ADVERTISEMENT

Issue Navigator

Volume 15 No. 05
Earn CME
Accepted Papers





Scientific Investigations

Validation of the MediByte Portable Monitor for the Diagnosis of Sleep Apnea in Pediatric Patients

Ahmed I. Masoud, BDS, MS, PhD1; Pallavi P. Patwari, MD2,3; Pranshu A. Adavadkar, MD2; Henry Arantes, RPSGT2,3; Chang Park, PhD4; David. W. Carley, PhD4
1Department of Orthodontics, Faculty of Dentistry, King Abdulaziz University, Jeddah, Kingdom of Saudi Arabia; 2University of Illinois Hospital and Health Sciences System, University of Illinois College of Medicine, Chicago, Illinois; 3Rush Children's Hospital, Rush University Medical Center, Chicago, Illinois; 4Departments of Biobehavioral Health Science, Medicine and Bioengineering, University of Illinois, Chicago, Illinois

ABSTRACT

Study Objectives:

Polysomnography (PSG) is considered the gold standard in the diagnosis of sleep apnea. In pediatric patients, because of limited availability and access to laboratory-based PSG, there can be significant delays in the diagnosis and management of sleep apnea that can result in progressive associated comorbidities. The main objective of the current study was to test the diagnostic value of a portable sleep monitor (PM), the MediByte, in comparison with laboratory PSG in pediatric patients wearing both setups simultaneously.

Methods:

A consecutive series of pediatric patients referred to the University of Illinois Sleep Science Center wore the MediByte during simultaneous PSG. The apnea-hypopnea index (AHI) was calculated for PSG and both manual and autoscoring functions of the PM. Pearson correlation and Bland-Altman plots were assessed.

Results:

A total of 70 patients successfully completed simultaneous PSG and PM studies (median age 10.8 years). The AHI obtained both manually and automatically scored PM studies strongly correlated with the AHI obtained from the PSG (r ≥ .932, P < .001). The oxygen saturation obtained by the PM showed significant correlation with that obtained by PSG among children aged 12 to 17 years (P < .001), but not among children aged 7 to 11 years (P ≥ .24). The sensitivity and specificity for detection of severe sleep apnea diagnosed by PSG (AHI ≥ 10 events/h) using both PM scoring methods was very high (> 93% for both).

Conclusions:

Although PSG is still recommended for the diagnosis of sleep apnea, PMs can play a valuable role in diagnosing moderate and severe sleep apnea, especially in older pediatric patients.

Commentary:

A commentary on this article appears in this issue on page 685.

Citation:

Masoud AI, Patwari PP, Adavadkar PA, Arantes H, Park C, Carley DW. Validation of the MediByte portable monitor for the diagnosis of sleep apnea in pediatric patients. J Clin Sleep Med. 2019;15(5):733–742.


BRIEF SUMMARY

Current Knowledge/Study Rationale: Obstructive sleep apnea (OSA) afflicts millions of children and leads to disrupted sleep as well as cognitive, behavioral, and developmental disturbances. In-laboratory polysomnography (PSG) is the standard test for diagnosing OSA. In pediatric patients, because of limited availability and access to PSG, there can be significant delays in the diagnosis and management of OSA. This highlights the urgent need for alternative diagnostic methods validated for use in pediatric patients.

Study Impact: This is the first study to systematically evaluate the MediByte, a portable sleep monitor, in relation to PSG in pediatric patients. Using portable sleep monitors may improve early detection and intervention for pediatric OSA, which can minimize progressive associated comorbidities.

INTRODUCTION

Children are often overlooked in studies of obstructive sleep apnea (OSA), yet OSA affects 1.2% to 5.7% of children.13 Currently, with the increase in childhood obesity, pediatric sleep apnea has become an even bigger concern.3 Like adults, children with OSA experience recurrent partial or complete upper airway obstruction during sleep,4 resulting in many comorbidities which may be prevented by diagnosing and managing OSA in a timely manner.3

Overnight attended in-laboratory polysomnography (PSG) remains the gold standard for OSA diagnosis, but it is not without disadvantages. PSG is technically complex to perform, expensive, time consuming, and requires the patient to be at the laboratory.3,5 An alternative OSA diagnostic approach is to use a portable monitor (PM). The American Academy of Sleep Medicine (AASM) has approved the use of PMs as an alternative to PSG in certain situations.5

