Feasibility Study of a Novel Method of Contactless Radar Sensing for Sleep Monitoring across Adolescence (8-18yrs): Ambient-Teens

Feasibility Study of a Novel Method of Contactless Radar Sensing for Sleep Monitoring across Adolescence (8-18yrs): Ambient-Teens | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Feasibility Study of a Novel Method of Contactless Radar Sensing for Sleep Monitoring across Adolescence (8-18yrs): Ambient-Teens Lorna Caddick, Dawn Haughton, Joanna Inchley, Sharon Anne Simpson, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9085766/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Healthy sleep plays an important role in adolescent development and links between sleep, physical and mental health outcomes remain poorly understood. To capture normative trends and abnormal changes, we need methods of sleep measurement used continuously over long time-frames. Wearable devices are the most common objective method as an affordable and less invasive alternative to gold-standard polysomnography. However, wearables are not always well-tolerated by young people impacting adherence and sleep, hence, study designs typically collect data for 1–2 weeks. This study aimed to test the feasibility of radar sensing for assessment of sleep, in children and adolescents (8–18 years), and compare performance against standard objective (wrist accelerometery) and subjective (sleep diary) methods. Radar data were collected over four weeks with two concurrent weeks including accelerometer and sleep diary data. Following data cleaning, usable data from the radar sensor was significantly higher (86%) compared to accelerometer (61%; p < 0.05), with particularly low accelerometer adherence noted among boys. Radar showed good agreement with accelerometer data for sleep onset (mean difference − 0.96 minutes, 95% limits of agreement: -123.16 to 121.24) and wake times (mean difference 4.53 minutes, 95% limits of agreement: -69.45, 78.51), however, poor agreement for total sleep time (mean difference 38.96 minutes, 95% limits of agreement: -78.40, 156.31). Participant feedback indicated the radar sensor as the preferred method (80% vote). The results of this study support the value of radar sensing as a contactless, continuous, acceptable and passive method for longitudinal assessment of sleep by young people. Adolescence sleep technology longitudinal Figures Figure 1 Figure 2 1. Introduction Adolescence is a period of critical neurobiological and social development, and sleep disruption at this life stage is associated with multiple adverse physical and mental health outcomes (Bacaro et al., 2024 ; Castiglione-Fontanellaz et al., 2023 ; Dong et al., 2019 ; Hysing et al., 2016 ; Morales-Ghinaglia & Fernandez‐Mendoza, 2023 ; Zhang et al., 2017 ). Studies report adolescents experience a shift to later sleep times, driven by developmental changes to their circadian clock (Crowley et al., 2007 ; Hagenauer et al., 2009 ; Lucien et al., 2021 ; Taylor et al., 2005 ). Many young people struggle to meet recommended sleep guidelines, such as age-advised sleep duration and consistent sleep regularity (Gariepy et al., 2020 ). Therefore, it is important to better understand developmental changes to circadian and sleep timing across adolescence to avert negative trajectories that can have long-term implications (Johnson et al., 2018 ). 1.1. Requirement for continuous longitudinal methods of sleep data collection Intervention studies suggest that poor sleep could be a lifestyle-related modifiable risk factor, either directly, e.g. cognitive and behavioural therapy (Olaithe et al., 2024 ), or through other lifestyle changes, e.g. reducing screen-time (Martin et al., 2021 ). However, the relationships between sleep timings, normal developmental shifts and negative outcomes are poorly understood. Many studies have used sleep assessment methods over relatively short time-frames (1–2 weeks) or bursts of intervals, as until recently there have been limited methods that are able to record measures over developmental temporal scales (Bei et al., 2014 ; Duncan et al., 2018 ; Lunsford-Avery et al., 2020 ; Short et al., 2012 ). Young people develop at different rates and sleep can be interrupted over short periods by contemporaneous stresses e.g. illness, school transitions or exam stress. Therefore, we require methods that can capture how normative variation or conversely abnormal sleep patterns display over the longer term, at the day-to-day level over extended periods. 1.2. Feasibility of sleep devices “Nearable” sleep monitoring, such as radar or under-mattress sensors, have the benefit of contactless recording. Low tolerance of other objective methods (polysomnography and wearables) include resourcing specialist technicians, charging requirements and discomfort (Lyall et al., 2020 ). Nearables provide an opportunity to overcome these problems with measurement in the home-setting, that could minimise participant burden, optimise participant adherence, and provide insights over longer durations. Nearables are potentially tools for application in research studies to facilitate data acquisition periods that span the time course of an intervention in clinical studies, or developmental periods such as the transition from child to adolescence to adulthood. However, many nearable methods are yet to be assessed for feasibility and acceptability in young people. 1.3. Aims This study was co-produced with young people to test the feasibility of a contactless radar-based sleep sensor for use by adolescents aged 8-18yrs. The primary objective was to assess data quality and participant adherence over 1 month of radar-based sleep assessment in the home setting. A second aim was to compare the performance of the radar sensor for assessment of sleep timings against standard objective (wearable accelerometery) and subjective (sleep diary) methods. 