Use of wearable sensors to assess compliance of asthmatic children in response to lockdown measures for the COVID-19 epidemic

Between March and April 2020, Cyprus and Greece health authorities enforced three escalated levels of public health interventions to control the COVID-19 pandemic. We quantified compliance of 108 asthmatic schoolchildren (53 from Cyprus, 55 from Greece, mean age 9.7 years) from both countries to intervention levels, using wearable sensors to continuously track personal location and physical activity. Changes in ‘fraction time spent at home’ and ‘total steps/day’ were assessed with a mixed-effects model adjusting for confounders. We observed significant mean increases in ‘fraction time spent at home’ in Cyprus and Greece, during each intervention level by 41.4% and 14.3% (level 1), 48.7% and 23.1% (level 2) and 45.2% and 32.0% (level 3), respectively. Physical activity in Cyprus and Greece demonstrated significant mean decreases by − 2,531 and − 1,191 (level 1), − 3,638 and − 2,337 (level 2) and − 3,644 and − 1,961 (level 3) total steps/day, respectively. Significant independent effects of weekends and age were found on ‘fraction time spent at home’. Similarly, weekends, age, humidity and gender had an independent effect on physical activity. We suggest that wearable technology provides objective, continuous, real-time location and activity data making possible to inform in a timely manner public health officials on compliance to various tiers of public health interventions during a pandemic.


Results
Participants' characteristics. For 2020 study period, a total of 108 (57% males) asthmatic children, 53 in Cyprus and 55 in Greece, with an average age of 9.2 years were enrolled in the study, and contributed data between February 3rd and April 26th, 2020. All children had a physician's diagnosis of asthma, while 53% also reported wheezing episode(s) during the past 12 months, 45% unscheduled medical visit(s) for asthma, 26% emergency room visits for asthma, and 20% daily preventive anti-asthma medication during the past 12 months. Among study participants, 43% were characterized as having asthma severity 1, 43% asthma severity 2 and 15% asthma severity 3 (Table 1). For the 2019 study period, from February 3rd to April 26th, we measured mobility for 39 and 52 asthmatic children (59% males) in Cyprus and Greece, respectively, with an average age of 9.3 years.
There was no significant increase between level 2 and 3 periods (p value = 0.298), while the increase between level 1 and 2 periods was almost significant (p value = 0.055). In Greece, the adjusted mean changes in fraction time spent at home were more gradual and moderate, increasing by 14 Table 4). The differences in total steps decreases per day between levels 1 and 2 (p value = 0.212) and levels 2 and 3 (p value = 0.576) did not reach statistical significance.

Discussion
In this study, we have assessed changes in mobility of asthmatic children in Cyprus and Greece in response to different levels of public health interventions for the COVID-19 pandemic using GPS tracking, pedometer and heart rate sensors embedded in wearable watches. Our data imply that asthmatic children in both countries were highly compliant to public health measures, by changing their everyday routine and limiting their activity and mobility outside their homes.
In asthmatic children in Cyprus, we recorded a steep increase in the fraction of time spent at home from 44% to the very high 95% and a steep decrease in total steps per day from 8,996 to 6,270, demonstrating high compliance to the implemented three levels of interventions. In asthmatic children in Greece, we observed a more gradual, stepwise increase in fraction time spent at home from 52% at baseline to the also high 90% and a similar pattern of gradual decrease of physical activity from 8,527 to 5,439 steps/day. The subtle differences in the findings between the two countries can be explained by differences in the actual measures implemented in each country rather than by real differences in the compliance of asthmatic children to the interventions. In fact, intervention measures in Greece were slightly different to those in Cyprus. In Greece level 2 measures included only ban of public gatherings and closure of shops and worship places, and escalated to include personal transport and movement restrictions only in level 3 period. In Cyprus, personal mobility restrictions were implemented from level 2 measures and became stricter (only one movement per day) in level 3, although the corresponding changes in fraction time spent at home and steps per day between level 2 and level 3 periods were not significant in our study group.
