Hospital-Based Polysomnography May Overestimate Obstructive Sleep Apnoea Severity: Comparison of Hospital-Based and Home-Based Measurements with a Single-Lead Electrocardiogram Patch

Purpose: Obstructive sleep apnoea (OSA) is a global health concern, and polysomnography (PSG) is the gold standard for assessing OSA severity. However, the sleep parameters of home-based and in-laboratory PSG vary because of environmental factors, and the magnitude of these discrepancies remains unclear. Methods: We enrolled 125 Taiwanese patients who underwent PSG while wearing a single-lead electrocardiogram patch (RootiRx). After the PSG, all participants were instructed to continue wearing the RootiRx over the 3 subsequent nights. Scores on OSA indexes, namely the apnoea–hypopnea index, chest effort index (CEI), cyclic variation of heart rate index (CVHRI), and combined CVHRI and CEI (Rx index), were determined. The patients were divided into 3 groups based on PSG-determined OSA severity. The variables (various severity groups and environmental measurements) were subjected to mean comparisons and their correlations were examined by Pearson’s correlation coecient. Results: The hospital-based CVHRI, CEI, and Rx index differed signicantly among the severity groups. All 3 groups exhibited a signicantly lower percentage of supine sleep time in the home-based assessment relative to in the hospital-based assessment. Signicant positive correlations were noted between the variations in the supine percentage ( ∆ Supine%) and the OSA indexes. For the patients with high sleep eciency ( ≥ 80%), signicant correlations were observed between the ∆ Supine% and ∆ Rx index. Conclusion: The high supine percentage of sleep may cause OSA indexes’ overestimation in hospital-based PSG. Sleep recording at home with patch-type wearable devices may aid accurate OSA diagnosis. and supine percentage were higher than those at home and there are positive correlations between the variations in those indexes and supine percentage. Patch-type wearable devices may aid in more personalized obstructive sleep apnea severity determination. PSG


Introduction
Obstructive sleep apnoea (OSA) is a major health concern in modern society. A systematic review published in 2017 reported that OSA prevalence ranges between 9% and 38% in the general population.
(1) Moreover, OSA has been demonstrated to be associated with several comorbidities, including metabolic syndrome, cardiovascular diseases, and neurodegenerative diseases.(2, 3) Regarding OSA diagnosis, polysomnography (PSG) is the gold standard for determining the apnea-hypopnea index (AHI), which is used to classify OSA severity. However, PSG is complicated and inconvenient to implement. Patients typically undergo PSG with multiple leads on their bodies at a hospital sleep centre. The discomfort involved in the PSG itself can cause sleep disturbance. Moreover, relevant studies have indicated that the sleep parameters obtained using PSG could be underestimated or overestimated because of environmental factors or the rst-night effect. (4) Therefore, to assess OSA severity in contexts where the results are less likely to be affected by environmental factors, the American Academy of Sleep Medicine (AASM) has classi ed unattended home monitoring devices into 3 types. (5) The most common home sleep apnoea testing (HSAT) devices available are type-3 devices that include parameters for evaluating respiratory status, cardiac function, and pulse oxygen saturation (SpO2). Type-4 devices consider 1 or 2 of these parameters. Studies have applied wearable devices operated using various technologies, such as actigraphy, nger-based pulse oximetry, and single-channel electrocardiography, for the home-based recording of sleep parameters.(6-8) More recently, a biosensor that integrates an electrocardiography module and a 3-axis accelerometer was developed, demonstrating favourable reliability and accuracy in evaluating OSA severity. (9) Although various wearable devices have been used for OSA severity assessment, uncertainties remain regarding the differences between hospital-based PSG parameters and home-based sleep variables.
Several investigations have compared HSAT results and hospital-based sleep parameters. One study suggested that home-based AHI values are underestimated relative to hospital-based AHI values. (10) This disparity may be attributed to the lack of sleep stage measurement at home, which leads to overestimation of total sleep time. Another study reported night-to-night variability in the results of hospital-based PSG, with relatively weak correlations between test-retest AHI values. (11) Hospital-based AHI measurements might not accurately represent sleep status because they are easily affected by the different percentages of time spent in various sleeping positions depending on the scenario. (12) Consequently, uncertainty remains regarding the association between home-based and hospital-based measurements of OSA severity. Moreover, to the best of our knowledge, no studies have undertaken the acquisition of long-term sleep parameters or conducted in-depth evaluations based on variations in sleeping position.
To determine long-term home sleep parameters and prevent sleep disturbance caused by cumbersome instruments, the cyclic variation of heart rate index (CVHRI) can be used as a potential surrogate for screening OSA severity. This index is calculated according to the speci c heart rate alternation in progressive bradycardia when an apnoea event occurs and is followed by abrupt tachycardia on breathing resumption.(13) Several relevant studies have been performed to improve the algorithm and validate the associations between CVHRI and AHI.(14-16) CVHRI can also be directly determined by analysing single-lead electrocardiogram (ECG) signals. (17) The CEI is also a potential surrogate for OSA risk screening. Chest wall motion is directly affected by sleep respiratory events-that is, chest wall movement is reduced when an apnoea event occurs. Studies have asserted that sleep-disordered breathing events are characterised by chest wall distortion and paradoxical chest wall movement caused by the respiratory effort against airway obstruction.(18, 19) Hence, the CVHRI and CEI are potential alternative indexes for the non-invasive observation of sleep parameters over multiple days.
The primary objective of this study was to compare the data on sleep parameters obtained in overnight PSG at the hospital and over several days at home through a single-lead ECG patch with a 3-axis accelerometer (RootiRx), with the results expected to provide an in-depth understanding of how sleep position and the environment affect OSA severity. Moreover, we investigated changes in sleeping position in various sleep environments to determine the correlations between the percentage of sleep time spent in a supine position and OSA severity for both in-laboratory PSG and RootiRx.

