Quantifying the sources of discrepancy between total recording time and total sleep time in home sleep apnea testing: insights from home-based polysomnography

Abstract

Study Objectives:

To quantify the contribution of sleep onset latency (SOL), wake after sleep onset (WASO), and wake after sleep offset (WASF) to the discrepancy between total recording time (TRT) and total sleep time (TST) in home-based polysomnography (PSG) using patient-activated and deactivated monitoring devices.

Methods:

This observational study enrolled patients with a high pretest probability of obstructive sleep apnea who underwent unattended home-based PSG. We measured the duration of SOL, WASO, and WASF to quantify the discrepancy between TRT and TST. TRT was defined as the interval from device activation to deactivation by the patients. SOL represented the time from device activation to the first epoch of any sleep, WASO was the total amount of time spent awake after the sleep onset epoch until the last epoch of any sleep, and WASF was defined as the time from the last epoch of any sleep until the patient-initiated device deactivation. We also assessed differences in the apnea-hypopnea index between home-based PSG and type 3 sleep studies by reanalyzing home-based PSG recordings as simulated type 3 studies after omitting type 2 signals.

Results:

A total of 78 patients were included in the study. The mean TRT exceeded the mean TST by 19%, with TRT at 457 ± 78 minutes and TST at 383 ± 68 minutes. The mean difference between TRT and TST was 74 ± 53 minutes, attributed to SOL (30%), WASO (45%), and WASF (25%). There was considerable variability in the difference between TRT and TST among study participants, ranging from as little as 14 minutes to as much as 233 minutes. The mean apnea-hypopnea index in simulated type 3 studies (41 ± 29 events/h) was, on average, 23% lower than the mean apnea-hypopnea index recorded in home-based PSG (53 ± 30 events/h) (P < .001).

Conclusions:

There was significant variability in the gap between TRT and TST among patients at increased risk of obstructive sleep apnea undergoing unattended home-based PSG. WASO was identified as the largest contributor to this discrepancy, with notable contributions from SOL and WASF. Additionally, simulated type 3 studies underestimated the true apnea-hypopnea index compared to type 2 studies.

Citation:

Nikolopoulos A, Tatsis K, Tselepi C, et al. Quantifying the sources of discrepancy between total recording time and total sleep time in home sleep apnea testing: insights from home-based polysomnography. J Clin Sleep Med. 2025;21(6):1065–1072.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Sleep onset latency, wake after sleep onset, and wake after sleep offset are recognized contributors to the discrepancy between total recording time and total sleep time in home sleep apnea testing. However, the specific proportional contributions of these factors to this discrepancy have not been extensively studied. To address this, we investigated the difference between total recording time and total sleep time in home sleep apnea testing using data from home-based polysomnography studies, in which patients at increased risk of obstructive sleep apnea activated and deactivated the monitoring devices.

Study Impact: Our findings indicate that in unattended home-based polysomnography studies, there is significant variability in the discrepancy between total recording time and total sleep time. This gap mainly arose from wake after sleep onset, defined as the time spent awake after the sleep onset epoch until the last epoch of any sleep, although sleep onset latency and wake after sleep offset also contribute to the difference. Furthermore, the time between the last epoch of sleep and device deactivation, wake after sleep offset, exhibited the greatest variability.

INTRODUCTION

Obstructive sleep apnea (OSA) is the most prevalent sleep disorder globally.¹ Nearly half a billion people aged 30–69 years are estimated to have moderate to severe OSA worldwide, with prevalence rates exceeding 50% in some countries.² OSA is associated with a range of serious medical conditions, including cardiovascular complications, increased risk of motor vehicle and workplace accidents, metabolic syndrome, cognitive impairment, and reduced overall quality of life .^3–8

Although laboratory-based polysomnography (PSG) is considered the most accurate diagnostic tool, the utilization of home sleep apnea testing (HSAT), first introduced in the early 1990s⁹ has increased considerably during and after the COVID-19 pandemic, a trend that is expected to continue.¹⁰ The American Academy of Sleep Medicine (AASM) has recommended the use of HSAT for the diagnosis of OSA in uncomplicated adult patients presenting with signs and symptoms that indicate an increased risk of moderate to severe OSA.^11–13 An uncomplicated patient is defined by the absence of: (1) conditions that increase the risk of nonobstructive sleep-disordered breathing (eg, central sleep apnea, hypoventilation, sleep-related hypoxemia); (2) significant nonrespiratory sleep disorders that require evaluation (eg, central hypersomnolence disorders, parasomnias, sleep-related movement disorders); and (3) environmental or personal factors that hinder the proper acquisition and interpretation of data from HSAT.¹²

