2] reported findings from workshops with individuals affected by long COVID, suggesting design implications for tracking and sense-making of fatigue and postexertional malaise (PEM), as well as self-management aimed at PEM prevention. Their findings stressed the importance of lightweight capture of data to prevent symptoms worsening among users, conjointly with reducing the burden of identifying both triggers and the PEM-approaching stage by making use of machine learning (ML) techniques or noninvasive items such as heart rate variability (HRV) tracking, and finally, supporting self-management of fatigue based on the stage of PEM that a user is at.

We also note the study of Homewood and colleagues [11] who co-designed through asynchronous communication with the aim of catering to the needs of individuals with ELCs. The researchers discovered that this method helped participants to only engage with the study when their energy levels were sufficient to partake; in addition, they were able to extract meaningful insights into improving the support for individuals with ELCs. These included designing technologies that improve communication with their social networks, implementing automatic tracking of their fatigue levels and associated symptoms, and using crip approaches in future studies dealing with ELCs.

Studies have also found that if participants’ fatigue symptoms reached more severe levels, they may avoid interaction with technology to prevent overwhelming them [7,9,12]. For this reason, designing around the potential fluctuations in their energy levels is paramount [2,9,10,13,14], and one such way this can be achieved is by including automated logging methods, especially in scenarios where users track numerous types of data [2,9]. In the same vein, Mack et al [14] proposed tenets to improve the accessibility of technology, and from these, one tenet called for designs that account for rises in fatigue levels post exertion and consider the consequences of a task on an individual’s energy reserves.

The findings detailed in this section can be used to better evaluate the current design space of commercial apps and provide more informed implications for future app designs.

Other areas of HCI research targeting chronic conditions have focused on the usage of online resources to support CFS, as in the study of Brigden et al [15]. This study, along with several others, found that social media could be of value to individuals with chronic conditions as a means of discussion and social support surrounding their condition [7,10,15-17]. These studies highlight the importance of observing for social media–based functionalities in the apps to be reviewed.

Another study is by Best and Butler [12], where they sought to identify the physiological effects experienced by individuals with CFS when socializing in a virtual world. Their findings also include the previously mentioned user behavior where they withdraw from technology as their fatigue levels worsen, so as to prevent further deterioration [7,9,12]. As such, the level at which users must interact with the apps could be a useful metric to evaluate the apps and think of further design implications.

PEM is also known to be a characteristic of conditions such as long COVID [2]; therefore, the research in this field has value for the area of fatigue management. From the works in this field, we note the study of Homewood [13], who undertook an autoethnography where they used a Fitbit device for activity pacing purposes to manage their long COVID symptoms. This is also one of the studies that highlighted the importance of designing around the potential fluctuations in energy levels [2,9,10,13,14].

Another study of interest is that of Homewood et al [16], who conducted interviews with participants who had long COVID and made use of wearables for disease management, with findings highlighting the need for designs that are tailored to account for individual differences.

In addition, Pater and colleagues [17] carried out a 3-month-long cohort study on patients with long COVID, which involved the collection of both objective and subjective data. The authors emphasized that cognitive symptoms may make data gathering difficult, with participants finding it challenging to enter survey data regularly due to forgetfulness, which also made interviews burdensome [17]. They also posited that gaps in data tracked during their study were caused by participants forgetting to charge their wearable devices.

Finally, Mehdipour and Aharari [18] conducted a literature review along with case studies to analyze the marrying of the fields of AI, HCI, and Internet of Things (IoT)—which is defined by Kelly et al [19] as “a system of wireless, interrelated, and connected digital devices that can collect, send, and store data over a network without requiring human-to-human or human-to-computer interaction”—in regard to the analysis of data surrounding COVID. Their findings highlight that these fields aided the process of understanding large health datasets built from the data of numerous individuals.

These findings reveal the need for flexibility in tracking options to prevent overwhelming the user, and this exploration can be aided with HCI principles.

We also draw from research on chronic pain, as it is a condition for which self-tracking is used for self-management. Adams and colleagues [20] investigated preferences regarding the self-assessment of pain for the purpose of self-managing chronic pain, through which they also discovered the previously mentioned importance of tailoring designs to suit individual differences [16,20].

Singh et al [21] carried out a study to explore the use of movement sensing and sonifying wearables by individuals with chronic pain in their undertaking of functional activity. The researchers found that the device challenged the user’s initial beliefs on their limits by improving their bodily awareness, and this led to users feeling more confident in trialing activity pacing methods. This highlights the importance of tracking and identifying limits for their self-management efforts.