PMs have been used for sleep testing in adults because they can be used at home, are less expensive, quicker to deploy, more accessible, and as accurate as laboratory PSG in certain applications.5 Depending on the number and type of recording channels, there are four types of PMs.6 The AASM recommends that at minimum a PM should record airflow, respiratory effort, and blood oxygenation; this level of monitoring is found among type 3 PMs.5

In adults, several authors have demonstrated agreement between a PSG and type 3 PMs for diagnosing OSA, with very high sensitivity and specificity values especially for moderate and severe OSA. This agreement was found whether both methods were used simultaneously or at different times and in different settings. Diagnostic accuracy increased when manual scoring was used rather than relying on automatic scoring algorithms alone.79 However, such PM validation studies are scarce in pediatric patients when compared to adults.3,6,1012 The pathophysiology and management of OSA differ between adults and children, so validation studies based on adults cannot be extrapolated to children.3 The aims of the current study were: (1) to test the diagnostic value of a type 3 PM, the MediByte, in comparison with laboratory PSG in pediatric patients wearing both setups simultaneously, and (2) to compare the diagnostic value of manual versus automatic scoring of PM data in pediatric patients.

METHODS

A consecutive series of pediatric patients between the ages of 7 and 17 years referred to the University of Illinois at Chicago (UIC) Sleep Science Center for PSG were approached by the primary investigator who conducted the informed consent process. Participants were classified into two groups according to age: group 1 (7 to 11 years) and group 2 (12 to 17 years). Exclusion criteria included: (1) craniofacial anomalies, (2) neuromuscular disorders, (3) pregnancy (urine human chorionic gonadotropin test on participants with childbearing potential), (4) inability to tolerate two nasal cannulas in the nostrils or four effort bands on the torso, and (5) fewer than 4 hours of total sleep time (TST) demonstrated by PSG. All study procedures were reviewed and approved by the Institutional Review Board of the University of Illinois at Chicago. Power analysis revealed that a sample size of 17 would be sufficient for a two-tailed alpha error of .05 and a beta error of .2.

During the night, participants wore the PM while simultaneously undergoing a PSG. PSGs were performed in standard clinical fashion under supervision of registered polysomnographic technologists (RPSGTs) at the UIC Sleep Science Center. Each PSG comprised recording of electroencephalogram (F3/M2, F4/M1, C3/M2, C4/M1, O1/M2, O2/M1), chin and anterior tibialis electromyogram, bilateral referential electro-oculogram, single-lead electrocardiogram, oronasal airflow by thermistor and nasal pressure transducer, thorax and abdomen movement by respiratory inductance plethysmography (RIP), peripheral oxygen saturation (SpO2) by pulse oximeter, and acoustic recording for snoring. All signals were acquired, processed and stored using the ALICE5 digital system (Philips Respironics, Murrysville, Pennsylvania, United States). All PSGs were staged and scored using AASM guidelines13 by a single RPSGT with final review and interpretation by the pediatric sleep medicine physician.

The PM used was the MediByte (Braebon Medical Corporation, Canada), which is a US Food and Drug Administration-approved type 3 PM. It consisted of two effort bands (chest and abdomen) to measure respiratory effort, a nasal cannula pressure transducer to measure airflow, a finger pulse oximetry sensor to measure oxygen saturation and heart rate, a microphone to record snoring, and a body position sensor. Pediatric size effort bands and finger pulse oximetry sensors were used as needed. The effort bands were also used to hold the PM in place at the midsternum. Midsternal placement was more comfortable and allowed for reliable body position detection. Before each use the device was configured by installing a new battery and entering the patient's information and assigned code. Once acquired, PM data were downloaded and analyzed using the MediByte software (version 8.1, Braebon Medical Corporation, Canada). The studies were scored automatically by the software and again manually by an RPSGT. Studies were considered failed if they did not meet published acceptable standards for minimum duration (≥ 4 hours) of interpretable signal.14