According to AASM guidelines, a technically adequate HSAT device should incorporate at least the following sensors: nasal pressure, chest and abdominal respiratory inductance plethysmography, and oximetry; alternatively, peripheral arterial tonometry (PAT) with oximetry and actigraphy as an alternative in certain cases.¹² HSATs that lack sensors like actigraphy or a dedicated algorithm to estimate true total sleep time (TST) (type 3 HSATs) have 2 limitations that lead to underestimating the true apnea-hypopnea index (AHI). First, while PSG determines the severity of sleep-disordered breathing using AHI based on actual sleep time, HSATs estimate severity using respiratory event index based on monitoring time.¹² Second, the standard sensors in HSAT devices cannot detect hypopneas associated only with cortical arousals.¹³

Home-based PSG (type 2 HSAT) employs a portable device that operates without the supervision of trained sleep laboratory staff and records at least 7 channels—including electroencephalogram (EEG), electrooculogram (EOG), chin electromyogram, electrocardiogram or heart rate, airflow, respiratory effort, and oxygen saturation.¹⁴ Compared to laboratory-based PSG, home-based PSG is technically reliable and has good diagnostic accuracy for both ruling in and ruling out OSA.^15–17 There are 2 primary methods for conducting type 2 studies: first, the patient visits the laboratory to be “wired up” and then returns home, and second, a technician visits the patient’s home to perform the setup.¹⁸

The discrepancy between total recording time (TRT) and TST is commonly observed in nearly all home-based sleep studies and can result in an underestimation of true AHI.¹⁹ Furthermore, it is well established that in both type 2^15,17 and type 3^20,21 sleep studies TRT consistently exceeds TST because it includes periods of wakefulness, such as sleep onset latency (SOL), wake after sleep onset (WASO), and wake after sleep offset (WASF).²² Although the difference between TRT and TST varies among patients, the specific contributions of SOL, WASO, and WASF to this discrepancy have not been thoroughly investigated previously. Understanding how each of these factors impacts the difference between TRT and TST can improve the interpretation of HSATs, where monitoring conditions differ from a laboratory setting. To address this, we conducted an observational study involving unattended home-based PSG in a patient population at increased risk of OSA. The primary aim of this study was to quantify the contributions of SOL, WASO, and WASF to the discrepancy between TRT and TST. The secondary objective was to evaluate the AHI and compare the diagnostic accuracy of home-based PSG with simulated type 3 studies, generated by excluding type 2 signals.

METHODS

Study design and participants

This single-center observational study was conducted at the Sleep Unit of the Department of Respiratory Medicine at Ioannina University General Hospital, Ioannina, Greece. The study included consecutive patients with a high pretest probability of OSA who were referred to the hospital for diagnosis of sleep-disordered breathing between January 2021 and October 2022. During this period, the sleep laboratory was closed due to the COVID-19 pandemic, and all patients underwent either type 2 or type 3 HSATs. All procedures were performed in accordance with the ethical standards of the institutional review board at Ioannina General University Hospital, which approved the study protocol (approval number: 53/27-03-2019). All participants provided written informed consent and completed sleep-related questionnaires such as the Epworth Sleepiness Scale and bed-partner comments, if applicable.

A comprehensive medical history and examination, following routine clinical practice, were conducted by a sleep specialist. For every participant, the following data were recorded: age, sex, body mass index, smoking habits, alcohol consumption, neck and waist circumference, presence of comorbidities such as hypertension, diabetes mellitus, atrial fibrillation, chronic obstructive pulmonary disease, hypothyroidism, as well as snoring and observed apneas. Subsequently, a sleep study was scheduled and conducted in the participants’ homes using a portable type 2 monitoring device. PSG hook-up was conducted in the hospital and patients returned home fitted with the device. The study’s exclusion criteria comprised heart failure, respiratory failure, daytime hypercapnia, known active malignancy, chronic kidney disease, neurologic disorders, and bedridden status.