The successful management of the pandemic spread in Cyprus and Greece is probably due to the early introduction of lockdown measures in both countries by March 25 and 23, 2020 respectively, and the high compliance of vulnerable groups, and possibly the general population, who spent the majority of their time at home. The effectiveness of lockdown measures to reduce transmission of COVID-19 in several European countries including Greece was previously reported 28 . A recent modelling study evaluated the impact of the sequence of restrictions posed to mobility and human-to-human interactions on the virus transmission in Italy and found that they have reduced transmission by 45% 29 . As of 26 April 2020, 1.43 and 1.22 deaths per 100,000 population were reported for Cyprus and Greece respectively, which places them among the countries with the lowest mortality rates for COVID-19 in the EU/EEA and UK 30 . Our findings may also relate to those of several recent studies that reported reduced asthma morbidity in children during the COVID-19 pandemic, attributed to the reduction of the overall viral infections because of significant changes in their daily activity [31][32][33][34] .
It is well documented that wearable technology is a reliable, objective tool for monitoring numerous diseases and estimating adherence to medication 35 . However, to the best of our knowledge, there are no previous studies examining compliance to public health interventions using wearable devices. Previous reports on adherence to public health interventions during epidemics were based on telephone or mailed interviews and questionnaires [17][18][19]36 . These conventional tools have inherent limitations such as non-response, recall biases, lack of validation of self-reports, influence by concerns about being recognized as breaching quarantine and low-level spatio-temporal information. In contrast, the use of wearable devices provides objective, continuous, real-time location and activity data making possible to timely inform public health officials on the results of various tiers of public health interventions and ensure adequate decision making in escalating or de-escalating interventions.
Using a mixed effects model, we were able to find independent quantitative effects of several factors on the time participants spent at home (weekends, increasing age) and their physical activity (weekends, increasing age, humidity and gender). We also found an interaction effect of weekends on higher fraction of time spent at home and lower physical activity, which was reversed during the implementation of intervention measures in both countries. One possible explanation is that, under normal conditions, families spend more time out of home on weekdays and more time at home during weekends. During the pandemic period this was reversed, as working parents and children had to work or study at home during weekdays and spend more time for outdoor activities on weekends. However, the change of participants' daily behavior in both countries and the increase in their mobility on weekends during the enforcement of the lockdown measures requires further investigation. The independent effect of age on increasing fraction time spent at home and lower physical activity and female gender on lower physical activity in asthmatic children in Cyprus, agree with previous studies that reported lower physical activity in asthmatic girls 37 and older asthmatic children 38 . Year 2020 has been associated with www.nature.com/scientificreports/ lower physical activity in asthmatic children in both countries, which may relate to a systematic effect of either the recording devices used or implication of environmental factors we have not accounted for in the analysis. Digital technology has been widely used in the COVID-19 management, including tracking positive cases, informing citizens about the current epidemiological status in their region, performing virtual clinics and diagnostics through telemedicine and in disease modelling using big data analysis 39,40 . Our study presents another promising application of digital technology in the fight against the pandemic, by employing wearable sensors to provide objective and real-time information about population compliance to public health measures. Therefore, in a public health emergency, it is possible to employ wearable technology in a greater sample of the general population and enable rapid measurements of the public reaction to a particular set of interventions. GPS and physical activity data for example, are collected anonymously by telephone companies from their smartphone customers and theoretically may provide information about population's attitude and adherence to interventions taken in pandemics and/or natural disasters. Towards this aim, in Israel a cell-phone tracking system was used in SARS-CoV-2 positive individuals to facilitate contacts tracing 41 , while data from smartphone users in Italy have been analysed to estimate citizen mobility during the lockdown 42 . In this direction, future applications of wearable digital technology may include the development of appropriate tools and infrastructure for efficient, systematic integration and interoperability of different health and behaviour data sources into existing public health systems. This approach will facilitate the link between symptom-tracking apps, behavior-tracking apps, contact tracing, aggregate population mobility monitoring, healthcare access and e-health monitoring 40 .