Before any PM scoring was done two steps were executed. These steps were undertaken by the authors to reduce errors in scoring, especially in automatic scoring, that occurred during trial tests. The first was to adjust the study end time “lights on” mark to exclude inadvertent extended recording which would increase the total recording time (TRT). If the mark occurred before the recordings ended, this mark was left as is. If the mark occurred after all the recordings ended, the mark was moved back in time to the last point that had both SpO2 and airflow recordings, as long as no signal returned. The second step was to identify “bad data.” The MediByte software marked recordings with SpO2 artifacts as “bad data.” However, the recording time with bad data was not excluded from TRT and the software attempted to score apneas yet avoided scoring hypopneas that rely on oxygen desaturation. Not excluding these “bad data” can underestimate the apnea-hypopnea index (AHI) if hypopneas are not scored during these periods. In an attempt to compare scorable data between the PSG and the PM, all SpO2 artifacts were marked as “bad data” and excluded from the study. The number of minutes with bad data and the number of minutes with sustained oxygen saturation less than 60%, which we also considered bad data, were added together and recorded. If after bad data exclusion the remaining TRT was less than 4 hours, then the study was designated as a “failure” and its results were not included in comparisons. Moreover, excluding bad data might also underestimate AHI since bad SpO2 might be related to movement associated with apneas which will not be scored if the bad data is excluded. Therefore, automatic scoring results before bad data exclusion were also recorded as a means to compare both methods of automatic scoring.

Automatic PM scoring settings were set to 90% airflow reduction for 10 seconds to score obstructive, central, and mixed apneas using the standard AASM breathing effort. Scoring a hypopnea was set at 30% airflow reductions and 3% oxygen desaturation for 10 seconds. The AASM recommends respiratory events to last for two breaths, instead of 10 seconds, to score an apnea or a hypopnea in children age 12 years or younger. However, the MediByte software did not have a function to score based on number of breaths and hence the 10-second duration was used for PM automatic scoring of all participants.

Manual PM scoring was performed by the same RPSGT who scored the PSGs. The RPSGT was blinded to the results of the PSG. Because arousals could not be documented in PM recordings, arousals were not used in the definitions of central apneas or hypopneas as was the case in the PSG. Except for these definitions based on arousals, the RPSGT used the same AASM criteria to score the PSG events. All manually scored PM studies were reviewed by the same sleep physician who reviewed the PSG. Similar to the RPSGT, the sleep physician was blinded to the results of the PSG. In addition to SpO2 artifacts marked as bad data by the primary investigator, the sleep physician marked additional bad data if airflow cessation due to cannula dislodgment was anticipated during manual scoring. If the sleep physician came across cessation of airflow for 30 seconds or more without other physiological changes (oxygen desaturation, heart rate change, or paradoxical breathing), these minutes were also marked as bad data.

Statistical analysis was performed using IBM SPSS version 20. All variables were tested for normality and symmetry using the Shapiro-Wilk test for normality and nonparametric tests were performed where appropriate. The AHI was calculated by dividing the total number of apneas and hypopneas by TST for the PSG and TRT for the PM. Pearson correlation coefficients were calculated between PSG and PM estimates of AHI; with weak, moderate, and strong associations demonstrated by correlation coefficients of ≥ .1 and < .3, ≥ .3 and < .7, and ≥ .7 respectively. Statistical significance for all analyses was noted at α = .05. Statistical analysis was done three times; PM automatically scored, PM automatically scored with bad data excluded, and PM manually scored. For each type of PM scoring, correlations between PSG and PM were calculated for AHI, SpO2, oxygen desaturation index (ODI), and heart rate (HR). Pearson correlation was used regardless if the data were not normally distributed because here we are looking for a linear relationship and proposing PM as a replacement to PSG. Separate correlational analyses also were performed for the two age groups. Bland-Altman analysis was used to estimate the error associated with the PM by considering PSG as the gold standard. For PM classification prediction, PM AHI receiver operating characteristic (ROC) curves were plotted once after classifying OSA severity as mild (AHI 1.5 to < 5 events/h), moderate (AHI 5 to < 10 events/h), or severe (AHI ≥ 10 events/h), and again using three distinct PSG AHI thresholds: AHI ≥ 1.5, AHI ≥ 5, and AHI ≥ 10 events/h. To further evaluate classification prediction, cross tabulation was used to calculate sensitivity, specificity, positive predictive values, negative predictive values, and accuracy.15

