ReSTech project on Xiaomi wearable devices for monitoring and detecting obstructive sleep apnoea: observational study protocol

Abstract

Introduction

Sleep-related breathing disorders have become a significant public health concern due to their negative impact on the population’s quality of life and overall health. Despite being underdiagnosed, their prevalence has increased in recent years, particularly in cases of obstructive sleep apnoea (OSA). Early diagnosis and detection of OSA are essential for timely treatment to mitigate the physical and health consequences. While polysomnography remains the gold standard for diagnosis, its limitations have led to the adoption of nocturnal polygraphy as an alternative for diagnosis. The scientific community is seeking devices that enable continuous monitoring of sleep status and other relevant parameters in this population. This study aims to analyse a wearable device as a complementary tool for monitoring health status and daily activity in people with potential OSA.

Methods and analysis

This observational and cross-sectional study will be conducted at the Sleep Respiratory Disorders and Home Ventilation Unit of a Hospital Álvaro Cunqueiro in Vigo. The aim is to recruit 246 participants who meet the inclusion criteria. Specific statistical methods will be employed to evaluate the accuracy and quality of the data collected by the Xiaomi Mi Smart Band 9.

Ethics and dissemination

This protocol study has been approved by the Pontevedra-Vigo Ourense Research Ethics Committee (process number 2024/260). All participants will sign a statement of informed consent. Study results will be disseminated in peer-reviewed journal articles.

Trial registration number

NCT06606691.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This observational study was conducted in the real-life settings of the participants.

• The recruitment of over 200 participants with suspected obstructive sleep apnoea (OSA) allows for a broad and representative sample.

• The study compares data from a consumer-grade wearable device with data from nocturnal polygraphy.

• The integration of biomedical and healthy lifestyle data enables a more comprehensive assessment of overall health status and daily activity in individuals with suspected OSA.

• Monitoring is conducted over a single day, providing a useful snapshot of participants’ data, though it may not capture long-term variability.

Introduction

Sleep-related breathing disorders (SRBDs) have become a significant public health concern due to their impact on the population’s quality of life.^{1 2} Although these disorders are often underdiagnosed, experts report a recent increase in their prevalence, linked to conditions such as long COVID-19.^{3 4} Epidemiological studies indicate that obstructive sleep apnoea (OSA), the most common and severe SRBD, affects approximately 2%–15% of the adult population, 2% of the paediatric population and over 25% of the older adults.^{4 5}

OSA has significant implications for individuals’ daily lives, both in terms of health and social and economic levels.^{1 2} These disorders disrupt the quality and quantity of sleep, often resulting in significant daytime consequences, including excessive daytime sleepiness, sleep fragmentation, fatigue, concentration deficits, mood disturbances and other related symptoms.^{6 7} Beyond these immediate effects, OSA is associated with an increased risk of developing chronic conditions such as type 2 diabetes, hypertension, stroke and respiratory diseases.^{6 8 9} Moreover, the literature highlights that the negative impact of OSA on health status increases the demand for medical care, specialised treatments and hospital admissions, generating a considerable economic burden.^{10 11}

The link between OSA and various chronic diseases can also significantly impact physical and mental well-being.^{12 13} These disorders can lead to physical or cognitive exhaustion, stress, low productivity and difficulties in social participation, among other issues.14,16 Furthermore, they can hinder the adoption of healthy habits, daily routines and the performance of everyday activities.¹⁷ In the long term, consequences can be more severe, potentially increasing the risk of temporary or permanent workplace accidents, with socioeconomic repercussions for the population.²

The high prevalence of OSA, along with its frequent underdiagnosis and implications for health status and quality of life, highlights the importance and necessity of increasing public awareness and improving screening strategies to enhance the detection and treatment of these disorders.^{18 19}