Nevertheless, this kind of monitoring raises important ethical issues, as it could be considered as limiting individual freedoms and rights [43][44][45][46] . In our case, monitoring of these patients with wearable devices was already established as part of the ongoing MEDEA project that started much before the spread of COVID-19, and thus we had obtained timely ethical approval and written consent from participants. In the case of pandemics, where decisions and measures are taken within days or even hours, it is extremely difficult to obtain fast-track consent from individuals to record their attitudes with wearable technology. For this reason, future health surveillance tracking and contact-tracing tools and applications should be designed in a way that can preserve individual freedoms and assert patient autonomy in a socially responsible manner, while educating the population about the positive effects of health surveillance 47 .
An important limitation of wearable technology is the loss of signal of GPS tracking, especially in indoor environments, which introduces the challenge of how to treat missing values. As a response, automated microenvironment classification algorithms that include spatial and temporal buffering have been developed and validated, especially for air pollution exposure studies 48 and provide an effective way to account for missing location data. Our findings should be interpreted with caution and not be generalized directly to all children or to the general population, because soon after the appearance of the COVID-19 pandemic it became widely known that asthmatic patients may be at increased risk for severe disease 49 . Furthermore, asthmatic patients were more likely accustomed to protective behavioural changes even before the pandemic, as they were aware of the negative health effects of poor air quality 50,51 and the potential of atmospheric pollutants to further impair the respiratory system and facilitate viral infection, as it has been suggested for SARS-CoV-2 52 . These factors may have led to increased compliance to public health intervention measures for COVID-19, as compared to the general population. Further studies, focusing on the general population are needed to better assess the full extent of population compliance to public health intervention measures for COVID-19. Finally, our cohort lacks information about family characteristics during the pandemic, such as parents' employment status and time spent working from home, which may affect the outcomes of our study.

Conclusion
In conclusion, we implemented novel wearable technology methods to assess personal compliance to public health interventions aiming to contain the spread of a novel, highly contagious virus such as SARS-CoV-2. The successful implementation of public health interventions in Cyprus and Greece, which minimized COVID-19 related mortality in both countries, was reflected in the sharp mobility reductions recorded in the participants of our study early in the course of the outbreak. Wearable devices provide objective, continuous, real-time data that may timely inform public health officials on compliance to various tiers of public health interventions and ensure informed decision-making and strategic planning in the containment of epidemics, both at national and cross-national levels.

Materials and methods
Study setting. Asthmatic children were recruited from primary schools in Cyprus and Greece (Heraklion district, Crete) and were enrolled in the ongoing LIFE-MEDEA public health intervention project (Clinical. Trials.gov Identifier: NCT03503812). The LIFE-MEDEA project aims to evaluate the efficacy of behavioral recommendations to reduce exposure to particulate matter during desert dust storm (DDS) events and thus mitigate disease-specific adverse health effects in vulnerable groups of patients. Details of the study protocol and methods are presented in Supplementary File 1. In order to assess adherence to recommendations, participants are equipped with a wearable device (smartwatch) with several sensors. Participants are instructed to wear the smartwatch throughout the study period, during both DDS and non-DDS days. During non-DDS days, all participants carry out their usual daily activities. The first MEDEA study period took place during February-June 2019, the second study period during February-June 2020.