2. Methods 2.1. Participants Young People were recruited via engagement with young person advisory groups (YPAG), schools affiliated with the Scottish Schools Health and Wellbeing Improvement Research Network (SHINE) and word of mouth. This study was approved by the Edinburgh Medical School Research Ethics Committee and informed consent was obtained from parents/caregivers of participants aged 8–15 years and directly from participants aged 16–18 years. Study participants received a certificate, thank-you letter, individualised summary sleep report and vouchers up to £30 for completing study-related activities, at the end of their participation. 2.2. Protocols Participants took part in data collection between January and December 2024 (Fig. 1 ). Participants completed questionnaires at baseline, followed by 4 weeks of sleep assessment using a radar-based sleep sensor (Somnofy, Vitalthings, Norway). Concurrently, participants were asked to wear an accelerometer (AX3, Axivity, UK) and complete a sleep diary (Consensus Sleep Diary, (Carney et al., 2012 )) for 2 weeks. Participants completed a feedback questionnaire at the end of the study and were invited to take part in engagement activities. Detailed description of the methods can be found in the protocol paper (Caddick et al., 2024 ). Modified from protocol paper ( Caddick et al., 2024 ). Multimodal data acquisition included contactless sleep sensor (4 weeks), accelerometery (2 weeks), and sleep diary (2 weeks). Questionnaires included demographics survey and sleep and health questionnaires at baseline followed by a feedback questionnaire at the end of sleep data collection. Participant engagement included invitation to weekly check-in calls followed by an optional citizen science activity at end of data collection. All icons are designed by Freepik. This study implemented YPAGs, participant check-in calls and citizen science activities to engage young people in the development of ecologically valid research methods (Gale et al., 2025 ), (Caddick et al., 2024 ) and Supplementary Material. 2.2.1. Questionnaires Participants completed questionnaires to assess demographics, sleep routines and disturbances, physical development, and mental health (Munich Chronotype Questionnaire, MCTQ (Ghotbi et al., 2020 ; Roenneberg et al., 2019 ); Sleep Disturbance Scale for Children, SDSC (Bruni et al., 1996 ); Pubertal Development Scale, PDS (Carskadon & Acebo, 1993 ); Revised Childrens’ Anxiety and Depression Scale, RCADS (Chorpita et al., 2000 )), see Supplementary Materials. If participants were < 16 years, a parent/caregiver completed the SDSC, PDS and RCADS on their behalf. At the end of the study period, participants were asked to complete a feedback survey. All questionnaires were completed online (Online Surveys, version 3, JISC). Demographics Questionnaire Demographic information (e.g. age, sex and ethnicity) were collected alongside details of bedroom arrangement to inform placement of the bedside radar sensor. The Scottish Index for Multiple Deprivation (SIMD) was used to score socioeconomic status based on participant postcode and categorised into high (> 7) and low ( = < 7). Urban or rural location was defined using the Scottish Government Urban Rural Classification. Ethnicity was categorised from open-text responses as White or Asian/Mixed/Other based on Government Statistical Service harmonised categories for ethnicity. Revised Children’s Anxiety and Depression Scale (RCADS-47) The 47-item RCADS was scored for major depressive disorder (MDD) and total anxiety. Clinical cut-off T-scores categorised individuals into ‘Normal Range’ below 65 or ‘Borderline/Clinical Range’ equal to or above 65, as recommended by the survey developers (Chorpita et al., 2000 , 2005 ). Sleep Disturbance Scale for Children (SDSC) The SDSC contains 26-items relating to adolescent’s sleep disruption over the past 6 months. Participants were scored for sleep problems using the recommended cut-off score of 39 (Bruni et al., 1996 ). Feedback Questionnaire The feedback questionnaire consisted of multiple choice and open-text questions about the radar sensor, accelerometer and overall study participation. 2.2.2. Contactless radar-based sleep sensor All participants were asked to complete four weeks of sleep assessment using the radar-based sleep sensor. Participants received the radar sensor by post along with set-up instructions. Details of the set-up procedure are provided in Supplementary Materials. Data were acquired remotely from the device to a cloud-based data platform that could be accessed in real-time by the researchers via an Application Programming Interface (API). A ‘session’ was recorded by the radar sensor when movement was detected by the device for at least 30 minutes. Radar data was processed and analysed using a proprietary sleep algorithm based on temporal neural networks trained on movement and respiratory rate data. The algorithm returned parameters related to sleep timing and duration (Toften et al., 2020 ). Radar data were downloaded from the cloud-based data acquisition platform at the end of the 4-week data collection period. Sessions that did not include time asleep were excluded. To ensure only one recording per night for each participant, data were included if the session started after 18:00 and ended on the next day, or if the session started between 00:00 and 04:00 and ended on the same day. These were judged to be physiologically valid values to remove additional sessions containing sleep. Sessions were excluded if time in bed was < 3.5 hours, as recommended by the device manufacturers. Sessions were excluded when session start time equalled time at sleep onset, and when session end time equalled wake time. Sleep onset and wake times that were considered unlikely for this age group (onset time 04:00–20:00, wake time 12:00–04:00) were inspected manually and excluded if they did not match sleep diary entries or an issue with the device was detected. Further details are included in the Supplementary Materials. 