Validation and Performance of the Sleep Inertia Questionnaire in Central Disorders of Hypersomnolence

Abstract

Background

Optimal measurement tools for problematic sleep inertia, common in some central disorders of hypersomnolence (CDH), have not yet been determined. We evaluated the performance of the Sleep Inertia Questionnaire (SIQ) in CDH, and how well it distinguished hypersomnolent groups from controls, and IH (idiopathic hypersomnia) from narcolepsy type 1 (NT1).

Methods

This prospective, bi-centric study included 63 control, 84 IH, 16 NT1, 18 narcolepsy type 2 (NT2), and 88 subjective excessive daytime sleepiness (sEDS) participants, using ICSD-3 criteria. 126 (47.2%) participants were on any medication at the time of SIQ completion. We assessed construct validity of SIQ scores, and sleep inertia duration (SID), and compared them across diagnoses, controlling for age and center. We derived cutpoints to distinguish hypersomnolent patients from controls and IH from NT1. Sensitivity analyses for depression, chronotype, and medication were performed.

Results

The SIQ sum and composite score were significantly lower in controls than in other groups (p<0.0001), demonstrating outstanding ability to distinguish patients from controls (AUCs 0.92), without differences among hypersomnolent groups. SID (AUC 0.76) was significantly shorter in controls than in all hypersomnolent groups except NT1, and was shorter in NT1 than in IH or sEDS. Optimal SIQ sum cutpoint was 42 (J = 0.71) for patients versus controls. Optimal SID cutpoint in distinguishing IH from NT1 was 25 minutes (J = 0.39).

Conclusion

The SIQ has excellent ability to distinguish hypersomnolent patients from healthy controls, after controlling for depression, eveningness, and medication. SID is best at distinguishing IH from NT1.

Introduction:

Sleep inertia describes the transition period between sleep and the fully awake state. This period is characterized by impairments in vigilance and physical abilities, with desire to return to sleep. While sleep inertia seems to be present to some extent and duration in the general population, some individuals experience a severe and/or prolonged form historically referred to as “sleep drunkenness.” Roth et al. noted this severe form of sleep inertia to occasionally occur in healthy individuals often in the context of alcohol consumption, medication effect, or sleep deprivation.¹ Indeed, a population study found that 2.9% of the population had experienced an episode of severe sleep inertia past month – more commonly occurring in those who were shift workers or had consumed alcohol near bedtime.² Additionally, severe sleep inertia can be experienced by people with depression, delayed sleep phase syndrome,^3,4 and obstructive sleep apnea.⁵ Individuals with syndromic depression are more likely to report severe sleep inertia compared to individuals with mild-to-no depression.⁶ Sleep inertia has potentially dangerous consequences, especially in health care or military personnel who are woken abruptly in the night and required to make cognitively-taxing decisions.^7,8

Patients experiencing idiopathic hypersomnia (IH) are one group that may be particularly vulnerable to severe sleep inertia. The International Classification of Sleep Disorders – Third Edition-Text Revision notes sleep inertia to be a core, though not necessary, feature of IH.⁹ Reported rates of severe sleep inertia among patients with IH ranges from 36–58%,^10,11 though presence of sleep inertia is not specific for IH. People with other CDH diagnoses such as narcolepsy type 1 (NT1) and narcolepsy type 2 (NT2) may exhibit sleep drunkenness at times, although this appears to be much more common in people with NT2 than in NT1.^11–13 This shared symptomatology between IH and narcolepsy is not surprising, as differentiating clinically between IH and, specifically, NT2 can be extremely difficult given current limitations in relied upon measurement tools. For example the MSLT, which serves as the diagnostic gold standard, has poor test-retest reliability in distinguishing IH from NT2.^14,15 As such, the clinical care of patients experiencing CDH would greatly benefit from additional measures that have both strong sensitivity for detecting hypersomnolence and specificity for distinguishing across CDH.

The Sleep Inertia Questionnaire (SIQ) has been previously validated in persons with depression.⁶ Although it has been descriptively applied in a few recent studies of IH^16,17, detailed analyses of its performance in people with CDH has yet to be completed. The current study sought to validate the SIQ in sleep disorders associated with hypersomnolence, while also assessing the capabilities of the SIQ to distinguish (1) pathological sleep inertia in CDH patients from healthy controls and (2) as a tool to differentiate diagnoses within CDH based on sleep inertia symptoms.