Ethics
This study was conducted at the sleep centres of Shin Kong Wu Ho-Su Memorial Hospital (SKH; Taipei City, Taiwan) and Shuang Ho Hospital (SHH; New Taipei City, Taiwan). The study was approved by the institutional review boards of both hospitals (SKH: 20171003R; TMU-JIRB: N201709023), and written informed consent was obtained from all participants before any examination.

Study Population
We recruited patients with reported snoring or with suspected sleep-disordered breathing who were referred to the sleep centres of SKH and SHH between February 2018 and January 2019. The inclusion criteria were as follows: patients (1) aged between 18 and 80 years (2) who were not pregnant, (3) did not have a diagnosis of other cardiovascular diseases, and (4) had a total PSG recording time of >6 h. To reduce the possibility of OSA severity overestimation caused by short sleep time in the hospital setting, a large proportion of patients (103 of 125) with high sleep e ciency (SE) were recruited from the sleep centres to form 2 subgroups (SE ≥ 80% and SE ≥ 90%). The patients underwent PSG while wearing a wireless single-lead ECG monitoring patch (RootiRx, Rooti Labs, Taipei, Taiwan). After the PSG was completed, the patients were instructed to continue wearing the provided patch over the 3 subsequent nights to collect relevant sleep parameters under home sleep conditions. All the values for the hospitaland home-derived sleep parameters were used for further analysis and comparison.

PSG Results
PSG is a systematic process through which (1) physiological parameters are collected during sleep and (2) the underlying causes of sleep disorders are assessed on the basis of various physiological signals.
Notably, PSG is considered a standard method for diagnosing sleep-related breathing disorders, including OSA, central sleep apnea, and sleep-related hypoventilation or hypoxia (20). We obtained PSG recordings by using the Compumedics Grael PSG system (SKH) or the ResMed Embla N7000 and Embla MPR systems (SHH). We scored the sleep stages and respiratory events according to the updated standard diagnostic criteria and scoring guidelines of the AASM. (21,22) Licensed PSG technicians scored the results at both sleep centres, and these scores were con rmed by at least 2 other technicians to ensure accuracy. We determined the AHI value of each patient to classify them into the following 3 groups: no-tomild (AHI < 15 events/h), moderate (15 ≤ AHI < 30 events/h), and severe (AHI ≥ 30 events/h) OSA.

Home Sleep Recording
We obtained the home sleep parameters through observation with RootiRx. The technical details of this device and the de nition of the obtained sleep parameters, including the CVHRI, CEI, and combined CVHRI and CEI (Rx index), were documented in our previous study. (23) In the current study, the CVHRI, CEI, and Rx index were determined rst at the sleep centres through PSG and subsequently at home for 3 consecutive nights. The triaxial accelerometer in the device assessed the percentage of sleep time spent in different positions. All of the derived data were then separated into hospital and home data groups for comparison.