Study populations and recruitment. In the asthma panel study the target population were children aged from 6 to 11 years with mild to moderate persistent asthma. The eligibility criteria included a physician's diagnosis of asthma and at least one of the following: daily preventative asthma medication, wheezing episodes and/or Physical activity and GPS tracking. Physical activity and global positioning system (GPS) data were recorded between February 3 and April 26, for both study years (2019 and 2020) in Cyprus and Greece using the smartwatch. Data from study year 2020 were extracted to assess the participants' mobility before and during the enforcement of COVID-19 lockdown measures while data from study year 2019 that represent a completely restrictions-free mobility period were also used for comparison. The EMBRACE™ smartwatch (Embrace Tech LTD, Cyprus) was used for data collection. The smartwatch works as a stand-alone device and is equipped with multiple sensors such as pedometer, GPS and heart rate as well as an embedded sim-card for Wi-Fi data transfer. The software is capable of synchronizing the sensors, so the data are transferred to the cloud with the same timestamp. Data on GPS coordinates and steps/time unit, and heart rate are collected per 5-min intervals. Data synchronization with a cloud-based database is performed automatically when the smartwatch contacts the Wi-Fi network inside the participants' home. For each participant we recorded the total number of steps per day (24-h period). In addition, we defined the fraction of time spent at home as the ratio of time with GPS signal within a 100 m radius around the participant's residence divided by 24 h. The 100 m radius was defined as the maximum barrier to account for the accuracy of GPS signal in commercially available GPS receivers 53 . As signal accuracy in urban and especially indoor environments is further blocked or bounced repeatedly off buildings prior to being received 54 , we also classified 5-min intervals with no GPS signal as either "at-home" or "out-home", depending on the signal of the most recent valid GPS recording. Lastly, the days that participants did not wear the smartwatch were identified by absence of heart rate measurements, and were excluded from subsequent analysis. We also excluded from the analysis GPS and pedometer data for DDS days in Cyprus (5 days during February-April 2020 and 4 days during February-April 2019) and Crete-Greece (1 day during February-April 2020 and 2 days during February-April 2019) that may had further influenced the mobility of the participants.   www.nature.com/scientificreports/ with only one mobility permission per person per day (Weeks 9-12: April 01, 2020-April 26, 2020) (Fig. 3). In Greece, the first COVID-19 case was identified on February 26, 2020 and the study period is divided to: i) level 0 (baseline)-no public health interventions (Weeks 1-5: February 03, 2020-March 10, 2020), ii) level 1-social distancing measures, ban of all public events, bars, restaurants and schools closed (Week 6: March 11, 2020-March 15, 2020), iii) level 2-all retail shops and worship places closed, ban of gatherings of > 10 people (Week 7: March 16, 2020-March 22, 2020), iv) level 3-mobility restrictions except for subsistence and health needs (no daily limit) (Weeks 8-12: March 23, 2020-April 26, 2020) (Fig. 3).

Public health (non-pharmaceutical) interventions in Cyprus and
Statistical analysis. Basic demographics and clinical characteristics of the asthmatic children were summarized using mean (standard deviation) for continuous variables and percentages for categorical variables.
In an unadjusted analysis, the mean fraction time spent at home and mean total steps per day were compared between the different periods of public health interventions using ANOVA, while graphs were constructed in order to demonstrate the weekly variation in mobility before and during the implementation of public health interventions in Cyprus and Greece. Furthermore, the course of fraction time spent at home and total steps per day in asthmatic children participating in the second MEDEA study period during February-April 2020 and the same parameters' course in asthmatic children participating in the first MEDEA study period during February-April 2019 are displayed in separate graphs for Cyprus and Greece.
The changes in the daily levels of fraction time spent at home and total steps were further explored in a mixed effect model, which included a fixed effect term for the level of public health interventions and a random intercept for each participant. The mixed effect model was adjusted for the effect of age, gender, temperature, humidity, year, and weekend on mobility. In addition, sine and cosine functions were included to control for monthly variability in our data. Finally, for each parameter, we used an interaction term to test for differential change of mobility across the levels of interventions measures.
All statistical comparisons were performed using STATA 12 (StataCorp, TX) and a p value lower than 0.05 was considered as statistically significant. www.nature.com/scientificreports/