Methods:

Participants

Data were obtained prospectively from participants at two sites, the Emory Sleep Center and the University of Wisconsin-Madison Sleep Clinic. Patients were consecutively offered participation in the study and included if they had a diagnosis of a primary CDH based on ICSD-3 criteria (NT1, NT2, or IH) or if they were seeking evaluation for problematic sleepiness but diagnostic testing did not confirm any of these CDH disorders (the subjective excessive daytime sleepiness group, sEDS). Healthy controls had to be free of complaints of daytime sleepiness, with a normal Epworth Sleepiness Scale, and have no history of sleep disorders. A subset of controls also underwent PSG/MSLT. The controls were friends and family of CDH patients at Emory or were recruited across the campuses of University of Wisconsin-Madison and Emory University/Emory Healthcare. All participants provided informed consent through research protocols approved by the respective Institutional Review Boards at Emory University and the University of Wisconsin-Madison.

ICSD-3 Criteria for IH, NT1, and NT2

Based on the time period in which this study was conducted, ICSD-3 criteria were used for diagnosis. IH is a CDH characterized by daily sleepiness in the absence of cataplexy plus at least one objective measure of hypersomnolence (mean sleep latency ≤ 8 minutes on multiple sleep latency test (MSLT), or long sleep time [total 24-hour sleep time ≥ 660 minutes on polysomnography (PSG) or averaged over at least 7 days with wrist actigraphy with sleep log], after ruling out other causes of daytime sleepiness.⁴

NT1 is a CDH characterized by daily sleepiness, with either: 1) cataplexy and an mean sleep latency ≤ 8 minutes and ≥ 2 sleep onset REM periods (SOREMPs) on PSG/MSLT; or 2)] CSF hypocretin-1 concentration ≤ 110 pg/mL or <1/3 of control mean values.⁴

NT2 meets the same daily sleepiness and PSG/MSLT criteria as NT1, but with the following differences: 1) cataplexy is absent, 2) either CSF hypocretin-1 concentration has not been measured or it is ≥ 110 pg/mL or >1/3 of control mean values.⁴

Sleep Inertia Questionnaire

The Sleep Inertia Questionnaire (SIQ), developed and validated by Kanady et al. in patients with depression⁶, was given to all participants to self-administer. This questionnaire asks participants to rate 22 different items each on a scale from 1 to 5 (1=not at all; 5=all the time), considering how often they experience different challenges upon awakening; higher numbers indicate more frequent difficulty. The time frame “on a typical morning in the past week, after you wake up” is applied to all 22 items of the SIQ. The SIQ encompasses four separate constructs of sleep inertia: cognitive, behavioral, physiological, and emotional correlates of sleep inertia. For this work, we derived a sub-score for each of these four constructs by averaging responses of all items within a construct; in this way, each sub-score could have the same range of values, from 1 to 5. We then generated an SIQ composite (weighted) score by summing the four sub-score values. We also report the raw score, the SIQ sum, as a sum of all individual questions. The single SIQ question “How long does it take you to ‘come to’ in the morning?” was used as a separate categorical variable, in minutes, not incorporated into the SIQ total score.

At the Emory site, the SIQ was completed by patients during their clinical evaluation, typically their first clinic visit or the day of their MSLT. At the Wisconsin site, the SIQ was completed by patients on the day of their MSLT. Control participants completed the SIQ during their research enrollment or MSLT visits. Patients undergoing MSLT were weaned off medications known to impact MSLT results whenever clinically feasible/appropriate, such that many but not all patients completing SIQ during MSLT were unmedicated. Patients completing the SIQ at their first clinical visit were sometimes unmedicated but were often already treated with medication for their CDH and seeking a second opinion. Participants were considered to be unmedicated if they were taking no medications at the time of SIQ completion.

Sleep Evaluation and Sleep Questionnaires

All patients underwent similar clinical evaluation at both centers. All patients (not all controls) had PSG/MSLT, some with long sleep recording when clinically indicated. Recordings were scored by certified sleep technologists following AASM scoring rules active at the time of testing, generating standard metrics such as sleep staging, MSL mean sleep latency, and number of SOREMs.