In this regard, the monitoring of OSA has seen significant advancements with the evolution of technological devices enabling home monitoring, such as nocturnal polygraphy tests.^{20 21} These devices provide a viable alternative to polysomnography (PSG), considered the gold standard for OSA diagnosis.^{22 23} Nocturnal polygraphy has demonstrated high accuracy in detecting respiratory events, positioning itself as a practical and cost-effective option for the diagnosis of individuals potentially affected by an OSA.24,26 However, these devices cannot perform real-time sleep staging, and their use is generally limited to one or two nights of monitoring, making prolonged follow-up challenging.^{20 27 28} The limitations of these devices have led to the search and research of more affordable, practical and appealing alternatives for users, enabling continuous and more detailed health monitoring.^{29 30}

Advances in information and communication technologies have established wearable devices as key roles in the current health field.^{29 31} These devices allow for continuous monitoring of various biomedical parameters, including sleep, daily activity, heart rate (HR) and blood oxygen saturation.^{32 33} Scientific evidence shows that activity trackers and smartwatches promote healthy lifestyle habits and improve self-management of health, promoting participatory health.^{30 34}

The continuous evolution and growing impact of these devices on the market and daily life have positioned them as a significant research objective.^{28 34} Several studies have explored the use of these devices for sleep monitoring, mainly in healthy populations.35,37 Overall, these studies have highlighted that wearable devices present certain limitations, tending to offer greater accuracy (ability to detect sleep-wake stages) and sensitivity (ability to detect sleep stages) than specificity (ability to detect wake stages).38,41 In this view, the technological field has prioritised improving the accuracy and quality of data generated by these devices, which are increasingly adopted by the general population.⁴²

The American Academy of Sleep Medicine (AASM) and the scientific community highlight the potential of these devices as preliminary tools for evaluating sleep before conducting more detailed clinical tests.^{43 44} In this context, they emphasise the importance of conducting studies to assess the accuracy and applicability of these devices, both in individuals with and without sleep disorders, particularly those with insomnia or OSA.^{43 44} Accordingly, the AASM, in collaboration with the American National Standards Institute and the Consumer Technology Association, has established a series of evaluation criteria to ensure the validity and reliability of these devices in classifying sleep stages.⁴⁵

Few studies have focused on researching these devices in populations with OSA. Moreno-Pino et al validated the Fitbit Charge 2 and Fitbit Alta HR in 55 participants with OSA, since Gruwez et al examined the Jawbone Up and Withings Pulse O2 devices in a study involving 36 participants with OSA.^{46 47} Both studies reported acceptable sensitivity but low specificity for these devices, using a small sample.^{46 47} Additionally, Byun et al found that the Fitbit Charge 2 and Galaxy Watch 2 tended to underestimate total sleep time (TST) in participants with OSA. However, they did not specify the accuracy of these devices.⁴⁸ The authors emphasise the need for home-based studies to validate the effectiveness of these devices in daily environments. Furthermore, the high prevalence of undiagnosed OSA highlights the importance of assessing device accuracy in this population to prevent misinterpretation of sleep status.46,48

Based on previous studies, this research aims to analyse the accuracy and application of the Xiaomi Mi Band (Xiaomi) in the free-living environment of people with a potential OSA. The accuracy of this device has been researched against the PSG in a heterogeneous sample of 45 adults, highlighting acceptable sensitivity but low specificity.^{49 50} Therefore, this is the first study specifically designed to research the applicability of this device in a population with potential OSA.

Thus, the main objective of this study is to evaluate the effectiveness of the Xiaomi Mi Band as a complementary tool for monitoring health status and daily activity in individuals with suspected OSA. The secondary objectives are: (1) to assess the accuracy, specificity and sensitivity of wearable devices, such as wristbands and smartwatches, in measuring blood oxygen saturation, HR and activity compared with nocturnal polygraphy; (2) to analyse the effectiveness of these devices in identifying participants’ profiles with potential OSA using unsupervised learning techniques and (3) to assess the impact and performance of an artificial intelligence (AI) model in detecting and predicting health status, activity and sleep patterns in individuals with suspected OSA.

