JMIR Publications

ABSTRACT

Background:

Commencing in 2019, a pandemic caused by a novel coronavirus disease (COVID-19) spread across the world [1, 2]. This pandemic forced the implementation of remote monitoring capabilities and helped shape the World Health Organisation's (WHO) global strategy on digital health 2020-2025 [1]. Furthermore, it forced the world into periods of isolation and lockdown demonstrating the need for remote monitoring capabilities to aid case prioritisation and timely intervention and to ensure continuing high-quality patient care [3]. Access to healthcare services remotely can reduce the burden of in-person healthcare services and the economic and environmental costs of hospitalisation, transportation, and exposure to in-hospital infectious disease [4, 5]. However, remote monitoring of respiratory diseases cannot be widely adopted before determining its modes, feasibility, usability, and acceptability to patients. Respiratory diseases comprise a diverse spectrum of different conditions affecting all ages with varied symptoms and prognoses [1]. The increasing incidence of respiratory disease and high mortality are global issues [4, 6, 7]. In 2017, chronic respiratory diseases affected approximately 544.9 million people worldwide [6, 8, 9]. The actual costs and the long-term outcomes of patients with respiratory disease are challenging to predict due to the varied disease trajectories that individuals experience [10]. Asthma + Lung UK reported an estimated 12.7 million people with respiratory disease in the UK. Of these patients, 1.2 million were diagnosed with chronic obstructive pulmonary disease (COPD), which is the third leading cause of death worldwide [9], and more than 150,000 were diagnosed with interstitial lung disease (ILD). According to a recent Asthma + Lung UK report, the UK spends £11 billion on respiratory diseases each year, with 29% of that budget allocated to COPD [4, 11]. Recent studies demonstrated the importance of timely identification of exacerbations of COPD [12, 13]; therefore, longitudinal measurement of symptoms and physiological parameters have the potential to allow earlier detection [14].There is presently an unmet need in the care of patients with respiratory diseases [11, 15].Chronic respiratory diseases and those in the post-discharge period following hospitalisation due to COVID-19 are particularly vulnerable, and little is known about the changes in their symptoms and physiological parameters [16, 17]. New modalities of remote data collection, such as home spirometers, wearables, pulse-oximeters, and smartphone apps may provide the opportunity to improve self-management and offer better, more timely information for clinical assessment. Remote monitoring may help to bridge the gap between hospital and home for these patients [18]. However, questions remain about the feasibility and acceptability of remote monitoring of physiology and symptoms for patients with respiratory diseases.

Objective:

The ultimate goal of remote monitoring is to provide practical healthcare to people with respiratory disease, facilitate community-based self-management, support early exacerbation detection and reduce hospitalisation [19, 20]. In this prospective cohort study, we sought to gain a better understanding of how well patients with chronic respiratory diseases engage with remote monitoring system. We evaluate the feasibility, adherence, engagement, retention, acceptability, and usability of remote monitoring of respiratory disease using commercially available wearables (for heart rate, physical activity, and oxygen saturation), spirometry, and questionnaires. We hypothesised that remote monitoring using a finger oximeter, wearables, spirometry, and smartphone apps would be feasible, acceptable, and usable in patients with respiratory diseases, including COPD, ILD, and those affected by COVID-19 infection.

