An estimated 65 million people worldwide and 1.2 million people in the United Kingdom alone [1] have chronic obstructive pulmonary disease (COPD). COPD is projected to become the third leading cause of death by 2030 [2,3]. COPD entails a significant personal, economic, and societal burden [4].

COPD patients are predominately older adults, (ex)smokers, and from lower socioeconomic backgrounds [5]. Strategies are needed to help patients manage and monitor their condition over time and between medical assessments to ensure effective long-term management [6]. Importantly, COPD tends to occur alongside various comorbidities (such as coronary heart disease, lung cancer, anxiety, depression, and osteoporosis) due to shared risk factors (eg, aging and smoking) and shared underlying pathophysiological mechanisms [7]. This increases the disease burden, worsens patients’ prognosis, and leads to increased health care costs [7].

The Department of Health in the United Kingdom has issued recommendations for the prescription of apps as part of the care strategy for various long-term conditions, such as COPD [8]. As COPD patients may not recall the small daily fluctuations in their lung symptoms due to the highly symptomatic nature of the disease, telemonitoring via mobile health (mHealth) technologies may facilitate early intervention through daily monitoring of symptoms [2].

mHealth technologies have been shown to reduce costs associated with long-term COPD management [2]. Qualitative insights show that using mHealth to complement regular care is acceptable to both COPD patients and their health care professionals [9]. The advantages of Web-based health interventions include cost-effectiveness, round-the-clock availability, customizability to personal preferences, and anonymity (when compared with face-to-face interactions) [10]. However, mHealth apps also entail issues and risks, such as a lack of quality control and lack of evaluations of their effectiveness, and privacy and security risks [10-12]. A systematic review of mobile apps used for self-management of chronic conditions concluded that apps can potentially improve health outcomes in long-term conditions through improved symptom management [13]. A systematic review and meta-analysis on randomized controlled trials (RCTs) assessing mobile apps for COPD self-management found evidence for a lower risk of hospital admissions among app users as compared with usual care [14]. A further systematic review and meta-analysis of RCTs found significant improvements in health-related quality of life across 557 COPD patients who used smart technology compared with face-to-face or written support [15]. However, the evidence stemmed from only three studies and was deemed to be of poor quality [15]. Moreover, a review of relevant literature on apps as well as apps available in app stores for people with COPD identified a scarcity of published literature regarding the effectiveness of the apps [2]. There is a clear need for quality-controlled, effective, and acceptable health apps if mHealth is to play a significant role in efforts to prevent and manage diseases.

This study aimed to provide in-depth insights into the views of people with COPD and their caregivers regarding the use of apps in COPD, highlighting key topics and issues around usability that need to be taken into consideration during app development.

This study formed part of the large, multistakeholder project CityVerve. The CityVerve initiative resulted in the city of Manchester receiving Innovate UK funding for a 2-year period to work on a range of initiatives to apply technology to four separate use cases in partnership with the National Health Service, industry, and universities. The use cases comprised health and social care, transport and travel, culture, and energy and environment. One of the focuses of the health and social care case was COPD. The use of mHealth with this demographic was considered, and a mobile app was developed using a co-design approach and prototyping with participants who were diagnosed with COPD. Project members included academic researchers, software engineers, and health care professionals.

The approach taken in this study is similar to the iterative convergent mixed-methods design proposed by Alwashmi et al [16]. The authors propose the use of a mixed methods framework in which both qualitative and quantitative data collection and analysis are used in iterations to develop mHealth interventions and enhance the usability of such interventions [16].

Ethical approval for the study was granted by the University of Manchester (2017-2941-4477). An overview of the main phases can be seen in Textbox 1.

Development and evaluation of the app took place during the three main phases. The first phase consisted of patient and public involvement (PPI) work, leading to the development of mock-up app designs and subsequently a prototype app. PPI work was carried out by attending and hosting a number of events for COPD patients and their families. We also invited members of a COPD self-help group to attend several sessions to discuss their needs and preferences. Together with participants, we developed paper prototypes for the app. The main stakeholders involved in the project were patients, families, caregivers, and members of the CityVerve project team. As this phase constituted PPI work rather than research, sessions were not audio recorded. Notes were taken following discussion and observation of the participants.

The second phase involved a usability study with the prototype app, using the “think aloud” paradigm. Participants were instructed to set up the app, browse its features, and verbalize their thoughts about the app as they proceeded. Participants then completed the System Usability Scale (SUS) questionnaire [17]. The “think aloud” sessions were audio recorded and transcribed. In addition, 2 researchers attended each session to observe and record notes. Transcripts and observation notes were subsequently analyzed thematically using framework analysis [18]. This is a qualitative method where themes are represented in columns and participants in rows, with participants’ responses summarized in the associated theme’s column (Multimedia Appendix 1). This facilitates comparison across participants for the respective themes. All data were coded by 1 researcher, and a random subselection of quotations was coded by a second independent researcher to test for coding agreement between the reviewers (interrater reliability). Cohen kappa indicated substantial agreement (84.6% agreement; κ=0.75).