2.2.3. Accelerometery Accelerometers were configured using Open Movement GUI (OMGUI, V1.0.0.43) software. Raw acceleration data was collected at a frequency of 50Hz with a range of +/- 8g. Devices were posted to participants’ home or school address with instructions to wear on their non-dominant wrist for 2 weeks, and only remove for activities that involve water. On return of the devices, Continuous Wave Accelerometery data files (.cwa) were processed using an R package GGIR (v3.2.6) (Migueles et al., 2019 ). Accelerometer derived sleep variables were determined by timepoints classified as bouts of sustained inactivity (SIB) - periods of time during which there was no change larger than 5° in any angle over at least 5 minutes (Van Hees et al., 2015 ). Time at sleep onset was defined as the start of the first nocturnal SIB. Wake time was defined as the end of the last nocturnal SIB. Total sleep time was calculated as the sum of all nocturnal SIBs. GGIR processing excluded days that did not have at least 16 hours per day of wear time, as default. GGIR data processing options were selected so that short times of non-wear were inputted by the average at similar time points on different days of the week. In addition, sleep onset and wake times that were considered unlikely for this age group (onset time 04:00–20:00, wake time 12:00–04:00) were flagged for manual checking and were excluded if they did not match sleep diary entries or an issue with the device was detected. 2.2.4. Sleep Diary Consensus sleep diaries (Carney et al., 2012 ) were posted to participants, to be completed by hand, for 2 weeks, concurrent with the accelerometer. As per Consensus instructions, participants were instructed to complete the diary within an hour of waking and to enter the date as the morning they are filling out the diary. Time at sleep onset was estimated from the sum of time ‘tried to go to sleep’ and the minutes taken to fall asleep. Wake time was estimated from ‘time at final awakening’. Total sleep time was calculated as the minutes from sleep onset to wake time, subtracting minutes estimated awake during the night. Sleep diary data was excluded if there was not a measure of time at sleep onset or wake time. Some participants expressed confusion about the date to enter in the sleep diary. Therefore, alignment of diary wake time was checked against radar-measured wake time. This identified 4 participants that entered the date as the night of going to sleep rather than the morning of completing the diary, so these dates were adjusted. 2.3. Data Quality Data completeness was evaluated as the mean percentage of collected and clean data across methods. To compare across methods, only participants that completed all three methods were included in this analysis. Expected data were defined as the number of nights from the first to the last recorded session, per participant. Collected data were defined as all nights with a recording of time asleep > = 2 hours. ‘Clean nights’ were defined as specified by the data cleaning steps for each method. Out of the 48 participants, some participated on nights that crossed daylight savings (radar = 15, accelerometer = 9, diary = 10). Sleep onset, wake times and duration were adjusted for the clock change and included. 2.4. Adolescent Engagement During pre-study YPAG sessions, adolescents expressed eagerness to be engaged more frequently throughout the project. Therefore, we invited study participants to weekly check-in calls during their sleep measurement to discuss progress and share highly summarised illustrations of their own sleep data from the radar sensor. Minimal aggregate data were presented to minimise influencing participants’ sleep behaviours. All participants were invited to take part in an optional citizen science engagement activity at the end of data collection. Participants who completed the engagement activity were invited to a final study YPAG. In these sessions young people decided to create a video series aimed at other young people, to share their experience and study results. Further details about participant engagement can be found in the Supplementary Material. 2.5. Statistical Analysis Descriptive data are presented as mean and standard deviation (SD), or percentages. Key sleep variables for analysis were time at sleep onset, wake time and total sleep time (TST). Data completeness (mean percentage of collected and clean nights from expected nights) were compared across sleep data collection methods, using Friedman test with post-hoc pairwise comparisons using Wilcoxon signed-rank tests with Bonferroni correction. Exploratory analyses assessed biases in data completeness using Mann Whitney U test, or Kruskal-Wallis test if there were more than 2 groups, for sex, age-range, SIMD, location, total sleep disturbance score, total anxiety and total depression scores. Methods of sleep measurement were compared using Bland-Altman plots corrected for within-subject and between-subject variability and unequal numbers of repeats to look at the mean difference and limits of agreement (Bland & Altman, 1999 ). Linear mixed-effect models were used to determine if key sleep variables were influenced by sex, age-range, SIMD, location, sleep measurement method or weekday/weekend, which were included as fixed effects. Random intercepts were included for subjects and for subject-nights to account for individual differences and multiple devices per night. Age-ranges were compared for key sleep variables, using a linear mixed-effect model with a random intercept for subjects. Kenward-Roger approximation for degrees of freedom was used to compute a global p-value for each age comparison. Post-hoc pairwise comparisons were performed using estimated marginal means with Tukey’s adjustment. Significance level was set at p < 0.05. All analyses were completed using R version 4.4.1. In the development of the analysis code, Edinburgh’s Learning Machine (ELM), the University of Edinburgh's secure generative artificial intelligence platform was used, and manually checked, to ensure accuracy and efficiency. Participant feedback was summarised from multiple choice and open-text answers from the feedback questionnaire as well as notes from check-in calls. 2.6. Data Availability Datasets generated in this study will be deposited with the UK Data Service and available at https://doi.org/10.17605/OSF.IO/BKY4S . 