Methods:

Study design This study is a 3-arm prospective observational cohort study that evaluated the feasibility of remotely monitoring physiological parameters and symptoms via a full-scale comprehensive remote monitoring system (RADAR-base platform) over 6 months [21]. Patients were recruited from the Royal Free London Hospital and University College London Hospital (United Kingdom). The RALPMH detailed protocol has been published recently and can be accessed online [21]. The study was registered with ISRCTN: 16275601Ethics approval was given and patients provided written informed consent. Data collection This study used the RADAR-base platform, a comprehensive remote monitoring system created to provide high-quality, clinically reliable, real-time, and relevant data for both patients and clinicians [22]. The RADAR-base platform is an open-source mHeath platform used in previous studies such as RADAR-CNS and MDD [23, 24]. Data was collected using 4 main components. Firstly, REDCap was programmed to automatically send periodic questionnaires including demographic questionnaires (Date of birth, sex, ethnicity, height, weight, smoking history) and medical history and other health-related questionnaires [24]. Secondly, a regimen of questionnaires and speech and vocalisation sampling was administered using RADAR-base active mobile phone app. Thirdly, data were collected from a number of devices: continuous passive monitoring with the Garmin Vivoactive 4, and active monitoring outcomes via daily finger oximeter. Daily lung function measurements via Nuvoair spirometer in the COPD and ILD cohorts. An innovative approach to boost retention was piloted in this study: Protocol Holidays were offered to patients who wanted to drop out midway through the study with an option of returning at a chosen date. Outcome We evaluated feasibility using a range of participant metrics: adherence, engagement, retention, acceptability, and usability. Further details of the outcomes’ measures are found in the supplemental material Table 1. Analysis Adherence was assessed using final adherence rates for various data types, with slight variations in the methods depending on the data source. In some cases, our study protocol allowed for protocol holidays, which required removing gaps in data and combining the remaining data to add up to a total 180-day study period. Adherence rates are visually represented using bar charts with confidence intervals and Kernel Density Estimation plots. For time series data, the adherence rate was calculated using the formula: Adherence rate (% )= (No.of actual data points in a time window T)/(No.of expected data points in the time window T) × 100 Questionnaire data were evaluated by comparing the expected number of questionnaires based on the protocol to the actual number provided by the patient as a percentage. For Garmin wearable data, the adherence rate was calculated based on aggregation over hourly windows, and data with negative values (signifying the metric could not be calculated by Garmin) were marked as missing. Spirometry adherence rates were calculated based on the frequency of recordings per week. The ATS grading system provided with spirometry data was used to assess the quality and usability of measurements. These analyses enabled us to visualise the optimal protocol frequency for spirometry to limit burden whilst ensuring useful data. The adherence was also assessed based on the burden on the patient by grouping the results by the burden of data collection of different data types. These include: “active_questionnaire_light”: Light burden questionnaire tasks like PROMs. “active_questionnaire_heavy ”: Heavy burden active tasks such as recording on finger pulse oximeter or providing audio recordings. “spirometry”: The spirometry task using the provided smart spirometer “passive_wearable”: The collection from the Garmin wearable device without active user involvement. Engagement contiguity: We define engagement as a measure of contiguity of data i.e. the extent to which data is collected without gaps. The level of adherence is divided by the number of days to calculate engagement, which provides an indication of how consistently and continuously patients are contributing data to the study. These are represented as time series heatmaps to view the overall engagement and grouped by the burden of data contribution on patients. Hierarchal clustering was performed to understand the patterns of patient engagement. Engagement with the wearable was evaluated by using the Garmin wearable wear time which was calculated using the availability of heart rate data. If heart rate data was present, the device was considered worn at those times. This is an estimation and not a perfect approach to ascertain wear time as the device might not calculate heart rate in some instances, e.g. when the fit of the device strap is not correct, even though it is being worn. Responsiveness: A second measure of engagement looked at the time-to-respond to prompting by mobile phone notification [25]. Acceptability At the end of the study, subjects were emailed three questionnaires: Technology Assessment Measurement Fast Form (TAMFF), acceptability, and satisfaction questionnaires. Adverse events and safety (e.g. reported adverse events and encountered problems during the study were recorded in the REDCap log). Retention of patients in the study was evaluated with Cox regression proportional hazards analysis to calculate the time-to-event, the event being the drop-out of a patient. Kaplan Meier analysis [26] was used to calculate probabilities of retention at time points from enrolment dates. Kaplan Meier plots are grouped based on cohorts, and data types based on burden and protocol holidays with group differences assessed using the Log Rank test. Two ways of calculating observation periods and the definition of an event including or excluding protocol holidays: The first one includes periods with protocol holidays as missing data. An event date in this case is defined as the last data point for the patient. The second period excludes protocol holidays, and the start of the first protocol holiday is considered as the patient's exit from the study. This is done in order to simulate the retention in case no protocol holiday option was provided in the study (assuming patients would have otherwise dropped out).

