Profile of subjective-objective sleep discrepancy in patients with insomnia and sleep apnea

Abstract

Study Objectives:

Although subjective-objective sleep discrepancy has long been observed in patients with insomnia, the profiles of this discrepancy are poorly understood. Further, sleep discrepancy in insomnia with sleep comorbidities remains underexplored. We sought to better characterize sleep discrepancy among patient groups with and without insomnia and comorbid conditions such as obstructive sleep apnea (OSA).

Methods:

Using data from the Sleep Heart Health Study, we conducted a secondary analysis describing (1) the profile of self-reported and objective sleep measures in patients with insomnia (IS group; n = 73) and comorbid OSA (IS + OSA group; n = 143), compared with individuals with OSA only (OSA group; n = 296) and normal sleep control patients (NSC group; n = 126); (2) the comparative magnitude of sleep misperception between these 4 groups; and (3) the self-reported quality of life (QOL) in the 4 groups.

Results:

Subjective-objective sleep discrepancy existed in all 4 groups, including the NSC group. Controlling for age, sex, mental health conditions, sleep apnea severity, and objectively measured sleep time, the presence of self-reported insomnia had the strongest association with sleep discrepancy. In patients with insomnia, sleep onset latency was overestimated (7.8 ± 36.8 min in the IS group; P < .001 when compared to the NSC and OSA groups), with the largest differences seen in the comorbid IS + OSA group (15.0 ± 56.8 min). Insomnia conferred the most negative impact on QOL, with the combined IS + OSA group reporting the lowest QOL.

Conclusions:

Self-reported insomnia is associated with sleep discrepancy and negative QOL. Those with comorbid OSA reported the greatest sleep discrepancy and the lowest QOL. Future research is warranted to further understand individual profiles of misperception and insomnia phenotypes.

Citation:

Ma Y, Goldstein MR, Davis RB, Yeh GY. Profile of subjective-objective sleep discrepancy in patients with insomnia and sleep apnea. J Clin Sleep Med. 2021;17(11):2155–2163.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Differences between self-reported and objectively measured sleep parameters, known as subjective-objective sleep discrepancy or sleep misperception, have long been observed in patients with insomnia. Research is needed to more directly and comprehensively understand the associations between sleep discrepancy, insomnia, comorbid conditions such as sleep apnea, and their effects on overall health.

Study Impact: Patients with self-reported insomnia with comorbid sleep apnea report the greatest sleep discrepancy and the lowest quality of life. Future research is needed to better understand individual profiles of misperception and insomnia phenotypes.

INTRODUCTION

Differences between self-reported and objectively measured sleep parameters have long been observed in patients with insomnia and even in healthy individuals. Terminology used to describe such phenomenon includes subjective-objective sleep discrepancy, sleep (state) misperception, subjective-objective mismatch in sleep perception, sleep discrepancy, error in sleep estimation, or sleep perception error. The majority of existing studies on sleep discrepancy have been conducted in insomnia, the most common sleep disorder, with approximately 10%–30% prevalence in the general population.¹ Rather than being an issue of the accuracy of assessments or reporting errors, recent research suggests that sleep discrepancy is a clinically meaningful feature of insomnia and may affect the maintenance and treatment of insomnia.² Sleep discrepancy has been reported in patients with insomnia with Alzheimer disease,³ chronic fatigue syndrome,⁴ borderline personality disorder,⁵ major depressive disorder,⁶ and attention-deficit/hyperactivity disorder,⁷ and across the age spectrum in adolescent children⁸ and older individuals.⁹

The co-occurrence of insomnia and obstructive sleep apnea (OSA) is well documented and highly prevalent, ranging from 55%–84% of patients presenting to sleep clinics.¹⁰ Both conditions cause common impairments that largely overlap^11–13 and have negative impacts on operational readiness, personal health, well-being, and health care costs. Patients with OSA and insomnia experience nonrestorative sleep, fatigue, daytime sleepiness, decreased daytime function, cognitive dysfunction, mood disorder, and increased risks of chronic conditions. A diagnosis of OSA, however, is based on objective polysomnography (PSG) parameters, whereas insomnia diagnosis does not require a laboratory PSG unless there is a suspicion of another disorder. Patients with comorbid disease present challenges for clinical management and a standardized systematic approach to identifying and treating these sleep disorders when they co-occur.¹⁴ Notably, the high comorbidity between insomnia and OSA has made research on sleep discrepancy in these subpopulations both clinically and mechanistically relevant.

There is an increasing body of literature on sleep discrepancy in insomnia and OSA, although much remains unclear. Choi et al¹⁵ reported that accurate sleep perception was positively associated with the presence of OSA, although it was negatively related to the presence of insomnia. Bianchi et al¹⁶ reported that patients with insomnia symptoms, OSA, or both conditions all overestimated sleep latency but underestimated wake after sleep onset (WASO). Both studies were conducted using PSG in laboratory settings, which can alter sleep quality and increase sleep fragmentation¹⁷ and thus limit the generalizability of the results in the real world. To add complexity and heterogeneity to this area of study, it is likely that sleep discrepancy varies both inter- and intraindividually in magnitude, direction, and night-to-night frequency in most patients with insomnia.¹⁸ The corresponding heterogeneity of research findings, unclear thresholds for the characterization of sleep misperception, unknown neurophysiological mechanisms, and challenging treatment continue to make this an important area of research.