Statistical Analysis
All statistical analyses, the framework of which is presented in Figure 1, were conducted using SPSS software (IBM SPSS Inc., Chicago, IL, USA). First, we conducted the Shapiro-Wilk test to examine the normality of the continuous variables. The baseline characteristics of the patients in the OSA groups were compared, using one-way analysis of variance (normally distributed data) or the Kruskal-Wallis test (nonnormally distributed data) for the continuous variables and the chi-square test for the categorical variables. Subsequently, we performed the Student's t test (normally distributed data) or the Mann-Whitney U test (nonnormally distributed data) to compare the sleep parameters and positions obtained at the sleep centres and at home. The correlations between the variations in the percentage of sleep time spent in a supine position (∆Supine%) and the CVHRI (∆CVHRI), CEI (∆CEI), and Rx index (∆Rx index), were investigated through the Spearman rank correlation test. All tests were two tailed, and differences were considered signi cant at P < .05.

Sample Characteristics
A total of 125 patients were included. Table 1 presents the baseline characteristics of the patients according to OSA severity. In the sample, 33, 31, and 61 patients were classi ed as having no-to-mild OSA, moderate OSA, and severe OSA, respectively. No signi cant differences in age or sex were noted among the 3 groups. Regarding the participants' body pro les, a signi cantly higher body mass index (BMI) and higher neck circumference were observed in the moderate and severe OSA groups. Regarding the hypoxemia-related indicators, mean SpO 2 , minimum SpO 2 , and oxygen desaturation index (≥ 3%) were signi cantly lower in the severe OSA group than in the moderate OSA and no-to-mild OSA groups.   Data are expressed as means ± standard deviations.
All P values were derived from the (a) Mann-Whitney U test or (b) Student's t test depending on whether the data sets met the normality assumptions.

Variations in Sleeping Position and Sleep-Related Indexes in
Patients with High SE Table 3 presents the alterations in sleep parameters determined in the hospital and home settings in patients with high SE (> 80% in the hospital-based PSG). Notably, 103 patients had high SE, including 24, 24, and 55 patients in the no-to-mild, moderate, and severe OSA groups. Signi cant differences in the CVHRI, CEI, Rx index, and percentage of supine sleep time between the hospital-and home-based measurements were noted in the severe OSA group. In the moderate OSA group, the home-based Rx index and percentage of supine sleep time were signi cantly lower than those measured in the sleep centres. Moreover, in patients with SE > 90% (n = 46), the home-based measurements of the means of the Rx index and percentage of supine sleep time were signi cantly lower than the corresponding hospital-based measurements (Fig. 4). Data are expressed as means ± standard deviations.
All P values were derived from the (a) Mann-Whitney U test or (b) Student's t test depending on whether the data sets met the normality assumptions.
Correlations Between Sleep-Related Indexes and Sleeping Position in Patients with High SE Figure 5 presents the correlation between the ∆Supine% and ∆Rx index in patients with high SE. In patients with SE ≥ 80%, ∆Rx index was signi cantly positively correlated with ∆Supine% (r = 0.35, P < .01). In addition, in patients with SE ≥ 90%, ∆Rx index was signi cantly correlated with ∆Supine% (r = 0.29, P < .01).