In addition to completing the SIQ, participants completed several other questionnaires. The Epworth Sleepiness Scale (ESS) is a short 8-item self-report questionnaire measuring excessive daytime sleepiness, each item rated on a scale (0–3) of how likely it is to fall asleep in everyday/common daytime situations “in recent times”, yielding a composite score of sleepiness severity.¹⁸

The Hypersomnia Severity Index (HSI) is a 9-item self-report questionnaire rating “sleep in the past month” on a scale (0–4), yielding a composite score to assess severity and impairment of hypersomnolence.¹⁹

The Morningness-Eveningness Questionnaire (MEQ) is a 19-item self-report questionnaire composed of both Likert-type and time-scale questions to assess individual differences in morningness and eveningness, quering preferences in sleep and waking times and subjective “peak times” at which respondents feel their best.²⁰

The Short-Form Beck Depression Inventory was used at Emory, and the Inventory of Depressive Symptomatology Self-Report (IDS-SR) was used at Wisconsin to evaluate depression. The Short-Form Beck Inventory is a self-rating scale to assess depression and suggested that scores higher than 10 are associated with moderate and severe depressive syndromes.²¹ The IDS-SR is a 30-item self-rating scale on a scale (0–3) for each item to evaluate depressive symptom severity.^22,23

Statistical analyses

Demographics, clinical features, and SIQ scores were compared across diagnostic groups via ANOVA, controlling for unequal variances and with Tukey test of pairwise comparisons in the case of overall significant ANOVA, for continuous outcome variables. Chi-square or Fisher test were used for categorical outcome variables, with Bonferroni correction of pairwise comparisons. For SIQ scores, additional adjustment was made for center (Emory vs Wisconsin) and age, given significant differences in age across diagnoses. For the comparison of IH with and without long sleep duration, t-test was used for continuous outcomes.

To assess how well the SIQ related to other measures of sleep inertia (i.e., construct validity), we compared SIQ composite score, SIQ sub-scores, and time to wake up to the single question from the Hypersomnia Severity Index (HSI, “Please consider your sleep in the past month. To what extent do you think that you have difficulty waking up in the morning or from naps?”), dichotomized to those who endorsed difficulty waking up (scores of 3 (“a lot”) or 4 (“very much”) vs those who did not. On the Emory sample, we assessed for correlations between number of alarm clock rings needed to awaken and SIQ measures, and compared SIQ measures in those who did versus did not need another individual to help them wake up.

To assess whether the SIQ was capturing information not captured in other hypersomnolence assessments, we calculated Pearson correlation coefficients comparing SIQ measures with ESS, MSLT mean sleep latency, PSG total sleep time, PSG N3 percent, and overall HSI scores.

We then performed three sensitivity analyses. Because sleep inertia can be experienced by people with depression and with delayed sleep phase syndrome, we performed ANCOVA analyses, comparing SIQ measures across diagnoses while controlling for depression, age, and center, and, separately, scores on the Morningness-Eveningness Questionnaire (MEQ)²⁰ and age. Depression was modelled as a dichotomous variable (yes/no), using a Short-Form Beck Depression Inventory cutoff of 10 or higher (Emory sample) or a IDS-SR scale cutoff of 26 or higher (Wisconsin sample).²⁴ MEQ scores were dichotomized as evening type or not evening type within the Emory sample. Third, we compared SIQ measures across diagnoses while controlling for age and center, limited to those who were not on medications for hypersomnolence at the time of the SIQ completion.

Finally, we used logistic regression for receiver-operator characteristic analyses to determine how well the SIQ distinguished: 1) all hypersomnolent patients from controls; and 2) between IH and other CDH disorders. Cutpoints were selected to optimally balance sensitivity and specificity, informed by Youden scores. All analyses were performed in SAS version 9.4.