Methods and analysis

Study design

This project is an observational and cross-sectional study. Different variables from participants with potential OSA will be observed and recorded to identify associations between data collected from wearable devices, nocturnal polygraphy and assessment tools. Additionally, patterns or biometric phenotypes associated with OSA will be explored.⁵¹ The study also involves continuous monitoring of the key variables over 1 year. During this time, participants will wear a Xiaomi device continuously for 24 hours, undergo a nocturnal polygraphy test for one night and complete various assessment tools to gather comprehensive data on their health status (see figure 1). Participants were not involved in the design of the study.

Figure 1. Design of the study.

This study protocol follows the Standard Protocol Items: Recommendations for Interventional Trials 2013 checklist for study protocols for clinical trials (online supplemental files data).

Study setting

The setting of this study will be the sleep respiratory disorders and home ventilation unit of the Álvaro Cunqueiro Hospital in Vigo, Spain. This unit, which is part of the pneumology department, is focused on the analysis, diagnosis and treatment of individuals with SRBDs and respiratory issues, with a particular focus on OSA. Various clinical methods are applied in this unit, including continuous nocturnal pulse oximetry, nocturnal polygraphy and the prescription and management of continuous positive airway pressure therapy, among others. The study began in late February 2025 and is expected to be completed by February 2026.

Participants

Sample selection will be performed through intentional non-probability sampling based on the established inclusion and exclusion criteria.

The inclusion criteria are as follows: (1) being at least 18 years of age or older and (2) attending the sleep respiratory disorders and home ventilation unit for the nocturnal polygraphy.

The exclusion criteria are as follows: (1) having significant health complications that hinder active participation in the study and (2) presenting skin hypersensitivity or a known allergy to the material used in the covers or straps of the wearable devices that will be used as one of the measurement instruments in the study.

Recruitment

Participants will be recruited from the SRBDs and home ventilation unit, where they will undergo a nocturnal polygraphy test to detect potential OSA, regardless of this study. The participants will be assisted by the pneumologist responsible for the unit and their clinical team. During the initial meeting with each participant, the pneumologist responsible for recruitment and follow-up will inform them about the possibility of participating in the study and the requirements for those who meet the inclusion criteria. During this meeting, the ethical procedures to ensure anonymity and data confidentiality will be clarified.

After the presentation of the main characteristics of the research, each potential individual will receive an information sheet to consult the study details and make a decision before undergoing the nocturnal polygraphy test. Once the users return to the unit for testing, the pneumologist can resolve any doubts or queries. If applicable, the informed consent document will be signed by both the responsible professional and the participant, confirming their final decision to take part in the study. Additionally, if a participant has reading or writing difficulties, a witness must be present throughout the entire process to ensure that the research team duly follows all ethical procedures.

Outcomes

The primary outcomes of this study consist of the classification and estimation of sleep parameters (min), including deep sleep, light sleep, rapid eye movement (REM) sleep and time wake after sleep onset, as recorded by the Xiaomi Mi Smart Band 9 in individuals with suspected OSA.

Secondary outcomes include the monitoring of daily physical activity using the Xiaomi Mi Smart Band 9, specifically step count, duration (in min) and distance (in m). Positional changes and body movements during sleep will also be assessed using nocturnal polygraphy. Additionally, HR (mean, maximum and minimum, in bpm) and blood oxygen saturation (mean, maximum and minimum, in %) will be collected using both the Xiaomi Mi Smart Band 9 and the nocturnal polygraphy.

Further secondary outcomes include variables related to sleep quality and sleep habits, collected through questionnaires and tests. These instruments include items on sociodemographic data and specific questions on sleep quality and quantity, as detailed in the Measurements subsection.

All outcomes will be collected for a single day for each participant, including one night of sleep monitoring. The collection of these outcomes will be conducted continuously over 12 months.