Discussion
The present study compared OSA index values and the percentage of supine sleep time using RootiRx device in both hospital and home settings. Because patch-type wearable devices cause less interference to individuals' sleep than do PSG measurements, lower values for parameters such as the CVHRI, CEI, Rx index, and percentage of supine sleep time were noted in the home setting ( Table 2 and Fig. 2). Moreover, signi cant correlations were observed between the variations in the percentage of supine sleep time and the OSA index values (Fig. 3). In other words, greater changes in the sleep time spent in the supine position were correlated with greater changes in the OSA index values. Participants with SE ≥ 90% in hospital also exhibited higher Rx index values and higher percentages of supine sleep time in the hospital setting (Fig. 4). Furthermore, even in patients with SE ≥ 80% and SE ≥ 90% in hospital settings, signi cant correlations were observed between increased ∆Rx index values and ∆Supine% (Fig. 5), indicating that patch-type wearable devices such as RootiRx may cause less interference to sleeping position than does PSG even when these individuals exhibited high SE in hospital settings. According to our literature review, this is the rst study to investigate these correlations between variations in sleeping position and OSA indexes in hospital-and home-based data.
Higher OSA indexes were noted in the hospital than at home, indicating that in-laboratory PSG may overestimate OSA severity. This nding may be partially ascribed to environmental factors, such as the equipment, testing room, and bed, which may hinder changes in sleeping position.
Discrepancies between OSA severity measurements obtained in hospitals and at home have been documented. A study conducted in 1996 centred on the effects of PSG on sleeping position, suggesting that PSG may in uence the diagnosis of positional OSA. (24) In that study, 12 patients with positional OSA who had undergone standard PSG returned for 2 additional nights of study without the attachment of PSG leads. The mean percentage of supine sleep time (56%) was greater during the PSG night than during the non-PSG nights. A large-scale retrospective study (2019) on positional OSA treatment with the Sleep Position Trainer, a vibrating device, reported that the PSG apparatus caused an increase in the percentage of supine sleep time and may increase the measured OSA severity. (25) The median AHI decreased from 13.3/h to 10.3/h (P < .001), and 33% of the patients exhibited a change in OSA severity (AHI obtained in hospital settings vs adjusted AHI obtained at home). These outcomes support our ndings that PSG measurements may affect sleeping position and increase the percentage of sleep time in the supine position. Therefore, the effects of PSG equipment on sleeping position may lead to higher AHI values, leading to the overestimation of OSA severity.
The signi cant correlations between the ∆Supine% and differences in OSA indexes, including the ∆CVHRI, ∆CEI, and ∆Rx index, indicated that the increase in the percentage of supine sleep time and the corresponding increase in OSA severity might be a general pattern rather than being limited to speci c patient groups. Moreover, this nding has clinical relevance because if sleeping position is in uenced by the PSG apparatus and this causes signi cant overestimation of OSA severity, treatment strategies are likely to be affected. Some may argue that OSA severity as determined through PSG may be affected by other environmental factors. To address this concern, we analysed the sleep parameters in participants with high SE, in whom the possibility of OSA severity overestimation owing to sleep stage can be largely excluded. Because such patients had long sleep times in the hospital setting, the AHI values obtained from PSG and the OSA index values obtained from the RootiRx were not categorised according to short sleep time. In other words, in patients with high SE, the overestimation of OSA severity is more likely to be attributable to sleeping position than to alterations in total sleep time. Moreover, the ∆Rx index was signi cantly correlated with the ∆Supine% in the high SE groups. These results suggest that the overestimation of OSA severity in hospitals may be mainly due to patients' sleeping positions. Therefore, the home-based OSA index values likely represent the participants' actual OSA severity because they were not restricted by cumbersome PSG devices and could freely alter their sleeping position.
This study has some limitations. First, the RootiRx data set lacked sleep staging measurements obtained through electroencephalography. The RootiRx device determines the sleep stage of the wearer by using a validated algorithm, such as a fast Fourier transform and neural networks.(26-28) Although the accuracy of the predicted sleep stage and estimated total sleep time was approximately 85-90%, the arousal response or the precise percentages of rapid eye movement (REM) sleep and non-REM sleep could not be obtained. The associations between the rst-night effect, REM latency, and duration should be further explored. (29) Second, during RootiRx recording, the effects of environmental factors in the hospital and home sleep environments could not be controlled. Environmental factors include radiant temperature, air temperature, relative humidity, carbon dioxide concentration, illumination, and equivalent noise level. (30) Moreover, ECG signals are the mechanism of the RootiRx device. Thus, the CVHRI index could have been affected by abnormal heart rhythms, such as atrial brillation and ventricular tachycardia (with or without pacemaker implantation) and arrhythmia caused by any other type of cardiovascular condition. Hence, in patients with related heart diseases, the Rx index and CVHRI might not be accurate measures of OSA. In such patients, CEI may be a more suitable parameter for diagnosing OSA.

Conclusion
This study compared hospital-and home-based sleep parameters by using the RootiRx, a wearable device that uses a single-lead ECG patch. The current results provide evidence to support that hospitalbased PSG may overestimate OSA severity because patients spend a higher percentage of sleep time in the supine position in hospital settings. Home-based sleep recording with patch-type wearable devices may complement PSG in accurate OSA diagnosis.