Results:

Participants

We studied 267 participants [control, n= 63; idiopathic hypersomnia (IH), n= 84; narcolepsy type 1 (NT1), n= 16; narcolepsy type 2 (NT2), n= 18; subjective excessive daytime sleepiness (sEDS), n= 86]; 120 participants were from Emory and 147 from University of Wisconsin. The participants were included prospectively, from 2/8/16 through 11/22/19. There were 78.3% female participants with a mean age of 32.5 ± 11.1 years. IH patients were older than controls and the sEDS group (see Table 1). Among the IH group, 22.6% (n = 19) were diagnosed based on long sleep time ≥ 11 hours (LST) and 77.4% were diagnosed by MSLT. At Emory, seven NT1 patients had CSF confirmation of hypocretin deficiency, and two NT1 patients without CSF testing were HLA DQB10602 positive; the other two patients had neither CSF nor HLA tested. For the NT1 patients from the University of Wisconsin-Madison Sleep Clinic, HLA testing was positive in one participant; neither CSF nor HLA were tested in the remainder. 126 participants were on any medication at the time of the SIQ (47.2%).

Table 1.

Demographics, ESS, MSL, and SIQ scores by diagnosis.

	Control	sEDS	IH	NT1	NT2	p-value	Pairwise differences
n	63	86	84	16	18	--	--
Age	30.7 ± 8.7	29.3 ± 9.8	36.0 ± 12.4	35.6 ± 12.6	34.1 ± 10.1	0.001	IH > control, IH > sEDS
Female gender	44 (69.8%)	71 (82.6%)	71 (84.5%)	10 (62.5%)	13 (72.2%)	0.08	--
Emory participant	34 (54.0%)	22 (25.6%)	42 (50.0%)	11 (68.8%)	11 (61.1%)	0.0002	sEDS < all others
ESS	5.4 ± 3.1	13.6 ± 4.6	14.7 ± 5.1	17.7 ± 4.6	14.9 ± 5.7	<0.0001	Control < all others; sEDS < NT1
MSL (min)	14.3 ± 4.5	13.1 ± 4.6	8.3 ± 4.8	5.8 ± 6.4	4.7 ± 2.6	<0.0001	Control = sEDS > all others; IH > NT2
HSI	3.8 + 4.4	16.6 + 10.9	24.6 + 6.6	25.4 + 6.7	20.4 + 8.3	<0.0001	Control < all others; sEDS < IH and NT1; NT2 < IH
Evening type by Owl-Lark (Emory sample only)	2 (6.3%)	8 (36.4%)	14 (33.3%)	0	3 (27.3%)	0.01	--
Depression	2 (3.2%)	37 (43.0%)	32 (38.1%)	5 (31.3%)	7 (38.9%)	<0.0001	Control < all others
On medication at time of SIQ	14 (22.2%)	32 (37.2%)	61 (72.7%)	9 (56.3%)	10 (55.6%)	<0.0001	IH > Control and sEDS
SIQ sum	35.1 ± 10.9	67.0 ± 18.1	70.1 ± 18.1	64.3 ± 19.5	62.4 ± 22.2	<0.0001	Control < all others
SIQ composite	6.8 ± 2.1	12.8 ± 3.6	13.5 ± 3.6	12.3 ± 3.8	12.0 ± 4.3	<0.0001	Control < all others
SIQ cognitive	1.5 ± 0.7	3.0 ± 1.2	3.3 ± 1.2	3.2 ± 1.1	2.9 ± 1.3	<0.0001	Control < all others
SIQ physiological	1.4 ± 0.4	2.8 ± 1.0	3.0 ± 1.0	2.9 ± 0.9	2.6 ± 1.0	<0.0001	Control < all others
SIQ behavioral	2.6 ± 1.0	4.2 ± 0.8	4.3 ± 0.9	3.5 ± 1.2	4.0 ± 1.3	<0.0001	Control < all others
SIQ emotional	1.4 ± 0.6	2.8 ± 1.3	2.9 ± 1.2	2.8 ± 1.0	2.5 ± 1.2	<0.0001	Control < all others
SI duration (min)	14.8 ± 12.4	44.8 ± 50.4	44.8 ± 39.1	20.5 ± 20.8	40.0 ± 32.3	<0.0001	Control = NT1; Control < all others; NT1 < sEDS and IH

ESS (Epworth Sleepiness Scale); MSL (mean sleep latency); SI (sleep inertia); SIQ (sleep inertia questionnaire); sEDS (subjective excessive daytime sleepiness); IH (idiopathic hypersomnia); NT1 (narcolepsy type 1); NT2 (narcolepsy type 2). Values presented are mean ± standard deviation or number (percent). For demographics and clinical features, p-values reflect effect of diagnosis in ANOVA for continuous variables and Chi-square for categorical variables. For SIQ results, p-values are for effect of diagnosis in ANOVA, controlled for age and center. Pairwise differences are significant at adjusted p < 0.05.