3. Results 3.1. Sample Characteristics A total of 48 young people, with elective recruitment by sex and age brackets: 8–11, 12–15 and 16–18 years, participated in the Ambient Teens Sleep Study, see Table 1 . Table 1 Overview of participant characteristics by age range Age Range (years) Total 8–11 12–15 16–18 N 16 17 15 48 Sex, Female, N (%) 9 (56%) 9 (53%) 9 (60%) 27 (56%) Age, years, mean (SD) 10.3 (1.1) 13.0 (1.2) 16.7 (0.5) 13.3 (2.8) SIMD, mean (SD) 8.7 (2.3) 7.0 (2.2) 8.3 (2.2) 8.0 (2.3) Location, Urban, N (%) 15 (94%) 11 (65%) 12 (80%) 38 (79%) Ethnicity, White, N (%) 13 (81%) 17 (100%) 13 (87%) 43 (90%) Pet, N (%) 6 (38%) 11 (65%) 11 (73%) 28 (58%) Nightlight, N (%) 7 (44%) 4 (24%) 0 (0%) 11 (23%) Shared Bedroom, N (%) 1 (6%) 2 (12%) 0 (0%) 3 (6%) Radar Bedside Table Mounted, N (%) 12 (75%) 15 (88%) 14 (93%) 41 (85%) SIMD - Scottish Index for Multiple Deprivation (1-high deprivation score, 10-low deprivation score). Location denotes urban or rural postcode based on Scottish Government Urban Rural Classification. Due to a small sample size, ethnicity is reported as white/non-white. 3.2. Protocol adherence and data quality One young person withdrew from the study due to other commitments, all data already collected were included (13 nights). One participant’s radar data were removed as they did not routinely sleep in their own bed for the full night (20 nights). In total, seven of the forty-eight participants were excluded from the data completeness analysis as they did not participate in at least one method. One participant did not use the radar sensor due to difficulties setting up the device with a canopy bed (2%). Two participants did not complete the sleep diary due to lack of engagement or withdrawing from the study (4%). Five participants did not wear the accelerometer, due to study withdrawal or personal preference including initial discomfort (10%). Data completeness for participants that completed all three methods (N = 41) is shown in Table 2 . Mean percentage of collected data was radar: 89.6%, accelerometer: 78.6% and diary: 90.6%. Mean percentage of clean data was radar: 86.4%, accelerometer: 60.1% and diary: 88.7%. The accelerometer was significantly lower than both the radar sensor and diary for both collected and clean data. Four participants had no clean accelerometer data, due to only wearing at night. Table 2 Data completeness across methods Radar (4 weeks) Accelerometer (2 weeks) Diary (2 weeks) Mean ± SD Mean % ± SD Mean ± SD Mean % ± SD Mean ± SD Mean % ± SD Expected 28.85 ± 6.43 - 15.98 ± 3.95 - 14.46 ± 1.92 - Collected 25.76 ± 6.56 89.55 ± 14.97 13.10 ± 4.40 78.58* ± 21.87 12.95 ± 2.51 90.58 ± 17.66 Clean 24.88 ± 6.69 86.39 ± 15.90 11.08 ± 4.31 60.55* ± 29.01 12.68 ± 2.85 88.67 ± 20.16 Mean and SD for the number of nights available by each method of sleep monitoring at different levels of data completeness. Also presented as mean percentage of expected data for comparison across methods. N = 41, mean percentage of collected and clean accelerometer data were significantly lower compared to both radar and diary methods (Wilcoxon signed-rank pairwise comparisons with Bonferroni correction, * p < 0.05), other pairwise comparisons were not significant. Exploratory analysis found no significant differences for the mean percentage of clean radar or accelerometer data, based on sex, age-range, SIMD, location, sleep disturbance score, anxiety score or depression score (Table 3 ). However, females provided on average 67.5% clean accelerometer data compared with 52.5% from males. Comparatively, clean data from the radar sensor was on average 84.9% for females and 88.1% for males. Participants that reported high depression, high anxiety or high sleep disturbance scores were found to have a significantly lower percentage of clean data from the sleep diary, p < 0.05. Table 3 Bias in Data Completeness (as % of expected nights) across demographics Co-Founder Category N Mean % Clean Radar (SD) p- value Radar Mean % Clean Acc (SD) p- value Acc Mean % Clean Diary (SD) p- value Diary Sex Female 22 84.88 (18.28) 0.479 67.48 (23.29) 0.166 91.89 (18.64) 0.209 Male 19 88.14 (12.88) 52.54 (33.34) 84.96 (21.70) Age Range 8–11 12 86.85 (12.17) 0.595 53.84 (28.94) 0.398 91.67 (20.03) 0.367 12–15 15 90.25 (9.31) 60.22 (31.76) 86.62 (19.75) 16–18 14 81.86 (22.79) 66.67 (26.69) 88.32 (21.86) Location Urban 33 85.53 (17.35) 0.921 59.14 (29.97) 0.587 87.74 (22.12) 0.647 Rural 8 89.96 (7.00) 66.37 (25.62) 92.52 (8.12) SIMD High (> 7) 27 84.39 (18.44) 0.465 57.27 (30.54) 0.327 88.99 (19.24) 0.886 Low ( < = 7) 14 90.25 (8.55) 66.89 (25.67) 88.07 (22.58) SDSC High 22 84.18 (18.79) 0.626 60.80 (29.88) 0.559 83.68 (24.39) 0.043* Low 14 88.43 (11.61) 55.78 (31.06) 96.52 (9.52) Anxiety High 7 89.05 (8.92) 0.903 61.50 (29.54) 0.862 67.02 (29.92) 0.002* Low 34 85.84 (17.03) 60.36 (29.35) 93.13 (14.49) Depression High 8 78.35 (26.39) 0.410 60.59 (28.96) 0.856 68.26 (29.71) 0.003* Low 33 88.34 (11.98) 60.54 (29.47) 93.62 (13.63) Mean percentage of clean data from each method across demographics. Mann Whitney U test (2 groups), or Kruskal-Wallis test (3 groups), *p < 0.05 (exploratory analysis due to small sample size). 3.3. Device Comparison The radar sensor, accelerometer and sleep diary were compared for sleep onset, wake time and TST to assess agreement in sleep metrics between methods (Fig. 2 ). All methods showed good agreement for time at sleep onset (mean difference < 11 minutes) and wake time (mean difference < 5 minutes). However, the accelerometer recorded shorter total sleep time compared to the radar sensor with a mean difference of 38.96 minutes (95% limits of agreement: -78.40, 156.31) and against the sleep diary with a mean difference of 50.79 minutes (95% limits of agreement: -95.39, 196.98). A linear mixed-effect model was used to determine if key sleep variables were associated with covariates, and mean sleep times were reported across age-range and method and day, in the Supplementary Materials. 