Chronic obstructive pulmonary disease (COPD) is a common, costly, and incurable respiratory disease affecting 1.2 million people in the United Kingdom alone. Annually, it costs the National Health Service (NHS) £1.9 billion (US $2.43 billion), it requires over 1 million inpatient bed days due to acute exacerbations requiring hospitalization, and it places significant demands on health services [1-3]. The prevalence of COPD in poor, remote, and rural populations is twice that of cities [4].

Effective COPD self-management can reduce both exacerbation-induced hospital admissions and health service use when compared to standard care [5,6]. myCOPD is a digital health self-management technology designed to improve COPD outcomes and mitigate demands on services [7]. myCOPD modules include symptom scoring, inhaler technique, and a virtual pulmonary rehabilitation course. Previous studies indicate that myCOPD is associated with reduced inhaler technique errors, lower COPD Assessment Test (CAT) scores, and fewer hospital readmissions within 3 months of an exacerbation [8-11].

NHS Highland covers the largest geographical area of Scotland, contains regions of significant socioeconomic deprivation, and has a majority remote and rural population. Access to hospital services and delivering equity of care remains challenging, and digital health technologies represent one potential solution. In this pragmatic test-of-change study, we evaluated myCOPD and its effectiveness at reducing hospitalizations, inpatient bed days, and other health service use.

Of the 140 people invited to enroll in myCOPD, 20 (14.3%) declined to participate, mostly for technology-related reasons. Of the remaining 120, 7 (5.8%) were excluded, as they died during the study period, leaving a total of 113 participants (Figure 1). The average participant age was 69.3 (SD 8.2) years, and 51.3% (58/113) were female. A total of 70.8% (80/113) of participants were from remote and rural areas, and 75.2% (85/113) represented the three most deprived SIMD quintiles. Most participants (69/113, 61.1%) had moderate or severe COPD, and 20.4% (23/113) had very severe disease according to their Global Initiative for Chronic Obstructive Lung Disease score (Table 1) [14].

A total of 89 out of 113 (78.8%) participants activated myCOPD, with 56% (50/89) of them doing so on the day of enrollment and 90% (80/89) doing so within 1 month (Figure 2, A). Most active participants used at least one module and entered their symptom scores at least once (Figure 2, B and C; n=79, 89%). A total of 10 (11%) participants activated myCOPD but used no modules. Overall, 57% (n=51) of active participants recorded their CAT score one or more times, 39% (n=35) initiated pulmonary rehabilitation training, 24% (n=21) viewed educational course material, and 10% (n=9) watched at least one inhaler technique video. Out of 89 participants, 15 (17%) were very high users based on symptom scoring frequency (Figure 2, C), and 14 of these (93%) also used four or five myCOPD modules (Figure 2, B), suggesting a discrete subgroup of highly engaged users. There was no overall correlation between myCOPD engagement and participant age, SIMD status, or rurality (Multimedia Appendices 1 and 2).

To evaluate myCOPD effectiveness, we quantified the daily incidence of inpatient bed days, hospital admissions, home visits, clinic appointments, and out-of-hours care provision for an average of 375 (SD 32) days before and 239 (SD 46) days after myCOPD activation, for a total of 69,211 participant days (47,972 days before and 21,239 days after). There were no significant differences in hospital admissions, inpatient bed days, or other health service use before or after myCOPD activation, apart from a modest increase in home visits consistent with previous telemonitoring studies [15] (Figure 3, A). Even after excluding participants who did not activate their license, there remained no significant difference for any of the categories.

Subgroup analysis results can be seen in Multimedia Appendix 3, A-H. Although underpowered, subgroup analysis based on either module usage (Figure 3, B; Multimedia Appendix 3, E) or symptom scoring frequency (Figure 3, C; Multimedia Appendix 3, F) did identify trends toward reduced inpatient bed days and hospital admissions for highly engaged users. There were also increased home visits in all subgroups after myCOPD activation regardless of module usage or symptom scoring frequency (Multimedia Appendix 3, A and B). No other trends in health service use were observed based on subgroup analysis regarding clinic appointments (Multimedia Appendix 3, C and D) and out-of-hours contacts (Multimedia Appendix 3, G and H).