Type 2 portable monitoring devices

For the present study, we used the portable type 2 device: Philips Respironics Alice PDx Portable Diagnostic Sleep System (Murrysville, PA). A certified sleep technician connected the participants to the recording system in the early evening at the sleep unit and then patients returned to their own homes. Patients were instructed not to drive back home by themselves to avoid the detachment of electrodes. The signals recorded included 2 EEG derivations (F4–M1 and C4–M1), EOG, chin electromyogram, electrocardiogram, oronasal airflow thermistor, nasal pressure transducer, abdominal and thoracic plethysmography bands, oximetry, body position monitor, and snoring. Recordings included in this study had less than 10% artifact or signal loss, including issues like sensor displacement and signal quality, during the night. The patients were instructed to turn the device on at bedtime and off in the morning upon waking. The total analysis time of the recording was considered equal to the TRT and was defined as the time from device activation until the patient-initiated deactivation with electrodes still connected, or until all the electrodes were detached. SOL was defined as the time from patient-initiated device activation to the first recorded epoch of any sleep stage. TST was defined as the total duration of sleep, calculated from the sleep onset epoch to the last epoch of any sleep, excluding any wake time in between. WASO was defined as the total amount of time awake after the sleep onset epoch until the last epoch of any sleep. WASF was defined as the total amount of time from the last recorded epoch of any sleep until the patient-initiated device deactivation, or until all the electrodes were detached. All recordings were collected and interpreted manually by a certified sleep technician following the guidelines established by the AASM.¹⁴ Hypopnea was defined as a reduction of nasal pressure excursion of at least 30% from pre-event baseline for at least 10 seconds, associated with either a desaturation of at least 3% and/or arousal from sleep.

Type 3 studies

Type 3 studies were simulated by removing the channels available only in type 2 studies (EEG, electrocardiogram, EOG, electromyogram, and the thermistor sensor). The remaining channels used in simulated type 3 studies included nasal pressure, pulse oximetry, thoracic and abdominal respiratory effort belts, and body position. The AHI of simulated type 3 studies was calculated as the total number of manually scored respiratory events, multiplied by 60, and divided by the TRT in minutes, to yield events per hour. The simulated type 3 studies were interpreted manually in random order by a second, independent certified sleep technician following the guidelines established by the AASM. Hypopnea for the simulated type 3 studies was defined as a drop in nasal pressure excursion on the nasal cannula signal of at least 30% from the pre-event baseline for at least 10 seconds, associated with a desaturation of at least 3%.

Statistical analysis

Data normality was evaluated using the Shapiro-Wilk test. Paired continuous variables (eg, TRT vs. TST and type 2 AHI vs. type 3 AHI) were compared using the paired samples t-test for normally distributed data or the Wilcoxon signed-rank test when normality assumptions were not met. The accuracy of a type 3 home study was determined for different AHI thresholds, classified according to the OSA severity levels: mild OSA (AHI ≥ 5–14.9 events/h), moderate OSA (AHI ≥ 15–29.9 events/h), and severe OSA (AHI ≥ 30 events/h). Type 3 studies were rated as either true positive (TP), true negative (TN), false positive (FP), or false negative (FN) in comparison with the respective type 2 studies according to the applicable thresholds. We calculated the following metrics for the type 3 studies: accuracy: (TP+TN)/(TP+FP+TN+FN), sensitivity: TP/(TP+FN), specificity: TN/(TN+FP), positive predictive value (PPV): TP/(TP+FP), and negative predictive value (NPV): TN/(TN+FN). The analysis was carried out using SPSS version 26.0 (IBM Corp., Armonk, NY), and statistical significance was set at .05.

RESULTS

Participants and sleep studies

Out of 87 participants, recording data from 9 were excluded due to prolonged oximetry loss, complete recording failure, or premature termination of recording. The final study group comprised 78 participants, including 58 males (75%) and 20 females (25%). The participants had a mean age of 55 ± 11 years, a mean body mass index of 33 ± 6 kg/m², and an average Epworth Sleepiness Scale score of 10 ± 5. Their demographic characteristics and comorbidities are summarized in Table 1. All 78 recordings included at least 4 hours of technically adequate oximetry and flow data. Nine (11%) recordings had poor abdominal or thoracic belt signals during a portion of the recording, and 8 (10%) had intermittent oximetry signal loss. Data from these participants were included in the analysis as more than 90% of the total recording was available.