Sleep Evaluation and Questionnaires

As anticipated, ESS and HSI scores were significantly lower in controls than in any patient group (Table 1). There were few differences across CDH groups on these scores, with lower ESS in sEDS than NT1, and lower HSI in sEDS than NT1 or IH, and lower sEDS in NT2 than IH. Similarly, MSLT MSL was significantly higher in controls and sEDS than in the other patient groups, and higher in IH than NT2. IH patients were most likely to be taking any medication at the time of SIQ completion. Depression scale scores consistent with a diagnosis of depression was lower in controls than all other groups.

SIQ Analysis

The SIQ sum score, composite score, and all sub-scores were significantly lower in controls than in all other groups (p<0.0001), without differences among the sleepy participant groups (Table 1), controlling for age and center. Sleep inertia duration was significantly shorter in controls than in all sleepy groups other than NT1; sleep inertia duration in people with NT1 was significantly shorter than in those with IH or sEDS.

We compared the SIQ scores between those with IH with LST and IH without LST. IH with LST demonstrated significantly higher SIQ behavioral scores (4.5 ± 0.5 with LST vs. 4.2 ± 1.0 without LST, p=0.04), but there were no significant differences in SIQ sum scores (70.5 ± 13.4 vs 70.0 ± 19.3, p = 0.92), composite scores (13.6 ± 2.6 vs 13.5 ± 3.9, p = 0.91), SIQ cognitive scores (3.1 ± 1.1 vs 3.3 ± 1.3, p = 0.49), physiologic scores (3.1 ± 0.9 vs 3.0 ± 1.0, p = 0.61), emotional scores (2.8 ± 1.1 vs 3.0 ± 1.3, p = 0.62), or sleep inertia duration (35.3 ± 18.7 minutes vs 47.6 ± 43.1, p = 0.08) between IH with and without LST.

The presence of depression was associated with higher SIQ sum, composite, and subscale scores, as well as, sleep inertia duration (Table 2). However, when controlling for depression (in addition to age and center), the relationship of SIQ sum, composite, most sub-scores, and sleep inertia duration to diagnoses were unchanged. For the behavioral subscale, there was no longer a difference between NT1 and controls. Similarly, being moderately or definitely an evening chronotype on the MEQ was significantly associated with higher SIQ values (Table 3). However, when controlling for eveningness (and age), the relationship of SIQ sum and composite scores by diagnosis was unchanged. Relationships between sub-scores and diagnoses were also largely unchanged, with the exception of a non-significant overall model for SIQ behavioral and no significant difference between NT2 and controls for sleep inertia durations. Limiting analyses to people not on any medications at the time of SIQ completion (n=141), controlling for age and center, the relationship of SIQ sum, composite, and most sub-scores to diagnosis were unchanged. The overall model for SIQ behavioral was non-significant in this smaller sample. For SIQ emotional, control and NT2 participants were no longer significantly different. For SIQ duration, NT1 was no longer significantly different than controls or sEDS.

Table 2.

Depression is associated with SIQ scores.

	Depression (n=83)	No Depression (n=183)	p-value
SIQ sum	78.8 ± 15.4	51.6 ± 19.1	<0.0001
SIQ composite	15.3 ± 3.0	9.8 ± 3.6	< 0.0001
SIQ cognitive	3.8 ± 1.0	2.3 ± 1.2	< 0.0001
SIQ physiological	3.5 ± 0.9	2.1 ± 0.9	< 0.0001
SIQ behavioral	4.4 ± 0.8	3.5 ± 1.2	< 0.0001
SIQ emotional	3.6 ± 1.0	1.9 ± 1.0	< 0.0001
SI duration (min) *	51.6 ± 50.5	28.7 ± 31.3	0.0002

^*Of those participants who completed the sleep inertia duration section, the number of participants reporting depression was n=82, and those reporting no depression was n=178. SI (sleep inertia); SIQ (sleep inertia questionnaire). Values presented are mean ± standard deviation.