Sample size

This project aims to study approximately 263 individuals from different age groups and sexes who are suspected of having OSA. The power to detect clinically significant effects was calculated based on previous studies.^{52 53} A mean difference of 8.69 bpm (±13.85) is estimated between both measurement methods. Assuming a significance level of 0.05 and a statistical power of 95% (beta risk of 0.10), in a two-tailed test, a sample size of 47 participants is considered appropriate to detect a difference of 8.69 bpm or greater. Additionally, an SD of 30 for the differences between means has been estimated according to similar studies.

Measurements

Sociodemographic questionnaire

The sociodemographic questionnaire will be completed by the participants at the beginning of the study. It will collect sociodemographic data, health data and contemplated factors that may influence the quality and quantity of sleep. The items include year of birth, gender, height, weight, marital status, residential environment, dwelling unit, educational level, type of job, work time, health conditions, perceived stress level, health habits, sleep habits, physical activity level and medication.

Epworth Sleepiness Scale

The Epworth Sleepiness Scale is a tool used to assess daytime sleepiness.^{13 54} This self-administered questionnaire contains eight items that inquire about the likelihood of falling asleep in different everyday situations, such as reading or watching television. It evaluates various aspects of sleep, including amount, quality, duration, latency and efficiency.^{13 54} Each item is scored on a scale from 0 (indicating never falling asleep) to 3 (indicating a high probability of falling asleep). The total score can range from 0 to 24 points, with higher scores indicating greater levels of sleepiness and the potential for various sleep disorders.^{13 54}

Sleep diary

This questionnaire (see box 1) consists of several items related to duration, quality and quantity of sleep. It must be completed after sleeping with the devices used in this project. The items are related to bedtime and wake-up time, hours of sleep, sleep quality (scored on a Likert scale from 0 (very good) to 5 (very bad)), sleep onset (scored on a Likert scale from 0 (easily) to 3 (with difficulty)) and nocturnal awakenings.

Box 1. Sleep diary items.

The following items relate to your sleep pattern:

1. Time of devices colocations (hh:mm).

2. Lights-off time (hh:mm).

3. Lights-on time (hh:mm).

4. Bedtime (hh:mm).

5. Wake-up time (hh:mm).

6. Hours in bed (hh:mm).

7. Hours of sleep (hh:mm).

8. Sleep onset (easily/after a while/with difficulty).

9. Nighttime awakenings (number of times/duration/causes).

10. Sleep quality (very good, good, moderately good, bad, very bad).

Xiaomi Mi Smart Band 9

This wearable device is among the most popular globally due to its reasonable quality and affordability.⁵⁰ Several research studies have used this device in different populations, including older adults, workers with perceived stress and people with and without sleep disorders.^{50 55 56}

The Xiaomi Mi Smart Band 9 includes an accelerometer, gyroscope, 6-axis motion sensor, HR sensor and a blood oxygen level sensor.⁵⁷ The device features updated software that continuously records sleep (light sleep, deep sleep, REM sleep, wakefulness, TST, sleep onset and end times, blood oxygen saturation, breathing and naps), daily activity (steps, distance, activity time and calories) and HR (see table 1). In addition, the device calculates the stress level (classified using the terms relaxed, mild, moderate and high) and allows for menstrual cycle tracking.⁵⁷

Table 1. Nocturnal polygraphy and Xiaomi Mi Smart Band 9 variables.

Variable	Measure	Device(s)
Activity	Number of steps Movement (min) Intensity	Xiaomi Mi Smart Band 9
Sleep	Total sleep duration (min) Deep sleep duration (min) Light sleep duration (min) REM sleep duration (min) Bedtime (min) Wake-up time (min) Heart rate (bpm) Blood oxygen saturation (%)	Xiaomi Mi Smart Band 9
Heart rate	Minimum, average and maximum value (bpm)	Nocturnal polygraphy and Xiaomi Mi Smart Band 9
Blood oxygen saturation	Blood oxygen value Oxygen Desaturation Index	Nocturnal polygraphy and Xiaomi Mi Smart Band 9
Airflow	Apnoea-Hypopnoea Index Obstructive Apnoea-Hypopnoea Index Central Apnoea Index Oxygen Desaturation Index (3%) Duration and frequency of respiratory events	Nocturnal polygraphy
Thoracic and abdominal movements	Time in different body positions during sleep	Nocturnal polygraphy