Table 1.

Demographic and clinical characteristics of the study population.

	All Patients (n = 78)
Age (years)	55 ± 11
Male sex (%)	58 (75%)
BMI (kg/m2)	33 ± 6
ESS (score)	10 ± 5
Lifestyle
Active smokers n (%)	27 (35)
Previous smokers n (%)	24 (30)
Nonsmokers n (%)	27 (35)
Daily alcohol consumption n (%)	18 (23)
Comorbidities
Arterial hypertension n (%)	52 (67)
Atrial fibrillation n (%)	9 (11)
COPD n (%)	10 (13)
Dyslipidemia n (%)	25 (32)
Diabetes mellitus n (%)	9 (11)
Depression n (%)	10 (13)
Hypothyroidism n (%)	5 (6)

Values are presented as mean ± SD. BMI = body mass index, COPD = chronic obstructive pulmonary disease, ESS = Epworth Sleepiness Scale, SD = standard deviation.

Discrepancy between TRT and TST

In home-based PSG studies, the mean TRT was 457 ± 78 minutes, 19% longer than the mean TST of 383 ± 68 minutes (P < .001), with a mean difference between TRT and TST of 74 ± 53 minutes (Table 2). This difference was due to contributions from SOL (22 ± 20 minutes), WASO (33 ± 34 minutes), and WASF (18 ± 33 minutes), which accounted for 30%, 45%, and 25% of the mean TRT-TST discrepancy, respectively. A box-and-whisker plot of the TRT-TST difference, SOL, WASO, and WASF is shown in Figure 1. The percentages of mean TST, WASO, WASF, and SOL relative to mean TRT are presented in Figure 2. The difference between TRT and TST ranged from 14 to 233 minutes within the study group. WASO ranged from 1 to 177 minutes, with 35 patients (45%) having a WASO exceeding 20 minutes. SOL ranged from 4 to 82 minutes, with 15 patients (19%) exhibiting a SOL greater than 30 minutes. WASF ranged from 0.4 to 176 minutes, with 19 patients (24%) having a WASF of more than 20 minutes. WASF showed the greatest variability, with a large standard deviation and a significant difference between its mean value of 18 minutes and median value of 6 minutes.

Table 2.

Polysomnographic findings in home-based diagnostic PSG studies.

	Μean	SD	Median	Q1–Q3
TRT (minutes)	457	78	453	414–506
TST (minutes)	383	68	387	333–424
TRT minus TST	74	53	58	32–98
SOL (minutes)	22	20	16	7–31
WASO (minutes)	33	34	22	8–45
WASF (minutes)	18	33	6	3–20
AHI (events/h)	51	30	45	29–73
ODI 3% (events/h)	49	32	42	26–77
Sleep efficiency (%)	87	9	87	82–93
T90%	34	31	26	7–62
Mean SpO2 (%)	90	4	91	87–93
Minimum SpO2 (%)	72	14	75	64–82
Arousal index (events/h)	40	23	35	25–52
N3 length (%)	9	7	8	4–11
REM latency (minutes)	159	80	135	102–209

Values are presented as mean ± SD and median (IQR). AHI = apnea-hypopnea index, IQR = interquartile range, N3 = non-REM sleep stage 3, ODI = oxygen desaturation index, PSG = polysomnography, Q1 = first quartile (25th percentile of the data), Q3 = third quartile (75th percentile of the data), REM = rapid eye movement, SD = standard deviation, SOL = sleep onset latency, SpO₂ = peripheral oxygen saturation, T90% = percentage of sleep time with SpO₂ below 90%, TRT = total recording time, TST = total sleep time, WASF = wake after sleep offset, WASO = wake after sleep onset.

Figure 1. The contributions of SOL, WASO, and WASF to the discrepancy between TRT and TST in home-based PSG.

Box-and-whisker plot demonstrating the median (line), lower and upper IQR (box), whiskers extending to the highest and lowest values, and outliers presented as dots. IQR = interquartile range, PSG = polysomnography, SOL = sleep onset latency, TST = total sleep time, TRT = total recording time, TRT – TST = difference between total recording time and total sleep time, WASF = wake after sleep offset, WASO = wake after sleep onset.

Figure 2. Percentages of mean TST, WASO, WASF, and SOL in relation to the mean TRT.