Table 3.

Evening type (“night owl”) is more likely to have sleep inertia.

	Night Owl (n=27)	Not Night Owl (n=91)	p-value
SIQ sum	77.4 ± 16.6	55.6 ± 23.0	<0.0001
SIQ composite	14.9 ± 3.3	10.7 ± 4.5	< 0.0001
SIQ cognitive	3.7 ± 1.3	2.6 ± 1.4	0.0002
SIQ physiological	3.3 ± 0.9	2.3 ± 1.1	< 0.0001
SIQ behavioral	4.6 ± 0.5	3.5 ± 1.3	< 0.0001
SIQ emotional	3.2 ± 1.2	2.3 ± 1.2	0.0004
SI duration (min) *	55.7 ± 41.5	31.7 ± 35.8	0.005

^*Of those participants who completed the sleep inertia duration section, the number of participants reporting to be night owls was n=26, and those reporting not to be night owls was n=88. SI (sleep inertia); SIQ (sleep inertia questionnaire). Values presented are mean ± standard deviation.

Across all participants, those who endorsed substantial sleep inertia on the HSI demonstrated higher SIQ sum, composite, subscale and duration values than those who did not (all p-values < 0.0001). Similarly, those who reported needing another person to help them wake up had significantly higher SIQ sum, composite, subscale, and duration values than those who did not (all p-values < 0.003). There was a significant, positive correlation between the number of alarm rings needed to awaken a respondent and higher SIQ scores (Table 4). All SIQ scores demonstrated a moderate positive relationship to both ESS and overall HSI scores, while the sleep inertia duration showed a weak relationship to both ESS and HSI.

Table 4.

Relationship between SIQ and ESS/HSI/number of alarms rings needed to awaken.

	ESS	HSI	Number of alarm rings needed
SIQ sum	r=0.62, p<0.0001	r=0.66, p<0.0001	r=0.40, p<0.0001
SIQ composite	r=0.60, p<0.0001	r=0.66, p<0.0001	r=0.40, p<0.0001
SIQ cognitive	r=0.55, p<0.0001	r=0.58, p<0.0001	r=0.32, p<0.0001
SIQ physiological	r=0.61, p<0.0001	r=0.63, p<0.0001	r=0.31, p=0.0001
SIQ behavioral	r=0.53, p<0.0001	r=0.58, p<0.0001	r=0.44, p<0.0001
SIQ emotional	r=0.46, p<0.0001	r=0.55, p<0.0001	r=0.36, p<0.0001
SI duration (min)	r= 0.34, p<0.0001	r=0.35, p<0.0001	r=0.31, p=0.0003

ESS (Epworth Sleepiness Scale); HSI (Hypersomnia Severity Index); SI (sleep inertia); SIQ (sleep inertia questionnaire).

In distinguishing hypersomnolent patients from controls, sum scores and composite scores performed similarly (AUC 0.92 for both, Figure 1). SIQ sub-scores also had excellent discriminative ability (AUCs 0.85 to 0.93), with sleep inertia duration having AUC 0.76. Optimal cutpoint for the SIQ sum score in distinguishing hypersomnolent versus control groups was 42 (Youden’s index = 0.71). The optimal cutopoint for sleep inertia duration in distinguishing these groups was 15 minutes (Youden’s index = 0.42).

Figure 1:

Use of the SIQ sum score to distinguish CDH patients versus controls (A) and SIQ duration to distinguish IH patients versus NT1 patients (B)

In distinguishing IH from NT1 patients, the duration of sleep inertia symptoms had the highest AUC (0.73), with an optimal cutpoint of 25 minutes (Youden’s index = 0.39), and the SIQ behavioral subscale was similarly effective (AUC 0.70). The SIQ sum score, composite score, and remaining SIQ subscales did not perform well at distinguishing IH and NT1 patients (AUCs 0.53–0.59, model p-values > 0.05).

None of the SIQ outcomes evaluated were able distinguish participants with IH from those with NT2 (AUCs 0.53–0.62, model p-values all > 0.05) or those with sEDS (AUCs 0.47–0.58, model p-values all > 0.05).