Results:

Patient characteristics A total of 162 patients were assessed for eligibility from July 2021 through November 2022, (Figure 1). Sixty recruited patients (ILD n=20, COPD n=20, post-COVID-19 n=20), had a retention rate of 55(92%) across all three cohorts at 6 months. Five patients dropped out of the study: one patient died during the first week of the study, and another dropped out as they were feeling unwell. Two patients found the study protocol too time-consuming. The last patient dropped out due to privacy concerns. Figure 1. Flow chart of patients screened. The sixty patients included in the analysis had a median age of 64 (IQR=36-82) years, with slightly less than half being female (41%). The ILD cohort 11(55%) was more likely to be female than the COPD 25 (41%) and COVID-19 5 (25%) cohorts. Other general characteristics are reported in Table 1. Table 1: Summary of baseline characteristics of patients in the RALPMH (N= 60). Patient adherence The adherence rate with 95 percent confidence intervals and kernel density estimations [27] of adherence rate based on questionnaire frequency are shown in Figure 2. When looking at adherence of data grouped by frequency of questionnaires, the questionnaires that were issued with low frequency had highest adherence with adherence reducing as the frequency of questionnaires increased in all the 3 cohorts. There were some exceptions to this trend like the audio questionnaires and spirometry due to the high burden of the task. The “pulse_ox” questionnaire is another high-burden task which required users to measure the pulse rate and SpO2 on a finger pulse oximeter device and manually enter the readings into a questionnaire in the mobile app. Adherence to data from the Garmin wearable device is shown in Figure 2g. Overall adherence was decent between 70-90% for most data types. The pulse ox sensor (SpO2) had lower adherence due to the sensor being more sensitive to motion artefacts causing issues with calculation of SpO2 during the daytime. The ILD cohort had the best adherence to the wearable device followed closely by COPD. The COVID-19 cohort had the lowest adherence. a b Figure 2: Questionnaire adherence rates with 95% confidence intervals (a) and kernel density distributions (b) for each of the ILD cohort. c d Figure 2: Questionnaire adherence rates with 95% confidence intervals (c) and kernel density distributions (d) for each of the COPD cohort. e f Figure 2: Questionnaire adherence rates with 95% confidence intervals (e) and kernel density distributions (f) for each of the COVID cohort. g Figure 2g: Garmin wearable data adherence rates with 95% confidence intervals. Adherence to home spirometry Patients with ILD were found to be more adherent to weekly home spirometry than patients with COPD, with adherence rates of 94% and 84%, respectively (Figure 3a). 3836 (90%) of sessions were of acceptable quality according to ATS grading (grades A-E). The majority were considered Grade A (781; 18%) or Grade B (1926; 45%). Other grades were C (175; 4%), D (130; 2%), and E (824;15%). There were only (412; 8%) that were found unacceptable and mostly from the same group of patients. (Figure 3b). a b Figure 3. a Adherence to home spirometry b Quality of home spirometry Patient engagement Figure 4 illustrates patient engagement with wearables, active questionnaires, and spirometry using hierarchical clustering and grouped by the burden of data collection on users. The lighter colour represents a higher usage rate, whereas darker colours reflect a lower usage rate. Each row represents a user, and the right y-axis shows the user IDs. Clusters of patients with similar engagement are shown closer to each other on the plots. The expected maximum daily use of wearables is approximately 24 hours due to the time needed to charge the device each day (as instructed). The average daily use of wearables varied between patients and ranged from 7.72 hours a day to 23.58 hours a day. The average use across all users was 18.66 ± 4.69 hours daily, 123.91 ± 33.73 hours (5.16 ± 1.40 days) weekly, and 463.82 ± 156.70 hours (19.33 ± 6.53 days) monthly. Figure 4: Hierarchal clustering heatmaps showing patterns of patient engagement at timepoints from their enrolment dates for 3 cohorts (top to bottom). The plots are divided by the burden of data collection on users (high to low from left to right). COVID-19 cohort did not have spirometry in the protocol. In order to gain a deeper understanding of participant engagement, we examined the time between notification and completion of questionnaires (Figure 5). To facilitate analysis, we grouped participants into two categories - those who took protocol holidays and those who did not. We observed that overall participants who took holidays had faster response times compared to participants who did not opt for holidays. A similar trend was observed in the COVID-19 cohort which had faster response times to the questionnaires compared to the ILD and COPD cohorts. a b Figure 5: The time to answer a questionnaire which is defined as the duration between the notification of a questionnaire and the completion of the questionnaire. From top to bottom: with respect to time to answer a) Looking at the daily study-specific questionnaires, b) Looking at audio questionnaires across the three cohorts. Participant acceptability A high percentage of patients surveyed had a favourable perception of the device's ease of use and were willing to continue using the device, with 72% stating they would use it both day and night and 82% showing a willingness to use it overnight. (Figure 6a). In addition, the technology assessment measurement fast form (TAMFF) was utilised to evaluate patients' perceptions of