This study is the longest and largest evaluation of the digital health self-management technology myCOPD to date, the only one involving a predominantly remote and rural population, and the first to recruit patients from within community care settings using a pragmatic approach. Enrollment and engagement with myCOPD was popular, with 78.8% (89/113) of participants activating the technology and 89% (79/89) of these participants using at least one module or entering their symptom scores at least once. Only 14.3% (20/140) of people approached declined to participate in the study, and there was no correlation between participant enrollment, activation, or engagement and either age, socioeconomics, rurality, or disease severity, suggesting that these are not significant barriers to using myCOPD. This finding may help mitigate perceived risks of increased health inequalities associated with the use of digital health technologies as part of routine care provision.

Despite high levels of activation, myCOPD did not reduce overall demands on health services. These findings are consistent with the limited evidence supporting the use of COPD self-management technologies, but they contrast with previous studies involving myCOPD [8-11]. There are several possible explanations for this difference. First, our study involved community-based recruitment of stable patients with COPD irrespective of exacerbation frequency, whereas other trials recruited hospital-based patients immediately following an acute exacerbation where motivation to engage in self-management may be greater. Second, previous studies collected data for only 90 days and, therefore, evaluated acute rather than long-term myCOPD benefits [9]. Finally, it remains possible that cultural or socioeconomic differences between rural and urban participants might influence myCOPD engagement and impact. Our results highlight the need for further evaluation of myCOPD and other digital health technologies ahead of their widespread procurement and adoption as part of routine health services. It may be that myCOPD can function as an effective tool in reducing COPD exacerbations when offered to participants in hospital and at a time of crisis, whereas it may not function in this manner when offered to patients in the community who are not actively in crisis or experiencing an exacerbation.

Despite no overall reduction in health service use, we did observe trends toward reduced hospital admissions and inpatient bed days in a subgroup of highly engaged users. This suggests the technology may be clinically beneficial if it is highly used and suggests that a greater emphasis is needed for understanding the motivation to use digital self-management tools and how to promote increased, meaningful user engagement. Paradoxically, previous studies indicate that those patients who may benefit most from engaging with digital self-management technologies are the least likely to do so [16]. Our observation that highly engaged myCOPD users were indistinguishable in terms of age, socioeconomics, rurality, or disease severity suggests that the factors driving meaningful user engagement are complex and require further attention. This will necessitate increased collaboration among a wide group of stakeholders, including patients, throughout all stages of digital health technology design, development, and testing.

One potential limitation of this study involves differences in the amount of data we collected before versus after myCOPD activation (47,972 participant days before and 21,239 after). Our original design involved collecting an equivalent quantity of data before and after myCOPD activation, but a decision was made to terminate the study on March 1, 2020, due to the emergent COVID-19 pandemic. We mitigated the impact of this change by evaluating data according to daily rather than annual individual health service usage. Interestingly, and despite cessation of formal data collection, we observed increased myCOPD engagement among many participants after March 2020, which may reflect changes in health behavior when access to face-to-face services was limited. A further impact of the COVID-19 pandemic was our limited ability to evaluate the effect of seasonality on exacerbation frequency, and it is possible that the inclusion of data beyond March 2020 may have resulted in different outcomes.

In conclusion, although our study does not support implementation of myCOPD to reduce health service demands on NHS Highland on a population-wide basis, our results do indicate that some highly engaged patients may derive benefits. Thus, individuals can be encouraged to individually adopt myCOPD as part of their self-management care should they find it useful. Further research is needed to understand what motivates some individuals to engage with digital health technologies, in order to facilitate the design and development of clinically and economically effective self-management tools.