Discussion:

We have demonstrated that the SIQ has an excellent ability to distinguish hypersomnolent patients from healthy controls, even after controlling for depression and eveningness chronotype, while sleep inertia duration showed best ability to distinguish IH from NT1, relative to other assessed sleep inertia variables. We present novel data on salient SIQ cutoff scores for these comparisons.

In addition to analyzing the sum scores and four individual constructs (physiological, cognitive, behavioral, emotional) within the SIQ, we derived a SIQ composite score, to equally weigh each of the four subscales despite different numbers of questions in each subscale. We anticipated that such a composite score may be necessary, if relative contribution of individual constructs differed among hypersomnolent groups. However, we demonstrated very similar performance between the sum and composite scores throughout our analyses. Given that generating a single sum is easier than creating a weighted average of subscales, but with equal performance, we have reported cutoff values for the sum score, i.e., scores of 42 and higher provide excellent discrimination between patients and controls.

Furthermore, our data show that the SIQ corresponds well to other measures of sleep inertia such as the HSI single question item (i.e. construct validity): “Please consider your sleep in the past month. To what extent do you think that you have difficulty waking up in the morning or from naps?” The HSI is a self-reported measure of hypersomnolence that was developed and validated in psychiatric disorders¹⁹ but also is useful in sleep disorders.²⁵ However, the HSI only includes one question item that relates specifically to sleep inertia symptom. The SIQ, on the other hand, adds further value, when compared to existing scales, such as the HSI, because the SIQ includes 22 question items which are addressing symptoms specific to sleep inertia represented by four different constructs of sleep inertia. The SIQ may be a preferred choice of questionnaire in evaluating patients with CDH, in whom sleep inertia is a major debilitating symptom and target of treatment. Across all participants in our study, those with excessive sleep inertia quantified by the HSI, had consistently higher SIQ scores, sub-scores, and duration. Moreover, our study showed that the HSI score was lower in controls than all other groups, as we had expected; the somewhat lower HSI score in the sEDS group, along with their higher MSL, might suggest that they have a less severe phenotype, when compared to IH and NT1 (Table 1).

A future study should evaluate how well the SIQ corresponds to other measures that also include assessment of sleep inertia, such as the IHSS.²⁶ This questionnaire has been validated for the evaluation of IH severity, showing adequate psychometric properties²⁷. It assesses IH symptoms broadly, including 3 items about sleep inertia and sleep drunkenness after nighttime sleep, and 1 item after daytime naps.

Because sleep inertia can be experienced by people with depression and delayed sleep phase syndrome,^3,9 we compared the measures across diagnoses while controlling for depression and delayed chronotypes. As expected, the presence of depression or eveningness (‘night owl’) was associated with higher SIQ scores (Tables 2, 3). Importantly, our study confirmed that the statistical differences in SIQ composite score between healthy controls and hypersomnolent patients persisted even after controlling for depression and eveningness on the MEQ, indicating that the SIQ is a reliable measure of sleep inertia severity in sleep disorders despite the potential confounders of depression or delayed chronotype.

Previous studies reported that sleep inertia in healthy controls can take between 15 and 30 minutes to dissipate, with performance improving over time.^3,28 Other studies have found that performance may not stabilize until 2 hours post-waking, even after a sufficient sleep opportunity of 8 hours of nocturnal sleep.²⁹ The wide range of previously reported sleep inertia duration may be due to varied types of performance measurements and/or heterogenous group of subjects in the studies, because the intensity and duration of sleep inertia vary based on situational factors (abrupt awakening during slow wave sleep, biological night, and/or prior degree of sleep deprivation.)²⁸ Although duration of sleep inertia on the SIQ was not the most discriminative feature between hypersomnolent patients and controls, a cutoff of 15 minutes duration was best to distinguish these groups. In contrast, for the distinction between IH and NT1, duration of sleep inertia was the most discriminative feature, with an optimal cutpoint of 25 minutes.