RESULTS

A total of 95 participants had simultaneous PSG and PM studies. Of 95 participants 10 (10.5%) had to be excluded: 2 because they had fewer than 4 hours of TST in the PSG, and 8 because the participants could not tolerate the extra cannulas or effort bands. Furthermore, there were 15 failed PM sleep studies (< 4 hours of reliable SpO2 in PM studies) of the remaining 85 participants (17.6%), leaving 70 PM studies to be analyzed and compared to PSG (43 in age group 1 and 27 in age group 2). Shapiro-Wilk test for normality showed the data to be not normally distributed (P < .05) for all variables; hence, nonparametric statistics were used. Table 1 shows the medians and interquartile ranges (IQR) of AHI and clinical characteristics of successful, excluded, and failed PM studies. Table 2 shows sleep characteristics of the 70 successful PM studies automatically scored, automatically scored with bad data excluded, and manually scored compared to PSG displayed in medians and IQRs. There were 25 participants with mild OSA (median AHI = 2.8 events/h), 13 participants with moderate OSA (median AHI = 6.8 events/h), and 9 participants with severe OSA (median AHI = 44.3 events/h) diagnosed using the PSG.

Comparing AHI and clinical characteristics of successful PM studies with participants with excluded and failed PM studies.

jcsm.15.5.733.t01.jpg

table icon
Table 1

Comparing AHI and clinical characteristics of successful PM studies with participants with excluded and failed PM studies.

(more ...)

Sleep characteristics of automatically scored, automatically scored with bad data excluded, and manually scored successful PM studies compared to PSG.

jcsm.15.5.733.t02.jpg

table icon
Table 2

Sleep characteristics of automatically scored, automatically scored with bad data excluded, and manually scored successful PM studies compared to PSG.

(more ...)

Averages and Pearson correlation between PSG and PM for AHI, SpO2, ODI, and HR are shown in Table 3. Figure 1 shows the Pearson correlation for AHI between PSG and the different scoring modalities of the PM. To further understand desaturations, data for both age groups are displayed in Table 4.

Portable monitor averages and Pearson correlation.

jcsm.15.5.733.t03.jpg

table icon
Table 3

Portable monitor averages and Pearson correlation.

(more ...)

Pearson correlation between the apnea-hypopnea index from the polysomnography (AHI PSG) and the portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right).

jcsm.15.5.733a.jpg

jcsm.15.5.733a.jpg
Figure 1

Pearson correlation between the apnea-hypopnea index from the polysomnography (AHI PSG) and the portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right).

(more ...)

Portable monitor means and Pearson correlation for both age groups.

jcsm.15.5.733.t04.jpg

table icon
Table 4

Portable monitor means and Pearson correlation for both age groups.

(more ...)

Bland-Altman analysis plots to estimate the error associated with the PM are displayed in Figure 2. The mean errors were +0.04 (95% confidence interval [CI] −14.4 to +14.5), −0.02 (95% CI −13.2 to +13.2), and −2.5 (95% CI −15.8 to +10.7) for automatic scoring, automatic scoring with bad data excluded, and manual scoring of the PM respectively. The Bland-Altman plots also showed biases. Linear regressions demonstrated that the mean AHI explained a significant proportion of variance in the AHI difference for all methods of PM scoring (P < .001). As the mean AHI increased so did the AHI difference between PSG and PM (PM AHI decreased) indicating proportional bias. To better understand the bias, participants with average AHI > 22 events/h were excluded from the plots and the Bland-Altman plots were redone. The corrected Bland-Altman plots for the remaining 63 participants are displayed in Figure 3. The mean errors for the corrected Bland-Altman plots were +2.17 (95% CI −3.7 to +8.0), +1.92 (95% CI −4.1 to +7.9), and −0.59 (95% CI −5.5 to +4.3) for automatic scoring, automatic scoring with bad data excluded, and manual scoring of the PM, respectively. Results pertaining to classification prediction; ROC curve analyses and cross tabulation, are shown in Table 5 and Table 6.

Apnea-hypopnea index (AHI) Bland-Altman analysis for portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right). PSG = polysomnography

jcsm.15.5.733b.jpg

jcsm.15.5.733b.jpg
Figure 2

Apnea-hypopnea index (AHI) Bland-Altman analysis for portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right).

(more ...)

Cross-tabulation results of PSG and PM using AHI ≥ 1.5, AHI ≥ 5, and AHI ≥ 10 events/h.

jcsm.15.5.733.t05.jpg

table icon
Table 5

Cross-tabulation results of PSG and PM using AHI ≥ 1.5, AHI ≥ 5, and AHI ≥ 10 events/h.