acceptance, usefulness, satisfaction, and ease of use. The majority of patients found the RADAR active app efficient (74%), performance-enhancing (63%), productivity-increasing (64%), effective (79%), helpful (79%), and useful (83%). The majority of patients stated that they would likely choose the RADAR app (77%), likely to use the app (77%), and likely to use the app for future health monitoring (Figure 6b). Finally, the majority of patients found daily home spirometry with NuvoAir Spirometer to be highly acceptable, with high levels of agreement regarding acceptability, usability, and satisfaction (Figure 6c). a b c Figure 6. a Patients' acceptability of remote monitoring. b Technology acceptance model fast form. c Acceptability of daily home spirometry Patient retention Survival analysis was performed to understand user retention in the study. The data was right censored with an observation period of 180 days (the protocol period) and the event was taken as the last data point contributed by the patient in the study. Figure 7a shows the retention of patients in the 3 cohorts. Overall, 31 out of 57 patients (54.39%) had a retention of a full 180 days and 46 out of 57 patients (80.70%) had a retention of more than 150 days. This includes patients and data with protocol holidays. Figure 7b shows the survival plot for the 3 cohorts. Even though all three cohorts had very good retention at 0.5 survival probability of 179.0, 179.8, 178.9 days in ILD, COPD, and COVID-19 cohorts respectively, the COPD cohort had much better retention between 0.8 to 1 survival probability compared to other cohorts. a b Figure 7a. Lifelines of retention for the 3 cohorts. Blue lines represent users that engaged during the full protocol period of 180 days. Orange lines (with markers) show the users that stopped engaging before the 180-day protocol period. These are based on data averaged across all passive and active data sources including questionnaires and tasks in smartphone apps, Garmin passive data and spirometry data. Figure 7b. Kaplan-Meier curves for the 3 cohorts with 95% confidence intervals. These are based on data from all passive and active data sources including questionnaires and tasks in smartphone apps, Garmin passive data and spirometry data. c Figure 7. c Kaplan-Meier curves based on burden or patient effort required of the data source. “active_questionnaire_heavy” contains questionnaires which require an active action from the user such as an audio task or measuring the finger oximeter using a finger pulse ox device and reporting in a questionnaire in the app. “active_questionnaire_light” are normal form type questionnaires which do not require any additional effort. Spirometry can be considered a high-burden task. Protocol Holidays Protocol Holidays (PH) were offered to participants who wanted to drop out of the study as a means to boost retention. To compare the difference in retention between participants requiring PH vs those who did not we compared two groups. Figure 8a shows time-to-last-day Kaplan-Meier survival curves for the two groups of patients: group 1 those who did not take protocol holidays, and group 2 those who were offered and accepted a protocol holiday. The retention at 0.5 survival probabilities was 179.9 and 165.1 days in the groups 1 and 2 respectively. Further to investigate the differences in retention due to PH on each cohort, we plotted the retention curves grouped by cohort in Figure 8c. The ILD and COPD cohorts showed differences between the groups but the COVID cohort did not show an discernible difference. To simulate the effect of the data collection without PH we used the first protocol holiday as the study exit compared to the full non-contiguous data generated over the one or more PHs (Figure 8 b) Here we have 2 groups of data: one inclusive of data after the patient returned to the study from the holiday, and another where we consider the patient as having dropped out at the start of the first protocol holiday, thus marking that as their last day in the study. Note that the patients in both groups are the same; only the selected data differs. a b Figure 8a. Kaplan-Meier curves based on protocol holidays. The left plot shows retention of patients that did not take any protocol holidays while the right plot shows retention of patients that opted to take at least one protocol holiday. Figure 8b. Comparing retention for all patients based on including or excluding data during protocol holidays for calculating the retention time. c Figure 8c. Kaplan-Meier curves based on protocol holidays per cohort. A log rank test was performed to test for differences between the two groups (with and without protocol holidays) shown in Figure 8a. The results are shown in the Table 2a in the material Significant differences between the two groups were found. Log rank test was performed to test for differences between the two groups (with and without protocol holidays) per cohort shown in Figure 8c. The results are shown in the Table 2b in the material. Significant differences between the 2 groups were found in the COPD and ILD cohort while no differences were found in the COVID-19 cohort. Reasons for missing data To understand the reasons for adherence and retention, patients were contacted at various points throughout the study, particularly when missing data or low engagement was observed. The patients reported several reasons for the absence of data. Figure 9 shows the reasons for this missing data, arranged in order of frequency for the ILD cohort. The most commonly reported reasons for data absence were issues with Garmin sync, forgetting to charge the device, and travelling. Figure 9: Reasons for missing data ordered by count in the ILD cohort.