SOL = sleep onset latency, TRT = total recording time, TST = total sleep time, WASF = wake after sleep offset, WASO = wake after sleep onset.

Patient classification by OSA severity in type 2 and type 3 studies

The mean AHI was significantly higher in type 2 studies (53 ± 30 events/h) compared to type 3 studies (41 ± 29 events/h) (P < .001) (Table 3). On average, the AHI in type 3 sleep studies was 23% lower than that of type 2 studies. As a result, the distribution of patients by severity group differed between the 2 types of sleep studies, with type 2 yielding a higher AHI and indicating more severe disease than type 3 studies (Figure 3). For mild OSA (AHI = 0–14.9 events/h), type 2 identified 7 patients (9%) compared to 15 (19%) with type 3. For moderate OSA (AHI = 15–29.9 events/h), type 2 identified 15 patients (19%) compared to 17 (22%) with type 3. For severe OSA (AHI ≥ 30 events/h), type 2 identified 56 patients (72%) vs 46 (59%) with type 3. The accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of simulated type 3 studies are presented in Table 4.

Table 3.

AHI and TST in type 2 and simulated type 3 studies.

	Type 2 Studies	Type 3 Studies	P
AHI (events/h)	53 ± 30	41 ± 29	< .001
TST* (minutes)	383 ± 68	457 ± 78	< .001

^*TST is represented by TRT in type 3 studies. AHI = apnea-hypopnea index, TRT = total recording time, TST = total sleep time.

Figure 3. Proportion of patients by severity group in home-based PSG and simulated type 3 studies.

AHI = apnea-hypopnea index, PSG = polysomnography.

Table 4.

The diagnostic accuracy, sensitivity, specificity, PPV, and NPV of simulated type 3 studies across different severity levels of OSA.

	AHI > 5 events/h	AHI > 15 events/h	AHI > 30 events/h
Accuracy	0.96	0.88	0.88
Sensitivity	0.96	0.88	0.84
Specificity	1	0.85	1
PPV	1	0.98	1
NPV	0.25	0.42	0.70

AHI = apnea-hypopnea index, NPV = negative predictive value, OSA = obstructive sleep apnea, PPV = positive predictive value.

DISCUSSION

In this study, patients with a high pretest probability of OSA who underwent an unattended home-based PSG had a mean TRT that was 19% longer than the mean TST. The factors contributing to this difference were WASO, which accounted for 45% of the difference, followed by SOL at 30%. WASF contributed 25% of the gap and showed the greatest variability among study participants. Previous home-based PSG studies found that TRT overestimated TST by 14–27.5%.^16,23,24

Traditionally, WASO in laboratory-based PSG studies has been defined as the time spent awake after sleep onset until “lights on.”²⁵ However, in our study, this definition was unsuitable due to the absence of a sleep technician to determine the exact “lights on” moment. In HSAT, the end of recording time is dictated by device deactivation. This can occur significantly later than the “lights on” time used in laboratory-based PSG studies. To address this limitation, we defined WASO as the time spent awake from the sleep onset epoch to the last recorded sleep epoch (sleep offset) and introduced WASF as the time from the last recorded sleep epoch until device deactivation. We instructed our participants to activate the device at bedtime and deactivate it upon waking. However, a subset of 19 participants (24%) left the devices activated long after waking up (more than 20 minutes). In our follow-up with these patients, some reported attempting to fall back asleep unsuccessfully, while others admitted forgetting to turn the device off despite being awake in bed. A similar observation was made in a previous study comparing home-based PSG with laboratory-based PSG in 40 patients.²⁶ In that study, WASO was defined as time awake from sleep onset to device deactivation. Results showed that WASO was significantly longer in home-based PSG compared to laboratory-based PSG: 97 ± 74 minutes vs 70.8 ± 42.4 minutes; P = .05. The researchers attributed this difference to the delay in pressing the “events” button to stop recording.²⁶