(more ...)

Cross-tabulation results of PSG and PM using mild, moderate, and severe sleep apnea definitions.

jcsm.15.5.733.t06.jpg

table icon
Table 6

Cross-tabulation results of PSG and PM using mild, moderate, and severe sleep apnea definitions.

(more ...)

DISCUSSION

The gold standard for sleep apnea diagnosis is to perform PSG in the laboratory. However, sleeping in an unfamiliar environment can materially affect sleep behaviors, and this concern increases when patients are children. Additionally, making the trip to have a sleep study performed in the laboratory is a concern for pediatric patients, because a legal guardian must be present during the study. Moreover, there can be significant delays in diagnosis and treatment of sleep apnea in pediatric patients because of limited availability and access to PSG and shortage of sleep laboratories with pediatric expertise. These factors, in addition to the PSG being complex, expensive and time consuming, highlight the urgent need for PMs validated for use in pediatric patients.1,16 This is the first study to systematically evaluate the MediByte, a type 3 PM, in relation to laboratory PSG in pediatric patients.

Our sample included 70 successful PM studies, which is the largest sample to date compared to similar previous studies in pediatric patients.3,10,11 A very large portion of the sample was black or Hispanic, which represents the general population of patients at the UIC (Table 1). In this study, exclusions and failures were separated. Excluded studies were done so by no fault of the PM and 8 of 10 of these exclusions in our study were because the participants could not tolerate the extra PM attachments. Examining clinical characteristics of excluded studies in Table 1 shows that this group had a median body mass index (BMI) z-score of 2.4, compared to median BMI z-score of 1.9 in the participants with successful studies. This might be an indication that the reason this group could not tolerate the extra PM attachments was because the patients were overweight. The median PSG AHI in the excluded studies was 2.8 which was, interestingly, identical to the median PSG AHI of the successful PM studies, suggesting that excluding these participants did not bias the sample by removing potentially more severe sleep apnea cases.

Failed studies, however, are due to failure of the PM recording acquisition process. This can range anywhere from the device not being turned on, insufficient battery life, or data loss all the way to poor airflow, SpO2, or RIP recordings. The main reason for failure in all of our PM studies was there being fewer than 4 hours of good SpO2 recording. Similarly, two previous studies used type 2 PM in children aged 5 to 12 years and reported poor oximetry as the most common reason for failure.16,17 Studies where type 3 PM was used simultaneously with PSG do not list the most common reason for failure.3,10,11 During the study, even though a pediatric finger pulse oximeter was used, the primary investigator noticed more failed studies in smaller-sized participants with small fingers. Examining the height, weight, and BMI z-score of failed studies in Table 1 shows that these values were on average less than they were for successful studies. This might confirm the observation of failed studies being more common in smaller participants.

Initially in our study, there were 9 of 34 failed studies (26.5%). Later, with the decision to add extra tape when applying the finger pulse oximetry sensor, there were 6 of 36 failed studies (16.7%). We recommend using extra tape especially when using the PM in patients with small fingers and in younger patients because they tend to move more and are more prone to PM failure.6 Contrary to what we found in children, other authors have reported that obesity in adults is associated with more home sleep study failures.14 Yet these reports are based on unattended sleep studies at home.

It is important to point out that failures due to poor nasal flow in our study, and probably in other studies where PSG and PM are studied simultaneously, are underestimated. When the RPSGT noticed poor or interrupted airflow in the PSG recording, the PSG cannula was adjusted and, consequently, the PM cannula was also adjusted during the process. Additionally, the primary investigator was trained to set up and apply the PM. If the parent had to set up the device at home, we would anticipate more overall failures. For these two reasons (RPSGT adjusting the cannula and the primary investigator setting the PM), we expect the 17.6% failure rate would be higher had the PM been used at home. When using the PM in pediatric patients at home, we recommend, in addition to extra tape, that parents be properly trained and that they go in and make sure the attachments are in place during the night if possible.