The Xiaomi device connects via Bluetooth with its application, Mi Fitness, where the data are synchronised and displayed.⁵⁷ This application allows the export of biomedical parameters such as sleep, daily activity, HR and blood oxygen saturation data in CSV format.⁵⁷ This application (see figure 2) enables data to be transmitted to the device’s cloud. However, this option has been disabled for this study to ensure that the data remain exclusively accessible to the research team at all times.

Figure 2. Wearable device synchronisation process. SFTP, Secure File Transfer Protocol.

Thus, this study will record the device’s data using a custom-developed application that allows us to access the database files of the Mi Fitness application. These files, stored in SQLite format, are named ‘fitness_data’ and ‘fitness_summary’. Once the files are accessed, they are securely transferred via Secure File Transfer Protocol (SFTP) to a server, where the data are processed for analysis and visualisation.

Nocturnal polygraphy

Nocturnal polygraphy is a portable clinical tool for diagnosing OSA.^{58 59} In this protocol study, nocturnal polygraphy will be performed using cardiorespiratory polygraph Embletta (Natus Medical Incorporated). This device (see table 1) records parameters related to breathing, including nasal and oral airflow through pressure sensors, body movements, blood oxygen saturation via a pulse oximeter, HR and body position during sleep using position sensors.^{58 59}

Polygraphy parameters will be scored by the criteria established by AASM,⁶⁰ and recordings will be exported in CSV files and the European Data Format+format.⁶¹ Data will be interpreted by a specialised pneumologist and their team. The following indices will be calculated: Apnoea-Hypopnoea Index (AHI), Obstructive Apnoea-Hypopnoea Index, Central Apnoea Index, number of apnoeas and hypopnoeas, and Oxygen Desaturation Index (ODI 3%). Additionally, the average, maximum and minimum values of blood oxygen saturation, HR and the effects of body position will be collected.

Procedure

On the day of the nocturnal polygraphy test, participants will complete a sociodemographic questionnaire and the Epworth Sleepiness Scale through the Research Electronic Data Capture Consortium⁶² with the assistance of a technical team if needed. Once this is completed, the technical team will explain the correct procedure for using the nocturnal polygraphy and the Xiaomi Mi Smart Band 9, making sure that the participants have understood the process. Thus, participants will receive the nocturnal polygraphy, the Xiaomi device and a manual with detailed instructions for proper use.

Participants will sleep with both devices for one night. After the test at home, they will fill out a sleep diary, in which they should add the time of position of both devices, lights on and off time and time in bed. Once the devices are returned to the unit, the technical team will simultaneously synchronise the data from both devices, recording the synchronisation time. Data from the Xiaomi device will be synchronised using the self-developed application (see figure 2). The phone with the software will be located in the SRBDs and home ventilation unit.

Moreover, the data will be pseudonymised at the time of collection, and the confidentiality of the information and the anonymity of participants will be maintained throughout the study. At the end of the project, each participant’s data will be stored for future studies if they provide consent.

Statistical methods

The statistical analysis of the data will be performed using R software (V.4.1.2; R Foundation for Statistical Computing) and Python software (V.3.13.0; Python Software Foundation). This study will conduct various types of analyses based on the objectives outlined.

The process of data analysis will involve cleaning/preprocessing, descriptive statistics and processing to extract useful information for decision-making. Preprocessing will correct errors to avoid bias, descriptive statistics will summarise key patterns and processing will extract actionable insights.