Conclusions:

Discussion To the best of our knowledge, our study is the first to demonstrate the feasibility and acceptability of the most comprehensive real-time remote monitoring program in patients of a variety of lung diseases using commercially available wearables (for heart rate, activity, and SpO2), spirometry, smartphone, and app questionnaires. Despite the COVID-19 pandemic lockdown, and the technical complex protocol, patients showed high levels of retention, adherence, acceptance and reported positive experiences. Patients demonstrated a higher adherence to passively collected data than actively collected data, which is consistent with what has been previously reported [28]. Adherence was inversely associated with the burden or effort required to complete the data. Patients were more likely to adhere to weekly actively collected data rather than daily collection. Patients also showed more adherence towards the app questionnaire over the finger oximeter and spirometry data collection which require extra effort to complete. This can be attributed to the RADAR-base app concept and approach, which was designed to minimise patient burden and prevent inaccuracies in entering manual data. RADAR-base is capable of delivering timely notifications to alert patients when to complete tasks. This feature assisted in minimising data loss encountered in previous studies [22, 29]. Furthermore, the use of real-time data transmission reduces the amount of data loss encountered in remote monitoring [29, 30] by providing a direct notification of any significant events or missing data. Engagement and Retention Figure 4 shows three distinct clusters of patient engagement patterns are discernible. Firstly, there are patients who exhibit high levels of engagement. Secondly, there are patients who display lower levels of engagement. Lastly, there are patients who demonstrate poor adherence across all data types for the majority of the study duration. Additionally, analysis of minor groups or patterns reveals that some patients begin the study with poor engagement but improve over time, while others start off well but become less engaged as the study progresses. The former is primarily observed in the “active_questionnaire_heavy” group across all three cohorts, which includes active audio tasks. This implies that the unique aspects of audio data collection may contribute to this phenomenon, and that training effects may lead to improved adherence later in the study. See supplementary material Figure 3b. Retention analysis time to the last data collected event was performed on data grouped by the level of burden imposed on the patient to complete the data. Spirometry and the questionnaires that require active action from the user (“active_questionnaire_heavy”) are most burdensome while wearable data completion being passive was least burdensome. Figure 7 shows the survival curves based on the burden of the data. As expected, spirometry and active_questionnaire_heavy have least retention at 0.5 survival probability of 176.8 and 179.2 days respectively while also having earlier dropouts when looking at 0.8 survival probabilities. This is followed by active_questionnaire_light and passive_wearable with retention of 180 days at 0.5 survival probability. The drop-off in retention in the passive_wearable data group was slowest when looking at survival probabilities between 1 and 0.8. More detailed plots for each data type are presented in the supplementary material Figure 7. Protocol Holidays We found that implementing 'protocol holidays' can boost retention in patients (who would otherwise have dropped out of the study). Interestingly, the patients that did opt for protocol holidays had lower retention than patients that did not take any protocol holidays giving insight into behavioural aspects of varied participants in remote monitoring (Figure 8a). This can serve as a way to inform the recruitment strategy for future studies [31]. Figure 8c displays the difference in retention between the groups per cohort. We observe that the ILD and COPD cohorts showed differences between the groups, but the COVID cohort did not show any discernible difference. These results are also corroborated by the log rank test results provided in the supplementary material tables 2a and 2b. Another way we analysed the impact of protocol holidays on retention was by grouping the data collected from patients. One group comprised of data after the patient returned to the study from the holiday, and another where we consider the patient as having dropped out at the start of the first protocol holiday, thus marking that as their last day in the study. Survival curves for these two groups are shown in Figure 8b. There are differences between the two groups, with the group that includes data during protocol holidays showing better retention, particularly between survival probabilities of 1 and 0.5 signifying the improvements in retention due to introduction of protocol holidays. Passive data collection Our findings suggest that the use of passive sensing wearables can effectively enable continuous remote monitoring in patients with respiratory diseases. We observed a higher adherence, engagement (Figure 2,3a), and retention rate (Figure 7) than has been previously reported [32, 33]. Passive data collection not only ensured minimal data loss but also reduced technical errors. The adherence and retention rate of passively collected data from wearables was notably higher than active data, especially among patients who showed little to no adherence to active data monitoring (as shown in Figure 3 and Figure 7c), supporting previously reported findings [25, 