We reported a 17.6% failure rate, which is comparable to that of other studies. Lesser et al.11 used Apnealink, a type 3 PM, on a group of children with obesity simultaneously with PSG and reported a failure rate of 13.8%. However, this is an underestimation of their failure because failure was defined by a study having fewer than 2 hours of usable data.11 In our study, if we count only studies having fewer than 2 hours of usable data as failures, we would only have 3 of 85 failed studies (3.5%). Similarly, Massicotte et al.3 used Apnealink simultaneously with PSG. They defined failure by fewer than 4 hours of TRT and reported 22% failed studies.

We compared the PSG to three scoring methods in the PM: automatic scoring, automatic scoring with bad data excluded, and manual scoring. Pearson correlation showed a strong significant correlation between the AHI from the PSG and the AHI from the PM in all three scoring methods (r ≥ .932, P < .001). The correlation was also strong for HR and ODI (r ≥ .854, P < .001), but only moderate correlation was found for oxygen saturation (r ≤ .364, P < .05). Examination of the correlations for the two age groups separately revealed that the correlations remained strong for AHI, ODI, and HR. However, for SpO2 the correlation remained significant for age group 2 (r ≥ .697, P < .001) but not for age group 1. The correlation coefficient for SpO2 went down to r ≤ .183 and P ≥ .240. We previously stated that failed studies because of poor SpO2 recording were more common in smaller patients. This finding here shows that even in successful PM studies in age group 1, the SpO2 was not as reliable as in age group 2.

Bland-Altman analysis for AHI was done for all three scoring methods using the PSG as the gold standard. The mean error for both automatic scoring methods was almost zero, whereas it was −2.5 for the manual scoring (Figure 2). A close examination of the plots shows that in all three analyses, but more so in the automatic scorings, there was a tendency for the PM AHI to be larger than the PSG AHI when the average AHI was below 22 events/h. This tendency then flipped when the average AHI was above 22 events/h and the PM AHI was smaller than the PSG AHI. The corrected Bland-Altman plots in Figure 3 show that for both automatic scoring methods the mean error was approximately +2.0, whereas it was −0.59 for the manual scoring. The initial mean errors of almost zero for the automatic scoring were obscured by the fact that the automatic scoring was overestimating AHI in participants with AHI < 22 events/h then underestimating AHI in participants with AHI > 22 events/h. The mean errors in the corrected Bland-Altman analyses are more of a true representation of the error in participants with AHI < 22 events/h. Proportional bias was still evident in manual scoring of participants with AHI < 22 events/h although to a much lesser extent. In participants with an AHI > 22 events/h the PM AHI predictability value becomes much less. Yet, clinically this might not be of great significance because severe sleep apnea in children is diagnosed when AHI is ≥ 10 events/h. The mean PSG AHI for the 7 excluded participants was 44.3 events/h. The mean PM AHI for the same 7 participants was 22 events/h for automatic scoring, 21.8 events/h for automatic scoring with bad data excluded, and 18.8 events/h for manual scoring. These 7 participants were all correctly identified by the PM as severe sleep apnea using the three scoring methods.

Apnea-hypopnea index (AHI) corrected Bland-Altman analysis after excluding AHI > 22 events/h for portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right). PSG = polysomnography.

jcsm.15.5.733c.jpg

jcsm.15.5.733c.jpg
Figure 3

Apnea-hypopnea index (AHI) corrected Bland-Altman analysis after excluding AHI > 22 events/h for portable monitor scored automatically (PMa; left), automatically with bad data excluded (PMax; middle), and manually (PMm; right).

(more ...)

The presence of oxygen desaturation artifacts was the main reason PM autoscoring overestimated AHI compared to PSG. This can be seen with only moderate correlation between oxygen saturation (SpO2) in PSG and SpO2 in PMs. Additionally, Table 2 shows that both the ODI and the time with SpO2 < 90% were greater in automatically scored PM compared to PSG. Artifacts in general were seen less in manual scoring because of the human aspect of manual scoring. For example, a desaturation lasting for over 2 minutes might be scored in the automatic scoring but a physician could tell if this was an artifact. This was evident by the larger ODI in Table 2 in automatic compared to manual scoring. Another example was if an apneic event lasted for more than 30 seconds and there was no desaturation, a physician could identify that the cannula became dislodged; thus, marking it as an artifact, not a true apnea, and not scoring it.