An important question in IH, still unresolved, is the potential clinical significance of a distinction between those with and without LST.³⁰ This distinction, previously described in the ICSD-2, was subsequently removed.^4,9,31 However, studies have shown that those who endorsed habitual long sleep (10 hours/night or longer) were more likely to endorse sleep inertia, suggesting an important phenotypic distinction based on sleep duration cutoffs of 10 or 11 hours;^16,32 this was consistent with the findings of a cluster analysis in which IH with LST clustered separately from IH without LST, with the LST cluster having the most difficulty waking from naps.¹¹ Our study showed that IH with LST demonstrated significantly higher SIQ behavioral scores than IH without LST (p=0.04), but there were no significant differences in the other SIQ scores (SIQ composite, physiological, cognitive, emotional), or in sleep inertia duration. This suggests substantial overlap in frequency and severity of sleep inertia in both forms of IH.

A key challenge in characterizing sleep inertia within CDH patients is that there is no single gold standard diagnostic tool for pathologic sleep inertia. The psychomotor vigilance test (PVT), a 10-minute reaction time test, has been evaluated as one such objective measure.^33,34 PVT performance was significantly impaired in patients with CDH compared with non-hypersomnolent controls at baseline, and particularly worsened post-nap in the morning for the CDH group compared with controls.³³ PVT performance was at its worst, particularly at 07:00 am and 07:30 am, in patients with long severe sleep inertia and sleep drunkenness, regardless of the sleep disorder type.³⁴ Future studies are needed to validate the SIQ against objective measures, such as the PVT, although self-reported and objective measures may also provide complementary information. Furthermore, a PVT device is not always available in every sleep clinic, and this test is more time-consuming, requiring additional staff to educate and monitor patients until the tests are completed. Therefore, questionnaires, such as the SIQ, are valuable in evaluating severe sleep inertia or “sleep drunkenness”. However, it is noteworthy that the SIQ was unable to distinguish objective IH from sEDS or from NT2, likely related to the substantial symptom overlap – and possible mechanistic overlap -- in these three groups.

This study has strengths and limitations. The strengths include being a multicenter study with a large sample size. Another strength is that our data compared the SIQ measures across diagnoses while controlling for depression and chronotypes, as well as evaluated for SIQ cutoff scores and sleep inertia duration cutoffs, providing novel data.

There are several limitations to this study as well. Some patients were on medications at the time of SIQ completion. However, use of medications did not significantly impact our results, such that the relationship of SIQ sum, composite, and most sub-scores to diagnosis remained unchanged when the analysis was limited to non-medicated participants. This may reflect the fact that medicated participants often presented to our facilities for re-evaluation due to lack of benefit of their current medications. For most medications used to treat CDHs, there is very limited data evaluating the extent to which they do or do not help with sleep inertia, although more recent studies have begun to incorporate this assessment. Including assessment of patients on medications may provide more generalizable data and a more “real-world” clinical scenario as we treat our patients with CDH.

Another limitation of the study is that the sEDS group is likely a heterogenous group of various sleep disorders, i.e. they did not meet ICSD-3 criteria for CHD or OSA on PSG/MSLT, therefore are not clearly representing a specific sleep diagnosis. There were fewer sEDS participants at the Emory center than the Wisconsin center, which emphasizes the benefit collecting data from multiple centers. The narcolepsy groups were smaller than the other groups. Not all participants in our control group completed PSG/MSLT or were directly evaluated by a sleep physician; as sleep disorders are quite frequent in the general population, this may pose another limitation.

Finally, a potential limitation of the current SIQ is that it instructs the participant to rate each item based “On a typical morning in the past week, after you wake up..”; however, the past week may not always be representative of regular sleep habits. Therefore, this questionnaire might require rephrasing to obtain more appropriate responses based on the participant’s regular sleep habits (e.g. “On a typical morning in recent times, after you wake up from your regular sleep habits, to what extend do you..”).

Despite these limitations, we have demonstrated that the SIQ has excellent ability in distinguishing hypersomnolent patients from healthy controls. Future studies should assess the SIQ for test-retest reliability, treatment effects, and other psychometric properties.

• Sleep inertia can be disabling in some central disorders of hypersomnolence.

• Optimal tools to measure sleep inertia severity have not yet been determined.

• The Sleep Inertia Questionnaire (SIQ) distinguishes sleepy patients from controls.

• Optimal SIQ sum cutpoint to distinguish patients from controls was 42 (J = 0.71).

• Sleep inertia duration distinguishes Idiopathic Hypersomnia from Narcolepsy Type 1.