Personal informatics is an interdisciplinary research area within the broader HCI discipline, focusing on how people use personal technologies in life to collect, organize, and use information about their daily lives, in order to support personal goals in a range of domains such as productivity, learning, well-being, and health [22]. In the space of chronic conditions, this research area contributes design principles for digital health technologies such as mobile apps aimed at supporting fatigue. HCI work on chronic conditions has focused on tracking symptoms [23,24] or data related to physical activity [22,25,26], heart rate [27,28], sleep [22,29], and various other general health aspects [23,24,30].

Personal informatics also has value in supporting interventions; for instance, Felipe et al [29] found that it can greatly supplement the intervention of pacing, which is commonly undertaken by individuals with ELCs [2,9] and/or chronic pain [29,31]. In addition, individuals with chronic migraines also made use of personal informatics to identify if their condition is worsening in order to take preventative measures [32], which is known to be an important aspect of pacing [2].

Moreover, personal informatics can help users with chronic health conditions understand their condition better [24,29,32], offload the burden of monitoring their condition [29], improve communication and discussions with clinicians [24,32], help identify when their condition worsens, which is essential for taking preventative action in time [32], and finally, aid users in managing and monitoring their condition in the long term [32].

We now report on the findings from the functionality review, namely, the functionalities of tracking, visualizing, assessing, providing interventions, and ensuring privacy. Table 1 provides the main functionalities found in each app.

The daily use of the apps during the expert evaluation provided a sufficient amount of data for the assessment of their visualization functionalities. The findings from this reveal the provision of visualizations of tracked fatigue, symptoms, and associated data through charts representing different types of data, color-coded symptom severity, temporal aspects, and data summaries (Multimedia Appendices 6 and 7).

Findings show that the most common charts for fatigue visualization are line charts (4 apps) and bar charts (2 apps), although not all apps visualized tracked fatigue data (2 apps), despite users’ interest in understanding their tracked fatigue for the purpose of self-management [9].

Fatigue is known to have numerous factors [2,9], and for this reason, users will want to reflect on other data sources as well. Findings indicate that 9 apps provide visualizations of a single data type, such as fatigue scores (6 apps). In addition, as the factors are known to be interdependent [2,9], users would benefit from being able to compare the data on the same chart. To this end, apps also offer visualizations integrating different tracked data (7 apps): most often fatigue data + symptom data (4 apps), fatigue data + health data (4 apps), health data + health data (5 apps), symptom data + symptom data (4 apps), symptom data + health data (4 apps), fatigue data + mood data (3 apps), symptom data + mood data (3 apps), physical condition + mental condition (1 app), or morning + evening mood (1 app). It was also noted that 3 of these apps providing visualizations of combined data do not have the option of integrating fatigue, which may limit users’ ability to better understand their fatigue by discerning factors that could affect it [9].

With respect to the number of different data used for such combined visualizations, most apps involve only 2 y-axis variables per visualization (6 apps), while 1 app allows an unlimited number of variables integrated in one app, which for numbers larger than 3 may be difficult to understand if not designed in an effective manner. Combined visualizations are often represented through multiple line charts, that is, one line for each y-axis variable (4 apps), combined bar and line charts (3 apps), combined scatter charts (1 app), parallel bar charts (1 app) as seen in Figure 3 from the Visible: Pacing for illness app [46], combined bar charts (1 app), and bar and dot combo charts (1 app). Of these, the most frequently used visualization was found to be combined line charts (4 apps). Given the interdependency of various aspects of fatigue [2,9], it is surprising that these visualizations do not look at them altogether in contrast to the isolated view, and this could be an area which future work could investigate. Finally, findings highlight that 1 app does not show combined visualizations, despite tracking different data.

From the 8 apps that provide visualizations for symptom tracking, half of the apps use warm colors such as red to indicate negative symptoms or high severity level (4 apps), while the others do not use such color codes (4 apps). The former highlights severity level, which can increase one’s concerns and be perceived as judgmental [9].

All 9 apps providing visualizations for historic data do so with a range of timescale options such as the last 7 days (1 app), last 2 weeks (1 app), last month (3 apps), last 2 months (2 apps), last 3 months (3 apps), or a custom timescale ranging from the last 10 weeks to the last 7 years (1 app), any 7 days (3 apps), any 2 weeks (1 app), any month (2 apps), any year (1 app), as well as a custom scale with a start and end at any 2 points in time (1 app). These various timescales are beneficial, with the most recent past supporting understanding while leveraging the memory of recent events impacting the displayed fatigue scores, and the far past supporting understanding broader patterns of how fatigue changes over time [57]. Moreover, findings show that 7 apps display only historic data, 1 app displays both historic and real-time data, and another app displays only real-time data. From the 2 apps that show real-time data, both apps use it for biofeedback interventions. In addition, the real-time data are measured with an HRV monitor for one app and with both a camera and an accelerometer for the other. Finally, one app visualizes HRV, and the other visualizes heart rate and breathing rate.