The third phase involved deploying the app on the Google Play store for a 4-month period to collect the self-reported symptom data and event data to evaluate the app over a longer period with a wider group of the target population. The app was advertised by disseminating a 150-word summary of the study (or in the case of Twitter, a 220-character summary) to relevant groups on social media, various mailing lists, websites, and newsletters, with a link to the app’s playstore page. For mailing lists, newsletters, websites, and Twitter, we approached the relevant admin who sent the information out or posted it to their website and Twitter account. For Facebook groups, we first approached the admin to obtain permission before posting. Targeted groups were either COPD-related (eg, COPD support groups) or likely to include an older demographic (eg, aging-related mailing lists) because COPD mostly affects older adults [19].

The app (Figure 1) was developed as a hybrid Web app using Apache Cordova and the Ionic framework and was designed to exploit the experience sampling method (ESM), which allows for the daily submission of data rather than retrospective completion, leading to potential memory bias [20]. Once downloaded, the app takes users through several setup screens. These screens collect some one-time data about the user (Textbox 2).

The main menu page displayed the current weather data, including temperature, humidity, and wind speed, as seen in the gray box in the first screen (home page) in Figure 1. The My symptoms page allowed users to record their general well-being, measured on a 3-point scale (great, so-so, and bad). Following this, they entered information about five key symptoms of COPD (breathlessness, coughing, mucus, tiredness, and sleep quality on a 4-point scale) and their medication use (Figure 2). The scales used to assess symptoms and well-being were adapted from the Britain Breathing app [21].

Participants could view their self-reported ratings for each symptom in the form of graphs presented on the My data page (Figure 1, center). This is accessed by tapping the View my symptoms menu option on the home page (Figure 1, left). Educational material about COPD was presented on the what is COPD? page. The About page contained information about the study and the participant information sheet. Each page had a help button displayed in the top right-hand corner, which included a video with an audio description of how to use that page.

Overall, users were enthusiastic about the concept of using an app to manage and monitor their condition. They welcomed the idea of being able to easily share information with family members, caregivers, and health care professionals, especially as several participants described difficulties communicating with doctors. Participants saw a value in the app for COPD patients in terms of recording, viewing, and monitoring symptoms, monitoring medication, and providing information about weather and characteristics of different locations to enable planning of activities. Figure 3 shows an example paper prototype developed with users.

On the basis of phase 1, we developed a prototype that was tested with 5 people (4 females and 1 male) in the usability study. This included 4 people, who self-identified as having COPD, and 1 caregiver. The participants comprised 5 volunteers (4 females, all aged ≥55 years), and all of them had clinician-diagnosed COPD for more than 5 years. Participants were recruited through a respiratory patient support group and a general practice patient forum, and all of them had prior experience of being members of research advisory or stakeholder groups; hence, they were familiar with medical terminology. All but 1 participant used a smartphone at least occasionally. Moreover, 1 participant stated that she did not usually use a smartphone. Participants gave an average score of 32.4 out of 55 (SD 8.47) on the SUS, suggesting that they rated its usability as poor. The results of the qualitative analysis of the audio transcriptions and observational notes reveal the main issues faced by users. The main themes identified are shown in Figure 4. The number in brackets indicates the number of times that theme was mentioned.

We found that many of the issues affecting our work with participants were similar to those faced by O'Connor et al [25], who attempted to co-design an app with dementia patients and their caregivers, including unfamiliarity with technology and incorrect perceptions about how users would interact with the technology. This led to the initial design ideas being significantly revised. So far, very little qualitative research regarding the perception of patients and their caregivers and how they experience technology has been carried out [26]. The findings of our study reinforce the importance of considering the unique users of the technology we are proposing. When working with people with chronic diseases, their age, education, and Information Technology literacy should be taken into account [2]. This is especially relevant for the COPD population.

Many design features commonly used in mobile apps are hitherto unknown to the COPD population, such as the sliders that were initially included in the app. Our research indicates that such design elements are not optimal for a COPD population. Previous research corroborates this by showing that older users require larger touch targets (minimum 15×15 mm) with sufficiently large gaps between touch targets (minimum 6 mm), ideally complemented through support functions such as speech input [27].