3.4. Participant Feedback The majority of participants (80%) preferred the radar sensor over the accelerometer and this was consistent across sex and age-range. Participants generally reported that the radar sensor did not affect their sleep (80% voted ‘not at all’), whereas 37.5% of participants reported that the accelerometer had some effect on their sleep. Overall, the radar sensor was viewed as somewhat easy to set-up, however, the ease of wearing the accelerometer covered a wider range from easy to impossible. Fifteen percent of participants did not have a bedside table and so used alternative mounting options. Providing both written and video instructions for setting up the radar sensor was considered helpful and participants in the youngest and middle age ranges mentioned requiring the support of caregivers to setup the radar device. Some participants (N = 11, 23%) shared that the accelerometer caused skin irritation and discomfort resulting in removing the activity band temporarily or completely. Participants, mostly younger, also reported issues with the accelerometer being too large (N = 8, 17%). Responses regarding obtrusiveness of the radar sensor were split evenly between ‘not at all’ and ‘only a little’. Many participants, in all age ranges, commented that it took a few days to get used to the radar sensor but then they forgot about it for the remaining weeks. When asked about their least favourite part of the study, 55% listed the sleep diary, mentioning the burden of remembering to complete the diary each day and stress in accurate recollection of sleep times, e.g. time to fall asleep or time spent awake during the night. Notably, the end-of-study engagement activity had a poor completion rate, 27% of participants. The post-study YPAG generated a short video series summarising their involvement and feedback in the study, see website. 4. Discussion Our results demonstrate that the radar-based sleep sensor was acceptable and feasible for use and preferred to other methods of sleep measurement, by young people aged 8-18yrs. Rates of data completeness for the radar sensor were significantly higher compared to the accelerometer. The radar sensor showed good agreement with the accelerometer and sleep diary for time at sleep onset and wake time. In addition, we note there was a lack of agreement between the accelerometer and other methods for total sleep time. 4.1. Protocol Adherence and Data Quality Data loss, particularly over longitudinal study designs, is expected but using methods that require minimal active input could optimise data completeness. Reasons for nights without a recording from the radar sensor included sleeping elsewhere (holidays, sleepovers or time spent at a second home) and disconnection of the radar device from power or internet connection. Mean usable data from the radar sensor was only 3% less than collected data, comparing favourably to the accelerometer which showed an 18% reduction. Excluded radar data were most commonly influenced by factors such as problems with device signals, time awake in bed with limited movement imitating sleep, and interference from pets. It is expected that future work can create algorithms to detect these issues, e.g. the influence of pets or daytime bedroom activity based on respiration rate and movement (Wang et al., 2019 ), which was outside the scope of this study. Only in one case was the radar sensor not able to be set up. Placement of the device near the participants bed can be more challenging for young people due to bedroom arrangements such as bunkbeds, raised beds, shared bedrooms or covers over beds such as nets/tents or cabin/canopy beds. However, alternative mounting options such as a gorilla tripod, that can be hooked around bedframes, provided a solution in most settings. Device placement needs to be considered for longer-term studies where it may be necessary for a device to remain in place for months or years without obstructing daily activities. A radar sensor may not be ideal for young people who regularly sleep elsewhere, or frequently leave their bed during the night. It was important to reassure participants' regarding privacy concerns, by providing a clear explanation of what was being recorded, with examples. While the initial setup was not described as difficult, participants and their parent/caregivers appreciated clear and concise video and written instructions as well as easy communication with the study team via phone or text. Accelerometer data loss included participants removing the device due to discomfort, forgetting to put it back on and reluctance to wear on holiday or for full days. It should be noted that compared to other studies which solely used wearables (Rowlands et al., 2018 ) participants were likely to show reduced adherence to methods that were disruptive, due to the concurrent use of three methods of monitoring sleep. Excluded nights from the accelerometer were most commonly due to participants removing the wearable during the day or only wearing at night, so they did not meet the requirement for at least 16 hours of daily wear time. It is possible to reduce the wear time requirement (Toftager et al., 2013 ), however this influences the detection of sleep timing for all nights. Although participants had been instructed to wear the device at all times, only removing it for water-based activities, some did report that during other activities such as sports or due to school restrictions they were required to remove the accelerometer. There were no significant biases found in data completeness for the radar sensor or accelerometer for different demographics, however sample sizes for each characteristic were small. Males demonstrated a lower proportion of usable accelerometer data than females, losing almost 50% of expected data. However, higher and more equal rates of data completeness were seen from the radar sensor. This suggests that the contactless passive method may help to maintain more balanced adherence rates across sexes. The lower proportion of usable sleep diary data in those with mental health or sleep problems, infers the difficulty to capture sleep information from subjective reporting in these young people. Average usable data from the radar sensor was 26% higher than for the accelerometer, notably over the longer 4-week period (radar) compared to 2 weeks (accelerometer). More available data from the radar sensor may, in part, be explained by real-time data uploads to an online server allowing communication with participants if a radar device went offline, compared to the local storage of data on the accelerometer. This strongly suggests that contactless passive methods of collecting sleep data have great potential for continuous use in longitudinal research designs with higher rates of adherence. 