From the apps that visualized fatigue, we note that 2 apps represented all data points per day on the chart, while the other 4 apps represented one data point per day. While 2 of the latter only record 1 fatigue score per day, the other apps either average the score (Guava: Health Tracker) or show the highest score for the day (Bearable - Symptom Tracker). This may be either a benefit or a drawback to self-trackers who deal with chronic fatigue. On one hand, representing all the data points could take more mental effort to digest, which could be remedied by per-day summaries; however, they hide the changes in fatigue, diminishing the insights obtainable from their tracking efforts. This highlights the value of visualizing a minimum of 3 data points a day, with options to show fatigue data before and after activities, as users typically seek to understand the impacts of activities for pacing purposes [9].

Findings reveal that 4 apps display data summaries. One app shows averages of any tracked variable within a chosen timescale. A second app shows the averages of symptoms and steps over a weekly or monthly period. A third app shows averages of morning and evening mood over 7, 30, or 90 days. The fourth app shows averages of daily scores within a yearly view visualization, and averages of monthly scores within a monthly view. These data summaries are important, as providing users a more condensed form of data could reduce the cognitive load of understanding their fatigue, which is important in preventing the apps from exacerbating their condition [2].

Despite the rich set of symptoms and tracked data, the top-rated apps for fatigue provide limited interventions for the self-management of fatigue, and in particular, those leveraging tracked data (Multimedia Appendices 10 and 11). Findings highlight a range of interventions provided by the apps, one of which is neuroeducation (3 apps); a treatment approach that aims to improve user well-being via educational lessons and exercises built around neuroscience concepts [62]. The apps in our review chose an approach centered on the concept of neuroplasticity and one such app is the Gupta Program Brain Retraining (Harley Street Solutions LTD) [63], as seen in Figure 4. Another intervention is psychoeducation (5 apps), which is delivered through articles informing users of their condition along with possible treatments and informing them of new advancements in research surrounding their condition, as seen in the Visible: Pacing for illness app [46] in Figure 4. Functionalities for social support (3 apps) were also implemented, in the form of community engagement through forum posts and social networking. These functionalities offer value to users with ELCs by providing them with a space for support while promoting discussion around their condition [7,10,15-17]. Moreover, 2 apps included biofeedback-based therapy, which aims to regulate the nervous system to reduce symptoms through the use of deep breathing (2 apps) or muscle relaxation (1 app) exercises, the latter of which has been known to reduce fatigue and may help individuals with CFS [64]. Finally, 2 apps were found to offer functionalities supporting pacing (2 apps).

Pacing is an important intervention, tailored for fatigue [1], which involves planning and allocating one’s energy toward activities to be completed based on their energy demands [9] through the so-called spoon management [9,65] or energy envelope approaches [9,66]. Pacing involves identifying what activities one can do, in addition to the extent that such activities can be carried out [67]; additionally, it entails stopping activities when breaching PEM and can also integrate the act of estimating energy levels [68]. Previous findings indicated the value of pacing interventions for managing fatigue in addition to preventing overexertion and PEM [1]. While not an intervention per se, 3 apps also provide support for energy estimates, which is important as it can provide input for pacing interventions.

Estimates of energy are manually tracked through self-reports (1 app), which is the traditional approach as a prerequisite for pacing. Interestingly, another 2 apps use a method of tracking daily energy estimates through HRV measurement in the morning, which is compared to the user’s baseline to provide either a stability score by one app (Figure 5) [46], or a readiness score by the other (Figure 5) [69]. These scores are used to notify the user of potential flare-ups of their conditions and, as mentioned before, can inform their pacing decisions [68].

Previous work suggested the potential of HRV data as an indicator of fatigue [70,71], albeit its use to estimate energy has been less explored. Such approaches to estimate both fatigue and energy levels are crucial to lower the burden of tracking and its cognitive load, which is especially beneficial for an individual who may be approaching PEM [2]. Interestingly, however, not all apps estimating energy also provide pacing interventions.

These interventions are provided only by 2 apps using different energy units. One app uses spoon management [9,65] as seen in Figure 6 [72], and the other leverages energy envelope approaches [9,66] as seen in Figure 6 [73]. The former has a maximum daily energy level set to 12 spoons, while the latter uses a less conventional and finer-grain maximum daily amount of 1000 energy units. Both apps provide predefined and customizable activity lists with predefined energy impacts for each of the listed activities, as well as gamification in the form of a limited-quantity source [74] to track and account for energy as a resource. Additionally, they both use traffic light–style red and green coding to visualize energy usage as positive and negative [75], and to deliver cautionary feedback as a consequence of extending one’s energy beyond the daily threshold amount.