31]. Garmin wearables provided effortless passive continuous data on heart rate, respiratory rate, activity, and oxygen saturation levels. However, further controlled benchmarking studies would be needed to evaluate the accuracy and usefulness of these data [20]. It also provided data quality indications in cases when the data is not of appropriate quality to calculate derived metrics, such as negative values in stress values ensuring high-quality data can be differentiated. Active data collection Previous studies using paper diaries have reported challenges with the implementation of PROMs (patient-reported outcome measures) and finger oximeter data in daily care and research [34-36]. Smartphone apps result in better data quality, lower cost, and faster completion time [20, 28]. In our study, ePROMs were successfully implemented using a smartphone app. The app also facilitated easier use of the finger oximeter, which was found to be acceptable and well-perceived. This suggests that online collection using a mobile application can facilitate the implementation of ePROMs. The burden of active data collection also impacted the adherence and retention with low-burden questionnaires (like form-based) having higher adherence and retention compared to high-burden questionnaires and audio recordings as shown in Figure 4 and 7c. Moreover, the COPD and ILD cohorts had higher adherence and retention because they were more hands-on with more contact from the study team, while the COVID-19 cohort had lowest as it was the most hands-off cohort (Figure 4,7b and 8c). On the contrary, faster responses to the questionnaires for the COVID-19 cohort were observed providing behavioural insights compared to COPD and ILD. Similarly, participants that opted for protocol holidays had faster response times for the questionnaires, again suggesting a pattern that participants with lower retention have faster response rates (Figures 8 and 4). Our findings suggest that considerations such as the app used, interface design, and the number, type and burden of tasks required should be taken into account in future studies. Overall, these strategies helped to improve patient adherence to the study protocol. Contrary to previous concerns about the feasibility and reliability of home spirometry due to technical issues [29, 37], our study demonstrated the feasibility of home spirometry, supporting findings from previous studies [20, 29, 37-41]. Specifically, the results indicate a higher adherence rate to weekly measurements of 84% in COPD and 94% in ILD (Figure 3a), compared to that previously reported by Turner et al.[42] at 72% in COPD, and by Johannson et al.[37, 39] at 90.5% and Noth et al. at 86% [39] in ILD. This may be attributed to the use of NuvoAir smart spirometry which features a user-friendly smartphone app and online portal, automatic notifications, periodic notifications through the RADAR-Base active application and the ability for clinicians to send manual reminders to patients. In contrast, previous studies utilised alternative data storage methods such as handwritten diary cards and manual downloads for which adherence rates were not reported, and quality checks were not applicable [43, 44]. Unlike previously reported home spirometry reliability [44] the data obtained in this study is considered to be of high quality and reliability due to its validation and quality check process. The device provides feedback on test quality using ATS grading guidelines (shown in Figure 3b) and depicts diagrams of inspiration and expiration. However, it is still believed that the study duration, frequency of tests required, data transmission method, and technical issues may have affected the overall quality and adherence rate [40]. Acceptability Although the results of our data collection revealed a high level of acceptability and satisfaction with remote monitoring, they also highlighted some areas for improvement. The Garmin wearable was generally well-received, particularly regarding ease of use and comfort during wear. However, some participants noted that the lights emitted by the watch disrupted sleep (Figure 6a). The RADAR app was predominantly perceived as useful, efficient, and user-friendly. A high majority of patients indicated that they would continue to use the app and recommend it to others (Figure 6b). Home spirometry with NuvoAir Spirometer was also favourably accepted (Figure 6c). Patients found it reassuring, easy to set up, and simple to perform consistent with the findings reported previously by Moor et al.[45]. However, some patients felt that it was burdensome to perform daily and that it did not alleviate test-related anxiety. Missing data, technology and technical issues In our study, we encountered several challenges related to remote monitoring technology and internet connectivity that may have impacted patient adherence to the study protocol. These issues included: 1) Technical difficulties which negatively impacted patients' adherence to the study protocol, 2) Difficulty for patients in maintaining motivation to complete required tasks, particularly when experiencing illness or receiving hospital care. 3) issues relating to the functionality of the applications, including freezing, and problems like automatic logout and test errors, requiring patients to log in repeatedly (Figure 9). To mitigate these challenges in future studies and ensure the success of remote monitoring studies for patients with respiratory disease, it is crucial to address the reported challenges and solutions encountered during our study. It is recommended that strategies such as providing training sessions, having readily available