4.2. Sleep Timing Adolescents displayed different sleep timings across age brackets, consistent with previous research (Colrain & Baker, 2011 ). Data from the radar sensor demonstrated that young people are not meeting age-advised sleep duration (Hirshkowitz et al., 2015 ), supporting reports that many adolescents are not achieving the recommended sleep guidelines (Kocevska et al., 2020 ), and there is a need to develop engaging interventions. Sleep timing was significantly associated with day (weekday or weekend). However, a limitation of this study was lack of reporting school days and free days. School holidays, study leave and sick days can all influence the time spent in the young person’s bedroom or bed. When using a radar sensor that detects movement within a radar zone in the bedroom, it is useful to note these changes to routine. In addition, engaging measures to record daily routines over long time-frames are required to further explore associations with sleep timing. 4.3. Device Comparison Preliminary analyses confirmed good agreement for sleep onset and wake time between the radar sensor, accelerometer and sleep diary. Radar-derived sleep duration displayed closer agreement with the sleep diary than the accelerometer. It has been shown in previous work that sleep diaries typically overestimate sleep duration (Arora et al., 2013 ) and that the Axivity AX3 can underestimate total sleep time compared to PSG (Meredith-Jones et al., 2024 ). The large differences in total sleep time across methods should be explored further. In adults, the radar sensor was shown to overestimate TST by 8 minutes compared to PSG (Toften et al., 2020 ), however PSG and radar have not been compared in young people. Adolescents typically display different sleep patterns so it is important for future work to do this validation. Wearable methods, often wrist-based accelerometery, are the most common objective method of measuring sleep timing. These may be preferrable as an affordable method, however other methods such as contactless radar-based sleep sensors could be comparatively affordable for future scalable studies. Further details and comparison can be found in the study summary video. Following validation in adolescents, radar-sensors could be considered as an alternative approach for those that cannot tolerate wearables or to explore changes continuously over longer timeframes. 4.4. Engagement Citizen science is encouraged as a method of engagement for adolescents to learn new skills and have greater involvement in research activities (Frazer et al., 2024 ). However, the optional end-of-study citizen science activities demonstrated particularly low rates of completion. The rotational study design limited the ability to do these in a group setting. Many young people viewed these activities as additional work and much preferred short but frequent online calls during data collection for engagement and learning. These calls gave participants the opportunity to ask questions about the study procedures but also to receive feedback incentives such as visualisation and explanation of summary sleep data. Young people noted that individualised data was of most interest to them. While care must be taken to ensure that data given to participants is clear and concise to limit misunderstanding and changes to behaviour, summary-level, aggregate data and considered timing of reports make this possible. 4.5. Strength and Limitations The small sample sizes are a limitation of this study, and our results on the performance of radar sensing compared to standard methods require replication in a bigger cohort of children and adolescents. Despite the small sample size, the qualitative data that we present on feasibility will facilitate future studies of radar sleep sensing in young people. Co-production was incorporated to ensure adolescents views were integrated into the study design. Furthermore, engagement strategies throughout the study encouraged participant retention, as shown with only one withdrawn participant. While this study gathered feedback from young people with different backgrounds, varying bedroom arrangements and family composition, the socio-economic diversity of the sample was limited. As this study is primarily a feasibility study, our small sample size allowed for comprehensive engagement of the participants, but limited power to analyse bias between groups or detect small effects. Representation across a large age range (8-18yrs) was prioritised over other demographics. Due to the young ages of participants, parents and caregivers of those under 16 completed three questionnaires on behalf of the young person for sleep disturbances, mental health and physical development. These measures are unlikely to be directly comparable with the self-reported responses from those aged 16–18. We aimed to demonstrate the ability to collect data over longer time frames, up to 8 weeks, as outlined in the protocol (Caddick et al., 2024 ), however, due to unanticipated time constraints this was not possible. 