Nonetheless, in the participants with AHI > 22 events/h the number of true desaturations increases significantly and oxygen desaturation artifacts become less of a factor. There are two main reasons why PMs would underestimate the AHI. The first reason is that the denominator for the AHI in PSG is the TST, whereas in the PM it is the TRT. The TRT is generally larger than the TST, which would lead to underestimation of the AHI. The second reason is that the AASM scores central apneas and hypopneas associated with arousals even if the duration and desaturation requirements are not met. These central apneas and hypopneas associated with arousals will be scored in the PSG but missed in the PM because no electroencephalogram is available. These two reasons were more of a factor in very severe OSA (AHI > 22 events/h) because there are many more events to count and miss.

The literature shows conflicting results with some studies showing PMs overestimating AHI,3,11 whereas others show PMs underestimating AHI.6,7,9,18,19 The reasons reported for overestimation are artifacts and scoring events while awake, whereas the reasons reported for underestimation are using the TRT in the PM instead of the TST, and missing central apneas and hypopneas associated with arousals. We found that PM overestimated AHI when AHI was < 22 events/h and underestimated AHI when AHI was > 22 events/h. This might explain why authors report conflicting results.

Results from ROC curve analyses showed that when simple AHI thresholds were used the PM performed very well with the area under the curve (AUC) consistently being above 0.864, which was significantly different from random chance (AUC = 0.5; P < .05) (Table 5). However, when cutoff limits to classify OSA severity were used, the PM was very good at identifying severe OSA, less so for moderate OSA, and poor for mild OSA. The AUC for mild OSA was not significantly different from 0.5 (P > .05) (Table 6). We found over 93% sensitivity and specificity for diagnosing severe sleep apnea (AHI ≥ 10 events/h) using all three scoring methods, which is very promising. For AHI ≥ 5 events/h, sensitivity and specificity were lower at 68.2% and 95.8%, respectively, using manual scoring. MediByte has not been validated in pediatric patients but authors have validated another type 3 PM, the ApneaLink.3,11 Lesser et al. used Apnealink in 25 participants with obesity and an average age of 13.6 years and applied it with simultaneous PSG. They found that for an AHI cutoff of > 5 events/h the sensitivity was 85.7% and the specificity was 83.3%.11 Massicotte et al. also used Apnealink in the same way in 35 participants with an average age of 11 years. They found a sensitivity of 94% and a specificity of 61% when an AHI cutoff value of > 5 events/h was used.3

In Table 5, the sensitivity for PM automatic scoring was higher than manual scoring when AHI thresholds of ≥ 1.5 and ≥ 5 events/h were used. This was because as previously stated, the PM automatic scoring tended to overscore, and overestimate AHI. This led to less false-negative results and consequently a higher sensitivity for automatic compared to manual scoring.

CONCLUSIONS

We compared a type 3 PM, the MediByte, simultaneously with PSG in children referred to an urban academic center. The AHI generated by the PM strongly correlated with the AHI from the PSG. Moreover, the sensitivity and specificity for detection of severe sleep apnea (AHI ≥ 10 events/h) using both manual and automatic scoring of the PM was very high (> 93%). Although PSG is still recommended for the diagnosis of sleep apnea, the PM may play an important role in diagnosing moderate and severe sleep apnea especially in older pediatric patients.

DISCLOSURE STATEMENT

All authors have read and approved the manuscript. The authors report no conflicts of interest.

ABBREVIATIONS

AASM

American Academy of Sleep Medicine

AHI

apnea-hypopnea index

AUC

area under the curve

BMI

body mass index

HR

heart rate

IQR

interquartile range

ODI

oxygen desaturation index

OSA

obstructive sleep apnea

PM

portable monitor

PSG

polysomnography

RIP

respiratory inductance plethysmography

ROC

receiver operating characteristic

RPSGT

registered polysomnographic technologist

SpO2

peripheral oxygen saturation

TRT

total recording time

TST

total sleep time

UIC

University of Illinois at Chicago

REFERENCES

1 

Marcus CL, Brooks LJ, Draper KA, et al. Diagnosis and management of childhood obstructive sleep apnea syndrome. Pediatrics. 2012;130(3):e714–e755. [PubMed]

2 

Kadmon G, Shapiro CM, Chung SA, Gozal D. Validation of a pediatric obstructive sleep apnea screening tool. Int J Pediatr Otorhinolaryngol. 2013;77(9):1461–1464. [PubMed]