Besides the physical impairments, our study highlights the importance of considering lack of familiarity and confidence in technology use. Our observations as well as participants’ verbalizations while navigating the app suggest that older users may require more help functions within apps that provide support regarding app usage as well as general smartphone usage (eg, how to tap items). Research shows that among those older than 65 years, the majority do not feel confident in using computers, smartphones, and other electronic devices [28].

We used ESM to capture patients’ symptom information, as it has been shown to provide an adaptive and personalized system for the monitoring and adaptation of treatment strategies as well as increasing ecological validity and reducing memory biases [20,29]. ESM has also proven to be a useful method in other conditions where the impact of symptoms varies over time [21]. Information about COPD was also included in the app. This was introduced following input from patients attending lung events, before app construction, as several patients highlighted a lack of information from their general practitioner about the condition.

Our study highlights that physical impairments and lack of familiarity can particularly affect initial registration processes, such as creating usernames and passwords. This step is crucial for health-related apps, as password protection is needed to protect sensitive data. On the basis of our findings, we recommend that this process be kept brief and simple, for example, by using a simple 4-digit pin. Registration pages should also be supported through a feedback system to identify errors immediately and reduce user frustration [27].

Given the demographic qualities of the COPD population, the development of apps for this target population may be reticent to present information in the form of graphs; however, our research suggests that graphs can be interpreted correctly and perceived as useful if they are presented in a simple, clear format. Previous research confirms that graphics and multimedia should be used sparingly and purposefully when targeting older users, and text alternatives should be provided where possible [27].

Another important insight from our work reflects on the validity of user feedback. Features that users appeared to struggle with according to observation notes were nevertheless described as fine by users, suggesting a social desirability bias [30]. This underscores the importance of developing apps not only based on users’ verbalizations and responses to specific questions (such as standardized usability scales) but on observations of user behavior while engaging with the app. This limitation has been noted in previous usability evaluations [31,32] and requires due consideration during app development.

Attrition rates are a known issue in digital health interventions [33]. The high level of attrition seen in our study is, therefore, not entirely unexpected. An understanding of why people discontinue use is important and worthy of further research given that retention is key for the management of chronic conditions over time [33]. This is especially challenging as chronically ill people may experience diary fatigue and be unlikely to keep accurate records of their condition, especially when unwell [34,35]. Another factor that may have contributed to the lack of use is that the reminder notification did not work on all devices due to an issue with the local notifications plugin path in a package of the Ionic framework.

The usage data indicate that the app was used as intended with most of the activity surrounding the recording and viewing of symptom data, which was the app’s principal purpose. The low usage of the help feature may suggest that the app was also fairly intuitive for most people. The biggest issue surrounds the lack of continued use of the app over time. More research is needed to identify the intrinsic and extrinsic reasons behind the attrition in this context. Longer-term analysis of app usage before dropout using in the wild (deployed in the real world) event capture may help to shed light on some of these reasons.

As a result of the usability work carried out with participants, we can infer several considerations for designing COPD support apps, including the following:

The implications specific to those with COPD include:

The principal recommendation based on our experience of working with a challenging target population on a large project with multiple stakeholders is that the population of interest, researchers, and developers should all be involved at every stage of the project and that an iterative approach be used to build a prototype. This prototype should then be tested in a wider environment with a larger group of the target population, where their interaction with the app is evaluated over a longer period to determine further issues and acceptability of the final intervention. The main stages are summarized in Figure 10.

Our PPI work involved participants from COPD support groups who expressed an interest in building a smartphone app. Such participants are more likely to be educated about COPD and may have more familiarity with technology than the standard COPD user. We attempted to mitigate this by capturing patterns of event behavior (ie button presses and transitions) from a wider group of users post deployment. There is an argument that such interventions in general do not adequately address the digital divide, as only those with access to such interventions in the first place may benefit. This rules out many older and low-income users [36]. Moreover, as our usability study was conducted with individuals who had a COPD diagnoses of more than 5 years’ duration, findings may not adequately represent those who were recently diagnosed. In addition, the time of year is likely to have had an impact on the reported symptom frequency and severity. Finally, it should be noted that the 3-point scale for well-being used in our app is not a validated measure. It was developed based on user engagement and has high face validity, but it is not clear how well it correlates with actual well-being.

We found that working with members of the target population at all stages of the project was a useful strategy; stakeholder engagement aids the development of research interventions that are both adaptive to the needs of the patient and the preferences of the provider [37]. This is different from a traditional researcher-led approach or a pure software engineering approach, such as agile. Placing the target population at the center of the work and iteratively building the intervention with the target users allows for the creation of a more acceptable final product. The combination of qualitative data analysis and data collected from open-source event capture tools also served to offer a further insight into app usage and dropout. We must try where possible to mitigate bias when carrying out such work, including social desirability bias and biases associated with participants who display more familiarity with technology than the typical target end user.