technical support, and offering flexible protocol and task adherence options are implemented to ensure optimal patient experience and adherence. In addition, missing passive wearable data were primarily due to recurrent background sync issues with the Garmin devices. These issues are common, and solutions are available on Garmin's support website, but these are not one-size-fits-all and a lot of effort, research and troubleshooting is required to be done by the study management teams. In addition, the need for manual activation of continuous always-on heart rate and oxygen saturation parameters, led to errors and difficulties in activation. Additionally, the bright red and green lights emitted by the sensors at the bottom of the Garmin watch were reported to disrupt patients' sleep. Key takeaways from this feasibility study The ease of use and passive data collection capabilities of wearables can facilitate the development of more in-depth research trials and the recruitment of larger study groups. Telehealth technology enables the ability to monitor patients continuously. Using these devices can prevent delays of important clinical trials and prevent postponement in drug clinical trials during crises such as COVID-19. Additionally, these devices offer a convenient and safe means of providing healthcare services to patients without the need for hospital visits. These utilisation of remote monitoring via wearables can also reduce potential exposure to highly infectious diseases. Telehealth technology potentially enables efficient and cost-effective healthcare research. Decrease the number of long commutes and improve access to healthcare for individuals in rural areas. Moreover, remote monitoring enables continuous monitoring of physiological parameters and symptoms through the use of individualised trained machine-learning models. In order to fully realise the potential of remote monitoring, future research should focus on the development of low-cost, high-quality wearable and remote monitoring devices. Additionally, it is important to consider compensation and incentives for patients in future studies, as some patients have noted the time commitment required for participation. implications for future telemedicine implementation efforts The study was conducted successfully despite COVID-19 pandemic lockdown. We implemented a simple remote recruitment and on-boarding strategy. We also tested modalities with a research team and clinicians and received feedback from patient representatives to better understand barriers and limitations facing the remote monitoring of patients with respiratory diseases. The low dropout rate shows the feasibility of remote monitoring in clinical research. Reasons for dropout were death, feeling unwell, travelling, being overwhelmed, and time commitment. High adherence rate can be attributed to involving patient representatives, simple remote recruitment, strong team communication, monitoring data streaming, providing regular calls, and ensuring ease of communication. The use of a smart finger pulse oximeter that can connect wirelessly to the mobile device and upload the data without extra effort from the patients will further improve adherence, and acceptability of finger pulse oximetry. Limitations and strengths Our study has several strengths, but it also has some limitations. The main limitation is the small sample size (n=60) and limited duration of data collection (6 months). This was acceptable for a pilot feasibility study, but a larger sample size and longer duration of data collection would be needed to improve the generalisability of the findings. Another limitation is that we used a convenience sampling technique of people willing to take part which may limit the representation of the general population. A more rigorous sampling technique would be recommended in future studies. Finally, patient recruitment was delayed and complicated by the COVID-19 pandemic which limited in-clinic spirometry data collection. The study's strengths stem from a successful data collection exercise, demonstration of daily home-based spirometry, and successful implementation of unique strategies such as protocol holidays and real-time exacerbation state detection. It included multi-dimensional investigation of engagement in remote monitoring shedding light on technical, statistical and behavioural aspects of engagement with insights into the impact of burden on adherence, engagement and retention. It provides a reference point for strategies and policies for future studies to improve adherence, retention and engagement and how to mitigate common challenges. Conclusion Our study provides valuable evidence supporting the feasibility and acceptance of remote monitoring technologies in patients with respiratory diseases. Patients were more likely to engage with wearables that collected passive data than with technologies that required active input. Flexibility in remote study design, such as allowing 'protocol holidays' e.g. for patient travel, fatigue, or loss of interest can also substantially enhance retention and will be more representative of a real-world clinical application of a chronic remote monitoring framework. Our study highlights the value of remote monitoring technologies in clinical research and patient care, despite its limitations and challenges. It sets a foundation for future research to improve on these aspects and continue to explore the potential benefits and applicability of remote monitoring, in particular wearables in broader contexts. Clinical Trial: The study was registered with ISRCTN: 16275601Ethics approval was given and patients provided written informed consent.