5. Conclusion Contactless radar sensing has the potential to provide enhanced insights into long-term sleep patterns across adolescence. Young people strongly preferred the radar sensor to other methods that are currently used as standard in sleep research and this study clearly demonstrated improved adherence to data collection via the passive radar device. Future work to establish the validity and scaling (including cost) of the radar sensor is required, such as performance testing against PSG in young people and collecting continuous data over months. Using a passive device has the potential to generate large datasets to better understand sleep timing and associated health outcomes across the length of adolescent developmental periods. Declarations Acknowledgements: We would like to thank all of the young people who took part in this project for providing valuable feedback, to the co-producers who helped co-design the project and created the engaging video series and to Vitalthings for their technical support throughout data collection. This work was funded by the Medical Research Council (MRC), grant number MR/X028917/1. JI, DH and JB were also supported by the MRC (MC_UU_00022/1) and the Chief Scientist Office of the Scottish Government Health and Social Care Directorate (SPHSU16). References Arora T, Broglia E, Pushpakumar D, Lodhi T, Taheri S (2013) An Investigation into the Strength of the Association and Agreement Levels between Subjective and Objective Sleep Duration in Adolescents. PLoS ONE 8(8):e72406. ttps://doi.org/10.1371/journal.pone.0072406 Bacaro V, Miletic K, Crocetti E (2024) A meta-analysis of longitudinal studies on the interplay between sleep, mental health, and positive well-being in adolescents. Int J Clin Health Psychol 24(1):100424. ttps://doi.org/10.1016/j.ijchp.2023.100424 Bei B, Allen NB, Nicholas CL, Dudgeon P, Murray G, Trinder J (2014) Actigraphy-assessed sleep during school and vacation periods: A naturalistic study of restricted and extended sleep opportunities in adolescents. J Sleep Res 23(1):107–117. ttps://doi.org/10.1111/jsr.12080 Bland JM, Altman DG (1999) Measuring agreement in method comparison studies. Stat Methods Med Res 8(2):135–160. ttps://doi.org/10.1177/096228029900800204 Bruni O, Ottaviano S, Guidetti V, Romoli M, Innocenzi M, Cortesi F, Giannotti F (1996) The Sleep Disturbance Scale for Children (SDSC) Construct ion and validation of an instrument to evaluate sleep disturbances in childhood and adolescence. J Sleep Res 5(4):251–261. ttps://doi.org/10.1111/j.1365-2869.1996.00251.x Caddick L, Gaggioni G, Haughton D, Brown J, Inchley J, Simpson SA, Lyall LM, Murray A, Romaniuk L, Krondorf L, Lopez LM, Smith DJ, Wyse C, Whalley H (2024) Ambient Teens Sleep Study: Protocol for a co-produced feasibility study in adolescents using a contactless radar-based sleep sensor. PLoS ONE 19(11):e0313286. ttps://doi.org/10.1371/journal.pone.0313286 Carney CE, Buysse DJ, Ancoli-Israel S, Edinger JD, Krystal AD, Lichstein KL, Morin CM (2012) The Consensus Sleep Diary: Standardizing Prospective Sleep Self-Monitoring. Sleep 35(2):287–302. ttps://doi.org/10.5665/sleep.1642 Carskadon MA, Acebo C (1993) A self-administered rating scale for pubertal development. J Adolesc Health 14(3):190–195. ttps://doi.org/10.1016/1054-139X(93)90004-9 Castiglione-Fontanellaz CEG, Schaufler S, Wild S, Hamann C, Kaess M, Tarokh L (2023) Sleep regularity in healthy adolescents: Associations with sleep duration, sleep quality, and mental health. J Sleep Res 32(4):e13865. ttps://doi.org/10.1111/jsr.13865 Chorpita BF, Moffitt CE, Gray J (2005) Psychometric properties of the Revised Child Anxiety and Depression Scale in a clinical sample. Behav Res Ther 43(3):309–322. ttps://doi.org/10.1016/j.brat.2004.02.004 Chorpita BF, Yim L, Moffitt C, Umemoto LA, Francis SE (2000) Assessment of symptoms of DSM-IV anxiety and depression in children: A revised child anxiety and depression scale. Behav Res Ther 38(8):835–855. ttps://doi.org/10.1016/s0005-7967(99)00130-8 Colrain IM, Baker FC (2011) Changes in Sleep as a Function of Adolescent Development. Neuropsychol Rev 21(1):5–21. ttps://doi.org/10.1007/s11065-010-9155-5 Crowley SJ, Acebo C, Carskadon MA (2007) Sleep, circadian rhythms, and delayed phase in adolescence. Sleep Med 8(6):602–612. ttps://doi.org/10.1016/j.sleep.2006.12.002 Dong L, Martinez AJ, Buysse DJ, Harvey AG (2019) A composite measure of sleep health predicts concurrent mental and physical health outcomes in adolescents prone to eveningness. Sleep Health 5(2):166–174. ttps://doi.org/10.1016/j.sleh.2018.11.009 Duncan S, Stewart T, Mackay L, Neville J, Narayanan A, Walker C, Berry S, Morton S (2018) Wear-Time Compliance with a Dual-Accelerometer System for Capturing 24-h Behavioural Profiles in Children and Adults. Int J Environ Res Public Health 15(7):1296. ttps://doi.org/10.3390/ijerph15071296 Frazer MT, Creaser A, Tatterton MJ, Daly-Smith A, Hall J (2024) Exploring children and young people’s experience of participating in citizen science–A qualitative evidence synthesis. PLoS ONE 19(7):e0304976. ttps://doi.org/10.1371/journal.pone.0304976 Gale EL, Manrai R, Caddick L, Murray A, Whalley HC, Smith D, Gardani M (2025) Co-production in sleep research: A scoping review of current practices and future directions. J Sleep Res. ttps://doi.org/10.1111/jsr.14476 Gariepy G, Danna S, Gobiņa I, Rasmussen M, De Matos G, Tynjälä M, Janssen J, Kalman I, Villeruša M, Husarova A, Brooks D, Elgar F, Klavina-Makrecka FJ, Šmigelskas S, Gaspar K, T., Schnohr C (2020) How Are Adolescents Sleeping? Adolescent Sleep Patterns and Sociodemographic Differences in 24 European and North American Countries. J Adolesc Health 66(6):S81–S88. ttps://doi.org/10.1016/j.jadohealth.2020.03.013 Ghotbi N, Pilz LK, Winnebeck EC, Vetter C, Zerbini G, Lenssen D, Frighetto G, Salamanca M, Costa R, Montagnese S, Roenneberg T (2020) The µMCTQ: An Ultra-Short Version of the Munich ChronoType Questionnaire. J Biol Rhythm 35(1):98–110. ttps://doi.org/10.1177/0748730419886986 Hagenauer MH, Perryman JI, Lee TM, Carskadon MA (2009) Adolescent Changes in the Homeostatic and Circadian Regulation of Sleep. Dev Neurosci 31(4):276–284. ttps://doi.org/10.1159/000216538 Hirshkowitz M, Whiton K, Albert SM, Alessi C, Bruni O, DonCarlos L, Hazen N, Herman J, Hillard A, Katz PJ, Kheirandish-Gozal ES, Neubauer L, O’Donnell DN, Ohayon AE, Peever M, Rawding J, Sachdeva R, Setters