Prolonged WASF in some of our study participants may also reflect increased sleep inertia, which is linked to conditions such as OSA, mood disorders, central hypersomnolence disorders, and delayed sleep phase syndrome.²⁷ The longer durations of SOL and WASO, in turn, are associated with conditions such as insomnia, chronic pain syndrome, restless legs, and nocturia.^28,29 The overlap between insomnia and OSA is common, with studies indicating that 30–50% of individuals with OSA may report considerable insomnia symptoms.³⁰ This high co-occurrence suggests a plausible relationship where the 2 conditions exacerbate each other.³¹ We believe that including WASF as an additional parameter in our analysis offers valuable insights. By measuring the duration of wakefulness after sleep offset, WASF can shed light on prolonged sleep inertia and difficulties returning to sleep after an undesired awakening during HSAT.²² This metric complements other sleep measurements, such as TST, SOL, and WASO, and enables a comprehensive assessment of wakefulness patterns in HSAT. Unlike laboratory-based studies that conclude abruptly when the lights turn on—regardless of whether the patient is still asleep—HSAT can capture “real-world” sleep behaviors more closely, making WASF a particularly relevant metric.

In our study, we diagnosed OSA using unattended home-based PSG. Previous studies have shown that home-based PSG represents a technically reliable and valid diagnostic modality in patients at increased risk of OSA without significant comorbidities.²³ A systematic review of 6 studies comparing home-based PSG against laboratory-based PSG concluded that unattended home-based PSG represents a reliable diagnostic tool for OSA, even in patients with comorbidities.³² The familiar setting of the home could minimize anxiety and allow the individual to adhere more closely to their usual bedtime routine.²⁴ However, limitations include the lack of video recording and exhaled carbon dioxide measurement, risk of data loss, and a general tendency toward overestimation of severity in mild disease and underestimation in more severe disease.³² In our study, 10% (9 of 87) of PSGs were excluded due to poor signal quality. A cross-sectional data analysis from the Sleep Heart Health Study (n = 6,802) estimated that during unattended home-based PSG, the abdomen, chest, thermocouple, and chin signals were usable for about 78–81% of the average time.³³ For the bilateral EEG, EOG, and oximetry signals, the mean usability was higher, between 91% and 93%.³³

The AASM has established scoring rules for HSATs.²⁵ In these guidelines “monitoring time” is defined as the total time recorded minus any periods of interference or time the patient was awake, as determined by actigraphy, body position, respiratory patterns, or the patient’s sleep diary. The respiratory event index is then calculated as the total number of scored respiratory events multiplied by 60 and then divided by the monitoring time, which serves as a surrogate for the AHI.²⁵ However, many sleep laboratories and commercial HSAT service providers continue to use the TRT instead of the more refined “monitoring time” suggested by the AASM guidelines.³⁴ Proponents of using TRT argue that the manual editing of the recorded data is time-consuming, adds extra costs, and may introduce bias without improving the accuracy of the AHI estimates significantly. In addition, some providers may lack experience with manual editing and certain HSAT devices do not support manual review of automated scoring.³⁴ In the current study, we intentionally used TRT, defined as the time from device activation to deactivation. We did not use “monitoring time” to analyze the simulated type 3 studies, as our primary objective was to highlight the differences between TRT and TST, and examine how these differences impact AHI calculation when comparing type 2 with simulated type 3 studies.

In our study, the discrepancy between TRT and TST showed substantial variability among participants, ranging from as little as 14 minutes to as much as 233 minutes. This wide range highlights individual differences in sleep-wake patterns and reflects variability in SOL, WASO, and the time taken to deactivate the device after final awakening (WASF). These findings highlight the need to consider each person’s specific sleep patterns when interpreting HSAT results. They also point to the value of integrating alternative methods to estimate true sleep time during HSAT, aiming for more precise assessments that account for individual variability. Potential solutions include actigraphy, self-reported sleep duration, autonomic signals, and automatic sleep staging.^35–45

Actigraphy tends to agree with actual sleep time in normal individuals but tends to overestimate sleep time and efficiency in patients with sleep-disordered breathing when these individuals lie awake in bed quietly.³⁵ Actigraphy shows high sensitivity (over 90%) for the detection of sleep but has limited specificity (0.35) for the detection of wakefulness.³⁶ An AASM task force review of 6 studies concluded that the use of integrated actigraphy to estimate TST during HSAT offers only slight improvement in diagnostic accuracy compared to using TRT alone, particularly in cases of severe OSA.³⁷ On the other hand, self-reported sleep duration has risks of both overestimation and underestimation of OSA severity, with discrepancies that often exceed 1 hour.^38–40 Non-EEG signals, such as autonomic signals derived from PAT (based on features extracted from time series of peripheral arterial tone amplitudes and interpulse periods).⁴¹ The combined use of PAT and actigraphy signals from Watch-PAT100 (a PAT recorder) can detect sleep stages with moderate agreement compared to PSG in both normal individuals and patients with OSA.⁴¹