The paired Student’s t-test will be used to compare the means of different sleep times of interest. Additionally, the Bland-Altman method will be employed to assess similarities between the two devices for each of the study variables (such as blood oxygen saturation, respiration, HR and activity).⁶³ A positive bias indicates that the device tends to overestimate a variable compared with the gold standard. A negative bias indicates that the variable is overestimated. Point estimates and their 95% CIs will be calculated. Furthermore, concordance analyses such as Cohen’s Kappa coefficient will be used with the following ranges: 0–0.2 (slight), 0.21–0.40 (fair), 0.41–0.60 (moderate), 0.61–0.80 (substantial) and >0.80 (almost perfect).⁶⁴

Two AI approaches based on machine learning will be used, supervised learning and unsupervised learning.^{65 66} In the supervised learning approach, the dataset will comprise data extracted from the devices, health and sociodemographic information, and data from the sleep diary and Epworth Sleepiness Scale. These results, based on the AHI, will serve as labels for predictive model training. The aim will be to develop predictive models that can support the creation of a new Intelligent Clinical Decision Support System to aid in participant screening and prioritisation based on severity. Response variables will be discretised into categories (eg, AHI<15, 15<AHI<30 and AHI>30) following international consensus guidelines on OSA. The dataset will be divided into training, validation and testing subsets, with model performance evaluated using Receiver Operating Characteristic (ROC) curves and area-under-the-curve metrics.

In the unsupervised learning approach, the dataset (comprising information derived from Xiaomi Mi Smart Band 9 and the other assessment tools) will be analysed using clustering algorithms to group participants based on similarities. This method does not require a response variable and aims to identify meaningful patient clusters. Statistical tests, such as χ² or Mann-Whitney U, will be used to assess differences between groups and evaluate the utility of data collected via wristbands. Additionally, once clusters are consolidated, polygraphy results (eg, AHI or hypoxic burden) may be incorporated to determine whether significant differences exist between clusters, providing further insights into the utility of wearable device data in clinical contexts.

Patient and public involvement statement

Patients or members of the public were not involved in the design and conduct of this study.

Discussion

The high prevalence and underdiagnosis of SRBDs, particularly OSA, highlight the need for strategies aimed at the early detection of these sleep disorders. Additionally, it is essential to monitor the health status of individuals with OSA while fostering greater awareness of the importance of sleep in this population. In this context, wearable devices emerge as valuable tools to complement standardised clinical tests such as nocturnal polygraphy for diagnosing these disorders.

This study is the first to evaluate the accuracy of the Xiaomi Mi Band in detecting and estimating sleep in people with suspected OSA. The findings from this research may contribute to a more precise use of this device and its potential application in clinical practice as a supportive tool. Furthermore, it highlights the role of wearable technology in advancing participatory health initiatives and driving technological transformation within the health context.

Ethics and dissemination

This study protocol was approved by the Pontevedra-Vigo Ourense Research Ethics Committee, under the number 2024/260. In addition, this protocol was registered in the Clinical Trials Protocol Registration and Results system on 23 September 2024, available at https://clinicaltrials.gov/study/NCT06606691.

The informed consent process will be applied to each participant. Participants will receive verbal and written information regarding the study’s characteristics and the implications of their participation. They will be provided with an information sheet to review the details and ask any questions they may have. Once it has been confirmed that they fully understand the information provided, they will decide whether to participate and, if they agree, will provide consent by signing the informed consent document.

The principal researcher will maintain the confidentiality of all data collected during the study. The research team complies with Regulation (EU) 2016/679 (General Data Protection Regulation) and national data protection laws (Organic Law 3/2018). Data processing will be overseen by the institutions conducting the research, with the support of data protection officers, who will implement secure protocols. Personal data will remain within the hospital information systems, protected through pseudonymisation techniques, and will not be transmitted outside the project scope. In addition, wearable device data will be captured via Bluetooth using specific software and stored on offline systems, ensuring data security.

On study completion, all pseudonymised data will be stored only with participant authorisation. The research team is committed to publishing the findings in high-impact journals, ensuring participants’ anonymity and reporting both positive and negative results.