3 

Massicotte C, Al-Saleh S, Witmans M, Narang I. The utility of a portable sleep monitor to diagnose sleep-disordered breathing in a pediatric population. Can Respir J. 2014;21(1):31–35. [PubMed Central][PubMed]

4 

Canadian Agency for Drugs and Technologies in Health. Portable monitoring devices for diagnosis of obstructive sleep apnea at home: review of accuracy, cost-effectiveness, guidelines, and coverage in Canada. CADTH Technol Overv. 2010;1(4):e0123[PubMed Central][PubMed]

5 

Collop NA, Anderson WM, Boehlecke B, et al. Clinical guidelines for the use of unattended portable monitors in the diagnosis of obstructive sleep apnea in adult patients. Portable Monitoring Task Force of the American Academy of Sleep Medicine. J Clin Sleep Med. 2007;3(7):737–747. [PubMed Central][PubMed]

6 

Tan HL, Kheirandish-Gozal L, Gozal D. Pediatric home sleep apnea testing: slowly getting there! Chest. 2015;148(6):1382–1395. [PubMed Central][PubMed]

7 

Parra O, Garcia-Esclasans N, Montserrat JM, et al. Should patients with sleep apnoea/hypopnoea syndrome be diagnosed and managed on the basis of home sleep studies? Eur Respir J. 1997;10(8):1720–1724. [PubMed]

8 

Mykytyn IJ, Sajkov D, Neill AM, McEvoy RD. Portable computerized polysomnography in attended and unattended settings. Chest. 1999;115(1):114–122. [PubMed]

9 

Driver HS, Pereira EJ, Bjerring K, et al. Validation of the MediByte(R) type 3 portable monitor compared with polysomnography for screening of obstructive sleep apnea. Can Respir J. 2011;18(3):137–143. [PubMed Central][PubMed]

10 

Rosen CL, Larkin EK, Kirchner HL, et al. Prevalence and risk factors for sleep-disordered breathing in 8- to 11-year-old children: association with race and prematurity. J Pediatr. 2003;142(4):383–389. [PubMed]

11 

Lesser DJ, Haddad GG, Bush RA, Pian MS. The utility of a portable recording device for screening of obstructive sleep apnea in obese adolescents. J Clin Sleep Med. 2012;8(3):271–277. [PubMed Central][PubMed]

12 

Kirsch DB. In-home testing for obstructive sleep apnea. Continuum (Minneap Minn). 2013;19(1 Sleep Disorders):223–228.

13 

Berry RB, Budhiraja R, Gottlieb DJ, et al. Rules for scoring respiratory events in sleep: update of the 2007 AASM Manual for the Scoring of Sleep and Associated Events. Deliberations of the Sleep Apnea Definitions Task Force of the American Academy of Sleep Medicine. J Clin Sleep Med. 2012;8(5):597–619. [PubMed Central][PubMed]

14 

Kapur VK, Rapoport DM, Sanders MH, et al. Rates of sensor loss in unattended home polysomnography: the influence of age, gender, obesity, and sleep-disordered breathing. Sleep. 2000;23(5):682–688. [PubMed]

15 

Simundic AM. Measures of diagnostic accuracy: basic definitions. EJIFCC. 2009;19(4):203–211. [PubMed Central][PubMed]

16 

Marcus CL, Traylor J, Biggs SN, et al. Feasibility of comprehensive, unattended ambulatory polysomnography in school-aged children. J Clin Sleep Med. 2014;10(8):913–918. [PubMed Central][PubMed]

17 

Goodwin JL, Enright PL, Kaemingk KL, et al. Feasibility of using unattended polysomnography in children for research--report of the Tucson Children's Assessment of Sleep Apnea study (TuCASA). Sleep. 2001;24(8):937–944. [PubMed]

18 

Pereira EJ, Driver HS, Stewart SC, Fitzpatrick MF. Comparing a combination of validated questionnaires and level III portable monitor with polysomnography to diagnose and exclude sleep apnea. J Clin Sleep Med. 2013;9(12):1259–1266. [PubMed Central][PubMed]

19 

Nigro CA, Dibur E, Malnis S, Grandval S, Nogueira F. Validation of ApneaLink Ox for the diagnosis of obstructive sleep apnea. Sleep Breath. 2013;17(1):259–266. [PubMed]