The development of deep neural network models capable of providing a classification of sleep stages using various signals, either alone or in combination with each other, shows promise.^42,43 These signals include EEG, electromyogram, EOG, electrocardiogram, respiratory rate, photoplethysmography, pulse transit time, heart rate variability, ballistocardiogram, actigraphy, cardiorespiratory coupling, and body movement frequency.⁴⁴ Many of these signals can be monitored and obtained by sleep wearable devices which are becoming smaller in size and easier to use by patients.⁴⁵ These advancements suggest that an effective automated sleep staging model based on non-EEG data for wide clinical use may become available in the near future.⁴⁴

Time dilution plays a significant role in the underestimation of true TST in type 3 studies.¹⁹ In the current study, the AHI from simulated type 3 studies was, on average, 23% lower than that from type 2 studies. As a result, simulated type 3 studies identified more than twice as many patients with mild OSA compared to type 2 studies. According to AASM guidelines, a “negative” HSAT in a patient with a high pretest probability of OSA warrants a repeat PSG, even when the HSAT is of adequate quality and duration.¹² Similarly, results from a large-scale study involving 11,049 participants, conducted by the European Sleep Apnea Database (ESADA) study group, showed that those assessed using polygraphy (with at least 4 channels) had AHIs approximately 30% lower than individuals evaluated by PSG.²¹

The main strength of our study was the use of home-based PSG. As the gold standard for sleep assessment, PSG enabled reliable measurement of TST, SOL, WASO, and WASF in this study. Additionally, conducting PSG at home might have reduced the impact of 2 common challenges frequently encountered in laboratory-based PSG: body positioning and the “first night effect.” A previous study has shown that people spent more time sleeping in the supine position during laboratory-based PSG (14.1%) compared to home-based PSG (7.1%).⁴⁶ The “first night effect” refers to disruptions such as poorer sleep quality, shorter sleep duration, longer time to fall asleep, delayed onset of rapid eye movement sleep, and less rapid eye movement sleep overall.⁴⁷ While this effect can also be observed in type 2 studies since patients need to adjust to the monitoring device, it is typically more pronounced in laboratory settings than in home environments.^48,49

This study has several limitations. First, because the home-based PSG was unattended, the quality of signals was suboptimal for some participants. As a result, 9 individuals were excluded from the analysis. Second, although participants were advised to sleep in their usual positions, the monitoring equipment may have affected their natural sleep patterns and body positions. Therefore, our findings might not be widely applicable to other patient groups, especially those with different comorbidities and those who use alternative portable devices for home sleep studies. Third, the PSG setup was performed at the hospital and then patients returned home with the device fitted. This might have led to a higher rate of PSG failures compared to home-based hook-up. Finally, this study was initially planned before the outbreak of the COVID-19 pandemic but was conducted during the pandemic.

CONCLUSIONS

The current study found that TRT in home-based PSG was, on average, 19% longer than the actual TST, mainly due to WASO, with additional contributions from SOL and WASF. The discrepancy between TRT and TST varied considerably among participants, reflecting individual differences in the time spent awake during the monitoring period. These findings emphasize the need for more precise tools to accurately estimate true TST in HSAT.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Institution where work was performed: Sleep Disorders Unit, Department of Respiratory Medicine, University Hospital of Ioannina, Ioannina, Greece. This work was financially supported by the Project: ΕP1ΑΒ-0028213, funded by the Operational Program: “Epirus 2014-2020 E.Y.D.E.P. Region of Epirus, Greece.” The authors report no conflicts of interest.

ABBREVIATIONS

AASM

American Academy of Sleep Medicine

AHI

apnea-hypopnea index

EEG

electroencephalogram

EOG

electrooculogram

HSAT

home sleep apnea testing

OSA

obstructive sleep apnea

PAT

peripheral arterial tonometry

PSG

polysomnography

SOL

sleep onset latency

TN

true negative

TP

true positive

TRT

total recording time

TST

total sleep time

WASF

wake after sleep offset

WASO

wake after sleep onset