Planning daily activities based on one’s energy can support not just pacing, but also structure and order, or routinization, with previous findings indicating its benefits for people living with CFS [9]. Routinization could be further supported by automatically prefilling in activity fields on apps to further reduce the user’s burden. This is important as a previous study has shown that fatigue symptoms can impede user engagement with self-management apps [9].

Our findings show that 6 of the apps offering interventions such as psychoeducation, neuroeducation, energy estimation, and biofeedback had provided either theoretical or empirical groundings of their methods and informed the user either through app help pages or the app’s website. From these, 4 of the apps cite academic work to support the efficacy of the intervention that they provide neuroeducation (Gupta Program Brain Retraining [Harley Street Solutions LTD]); pacing (Visible: Pacing for illness [Visible Health Inc.]), an energy estimation app; biofeedback (Juva Stress and Migraine [Juva Health Inc.]); and HRV-based interventions (Elite HRV [Elite HRV Inc]), respectively. However, only 2 of these apps (Visible: Pacing for illness [Visible Health Inc.] and Gupta Program Brain Retraining [Harley Street Solutions LTD]) provide empirical evidence supporting the effectiveness of the app itself. Two other neuroeducation apps (Freeme: ME/CFS and Long COVID [Freeme Health Ltd], and Curable: Chronic Pain Relief [Curable Inc]) cite academic work supporting all or some of the information provided to users in their educational material.

We now report on data from app privacy policies and cookie notices through 3 main identified themes, namely the target age groups of the apps, information surrounding the handling of data, and the usage and purpose of cookies (Multimedia Appendices 12 and 13).

Findings reveal that the majority of the apps (10 apps) restricted their use to nonadult individuals. From these, 7 apps mention that they do not provide services to users younger than 18 years, 2 apps do the same for users younger than 16 years, and 1 app restricted nonadult users unless the legal guardian or carer provides consent.

Findings indicate that for 11 apps, their developers disclosed that they stored data gathered through the app on company servers, while 1 app stored it with a third-party site called Sentry Inc. The latter has stated in its privacy policy that this is done only for anonymized crash data.

As the apps store their data outside the user’s device, it is important to reveal the parties with whom the data are being shared. Findings show that via the use of the app, users consent to sharing data with advertisers for 4 apps, 2 of which have revealed that personal data will be shared. Other parties with whom data will be shared in this manner include analytics (3 apps), anonymous crash data processing as mentioned earlier (1 app), and research (2 apps), but in an anonymized form.

Outside of the previously mentioned data, some apps also offered the option to opt-in to third-party data sharing with advertisers (1 app), analytics (1 app), the OS’s health app (1 app), third-party backup storage providers (1 app), and 5 apps present the option to share data with researchers, of which 2 apps claimed that the data will be anonymized, and a third app revealed that the data will be used for product development and health research. It is worth noting that opting in with advertisers, the OS’s health app, and backup storage providers involves the sharing of personal data.

Another important finding is the information surrounding the sale of user data; 3 of the apps were found to sell data, one of which revealed that they only sell personal information with the user’s permission. Seven other apps offered no information surrounding the sale of data, and the remaining 6 apps revealed that they do not sell. Regarding the latter, 4 of them explicitly stated no, 1 app claimed that they do not sell the information covered by the policy, and the final app revealed that it does not sell data from Californian residents. Meanwhile, 2 of the 3 apps selling data informed that the data were being sold to advertisers (2 apps) and analytics providers (1 app). Data surrounding the sale of information were mostly easy to find, with the majority (6 apps) highlighting the information under clearly labeled headings and subheadings, while 2 apps detailed that they do not sell it without highlighting or obfuscating the information. However, one app that sold personal data did not state it on the main page of the privacy policy and was instead presented on a separate page for Californian and US residents. While this page was linked in the privacy policy, the document did not point to the information regarding the sales being there and thus appears hidden to users outside the United States, as they may not explore sections irrelevant to them. This emphasizes the importance of clearly highlighting information regarding sales to users through the use of subheadings to improve discoverability.

Regarding ownership of data, only 1 app’s privacy policy offered information surrounding this, and it revealed that the developer is the data owner.

We also analyzed the app’s privacy policies and cookie notices for information surrounding cookies and have found that 8 apps used cookies for analytics purposes, with one of the apps only enabling cookies outside of app and website areas requiring a log-in to use, to protect their users’ health data. A total of 6 other apps stated that they made use of cookies for both advertisement and analytics, while the remaining 3 provided no information surrounding the use of cookies.