RC, Vitiello B, M. V., Ware JC (2015) National Sleep Foundation’s updated sleep duration recommendations: Final report. Sleep Health 1(4):233–243. ttps://doi.org/10.1016/j.sleh.2015.10.004 Hysing M, Harvey AG, Linton SJ, Askeland KG, Sivertsen B (2016) Sleep and academic performance in later adolescence: Results from a large population-based study. J Sleep Res 25(3):318–324. ttps://doi.org/10.1111/jsr.12373 Johnson D, Dupuis G, Piche J, Clayborne Z, Colman I (2018) Adult mental health outcomes of adolescent depression: A systematic review. Depress Anxiety 35(8):700–716. ttps://doi.org/10.1002/da.22777 Kocevska D, Lysen TS, Dotinga A, Koopman-Verhoeff ME, Luijk MPCM, Antypa N, Biermasz NR, Blokstra A, Brug J, Burk WJ, Comijs HC, Corpeleijn E, Dashti HS, De Bruin EJ, De Graaf R, Derks IPM, Dewald-Kaufmann JF, Elders PJM, Gemke RJBJ, Tiemeier H (2020) Sleep characteristics across the lifespan in 1.1 million people from the Netherlands, United Kingdom and United States: A systematic review and meta-analysis. Nat Hum Behav 5(1):113–122. ttps://doi.org/10.1038/s41562-020-00965-x Lucien JN, Ortega MT, Shaw ND (2021) Sleep and puberty. Curr Opin Endocr Metabolic Res 17:1–7. ttps://doi.org/10.1016/j.coemr.2020.09.009 Lunsford-Avery JR, Keller C, Kollins SH, Krystal AD, Jackson L, Engelhard MM (2020) Feasibility and Acceptability of Wearable Sleep Electroencephalogram Device Use in Adolescents: Observational Study. JMIR mHealth uHealth 8(10):e20590. ttps://doi.org/10.2196/20590 Lyall LM, Sangha N, Wyse C, Hindle E, Haughton D, Campbell K, Brown J, Moore L, Simpson SA, Inchley JC, Smith DJ (2020) Accelerometry-assessed sleep duration and timing in late childhood and adolescence in Scottish schoolchildren: A feasibility study. PLoS ONE 15(12):e0242080. ttps://doi.org/10.1371/journal.pone.0242080 Martin KB, Bednarz JM, Aromataris EC (2021) Interventions to control children’s screen use and their effect on sleep: A systematic review and meta-analysis. J Sleep Res 30(3):e13130. ttps://doi.org/10.1111/jsr.13130 Meredith-Jones KA, Haszard JJ, Graham-DeMello A, Campbell A, Stewart T, Galland BC, Cox A, Kennedy G, Duncan S, Taylor RW (2024) Validation of actigraphy sleep metrics in children aged 8 to 16 years: Considerations for device type, placement and algorithms. Int J Behav Nutr Phys Activity 21(1):40. ttps://doi.org/10.1186/s12966-024-01590-x Migueles JH, Rowlands AV, Huber F, Sabia S, Van Hees VT (2019) GGIR: A Research Community–Driven Open Source R Package for Generating Physical Activity and Sleep Outcomes From Multi-Day Raw Accelerometer Data. J Meas Phys Behav 2(3):188–196. ttps://doi.org/10.1123/jmpb.2018-0063 Morales-Ghinaglia N, Fernandez‐Mendoza J (2023) Sleep variability and regularity as contributors to obesity and cardiometabolic health in adolescence. Obesity 31(3):597–614. ttps://doi.org/10.1002/oby.23667 Olaithe M, Richardson C, Ree M, Hartung K, Wylde T, Bucks R (2024) Sleep in young people: What works now and where to? A meta-review of behavioural and cognitive interventions and lifestyle factors. Behav sleep Med 22(1):58–75. ttps://doi.org/10.1080/15402002.2023.2182305 Roenneberg T, Pilz LK, Zerbini G, Winnebeck EC (2019) Chronotype and Social Jetlag: A (Self-. Crit Rev Biology 8(3):54. ttps://doi.org/10.3390/biology8030054 Rowlands AV, Harrington DM, Bodicoat DH, Davies MJ, Sherar LB, Gorely T, Khunti K, Edwardson CL (2018) Compliance of Adolescent Girls to Repeated Deployments of Wrist-Worn Accelerometers. Med Sci Sports Exerc 50(7):1508–1517. ttps://doi.org/10.1249/mss.0000000000001588 Short MA, Gradisar M, Lack LC, Wright H, Carskadon MA (2012) The discrepancy between actigraphic and sleep diary measures of sleep in adolescents. Sleep Med 13(4):378–384. ttps://doi.org/10.1016/j.sleep.2011.11.005 Taylor DJ, Jenni OG, Acebo C, Carskadon MA (2005) Sleep tendency during extended wakefulness: Insights into adolescent sleep regulation and behavior. J Sleep Res 14(3):239–244. ttps://doi.org/10.1111/j.1365-2869.2005.00467.x Toftager M, Kristensen P, Oliver M, Duncan S, Christiansen L, Boyle E, Brønd J, Troelsen J (2013) Accelerometer data reduction in adolescents: Effects on sample retention and bias. Int J Behav Nutr Phys Activity 10(1):140. ttps://doi.org/10.1186/1479-5868-10-140 Toften S, Pallesen S, Hrozanova M, Moen F, Grønli J (2020) Validation of sleep stage classification using non-contact radar technology and machine learning (Somnofy®). Sleep Med 75:54–61. ttps://doi.org/10.1016/j.sleep.2020.02.022 Van Hees VT, Sabia S, Anderson KN, Denton SJ, Oliver J, Catt M, Abell JG, Kivimäki M, Trenell MI, Singh-Manoux A (2015) A Novel, Open Access Method to Assess Sleep Duration Using a Wrist-Worn Accelerometer. PLoS ONE 10(11):e0142533. ttps://doi.org/10.1371/journal.pone.0142533 Wang P, Zhang Y, Ma Y, Liang F, An Q, Xue H, Yu X, Lv H, Wang J (2019) Method for Distinguishing Humans and Animals in Vital Signs Monitoring Using IR-UWB Radar. Int J Environ Res Public Health 16(22):4462. ttps://doi.org/10.3390/ijerph16224462 Zhang J, Paksarian D, Lamers F, Hickie IB, He J, Merikangas KR (2017) Sleep Patterns and Mental Health Correlates in US Adolescents. J Pediatr 182:137–143. ttps://doi.org/10.1016/j.jpeds.2016.11.007 Additional Declarations The authors declare no competing interests. Supplementary Files AmbientTeensSupplementaryMaterial20260303.docx Ambient Teens Feasibility Study Supplementary Material Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Napping, sleep and affect in late- and short- sleeping adolescents: An actigraphy and daily diary study.#247543
• Baseline Moderators of a Transdiagnostic Adolescent Sleep and Circadian Health Intervention's Effectiveness at Improving Circadian Sleep Health: A Pilot Investigation#272193
• Adverse experiences in childhood and adolescence in relation to sleep health in adulthood: An overview of systematic reviews and a global evidence map.#610548
• Using digital measures to predict sleep impairment in adolescents with mood disorders treated with antidepressants.#1092211
• Critical Smoke Alarm Characteristics to Awaken Children From Stage 4 Sleep#881083