Within this context, we conducted a secondary analysis of a large, publicly available database—the Sleep Heart Health Study—aiming to (1) better characterize and describe the profile of self-reported and objective sleep measures in patients with insomnia with and without OSA; (2) explore patterns of sleep discrepancy and factors associated with sleep discrepancy in patients with insomnia, OSA, and their comorbid presentation; and (3) examine quality of life (QOL) in patients with insomnia with and without OSA.

METHODS

Dataset and included individuals

Data used in this analysis were derived from the Sleep Heart Health Study, a multicenter cohort study implemented by the National Heart, Lung, and Blood Institute to determine the cardiovascular and other consequences of sleep-disordered breathing. Participants’ baseline visits were between 1995 and 1998 from existing community-based cohorts as described previously.^19,20 Eligible participants were at least aged 40 years and were not receiving active treatment for OSA (eg, continuous positive airway pressure, oral appliance, and oxygen therapy).^19,20 At baseline visits, trained research investigators conducted interviews. Sociodemographic characteristics, sleep habits, cardiometabolic factors, overall health, medication use, anthropometry and blood pressure measurements, and questionnaires (eg, the Epworth Sleepiness Scale [ESS]) were included. The dataset included 5,805 patients who successfully completed baseline PSG at home using a portable monitor (Compumedics P-series, Abbotsford, Victoria, Australia). Sensors were placed and equipment was calibrated during the evening home visit by a certified technician. Recordings included electroencephalogram (EEG), electrooculogram, electrocardiogram, chin electromyogram, pulse oximetry, chest and abdominal excursion by inductance plethysmography, airflow by thermal sensor, and body position channels. All PSG recordings were scored using Rechtschaffen and Kales criteria at a centralized reading center by trained research technicians blinded to all clinical data.²¹

From the original dataset, we excluded the data if (1) overall PSG signal quality was less than “good” (all data were labeled with different levels of quality after manual check), (2) more than 30 minutes of the sleep period had either lost or unscorable EEG data, (3) recording either started or ended in sleep, (4) the sleep latency was not reliable, or (5) patients had substance use (eg, liquor, coffee, usual medication) before sleep.

At baseline, participants completed a sleep habits questionnaire that elicited standardized information on sleep symptoms and patterns. Four insomnia-related questionnaire items²² were as follows: “Have trouble falling asleep,” “Wake up during the night and have difficulty getting back to sleep,” “Wake up too early in the morning and be unable to get back to sleep,” and “Take sleeping pills or other medication to help you sleep.”

Patients with insomnia were identified based upon self-report of at least 1 of the above symptoms with affirmative responses occurring almost always (16–30 nights) per month. Patients without insomnia were identified if they never (zero nights per month) or rarely (< 1–2 nights per month) had any of the above symptoms.

Given our interest in evaluating sleep discrepancy in phenotypes, we further classified the patients into 4 subgroups according to the apnea-hypopnea index (AHI; ≥ 3% oxygen desaturation recorded per hour of sleep): (1) normal sleep control patients (NSC group; no insomnia, AHI < 5 events/h), (2) OSA only (OSA group, no insomnia, AHI ≥ 5 events/h), (3) insomnia only (IS group, insomnia, AHI < 5 events/h), and (4) IS with OSA (IS + OSA group, IS with AHI ≥ 5 events/h).

This is a secondary analysis with only publicly available deidentified data. It does not involve a research protocol requiring approval by the relevant institutional review board or ethics committee.

Self-reported and objective sleep measurements

Self-reported sleep measurements were derived from the morning survey that was conducted after the sleep study night. Questions in the morning survey included the following: “How long did you sleep last night” (defined as self-reported total sleep time [sTST]); “How many minutes did it take for you to fall asleep at bedtime last night” (defined as self-reported sleep onset latency [sSOL]); “Did you have difficulty falling asleep last night?” (yes or no); “On a scale of 1–5 (1 = short, 5 = long), how was the length of your sleep last night?”; “On a scale of 1–5 (1 = light, 5 = deep), how was the depth of your sleep last night?”; and “On a scale of 1–5 (1 = restless, 5 = restful), how was the quality of your sleep last night?” In addition, the ESS was used to assess daytime sleepiness. The ESS is a self-administered questionnaire with 8 questions. Respondents were asked to rate, on a 4-point scale (0–3), their usual chances of dozing off or falling asleep while engaged in 8 different activities.^23,24 The ESS score ranges from 0–24, with the higher score indicating more sleep propensity in daily life or a higher level of daytime sleepiness.

Objective sleep measurements were derived from PSG scoring, including objective TST (oTST), objective SOL (oSOL), rapid eye movement (REM) sleep latency, sleep efficiency, arousal index (total number of arousals per hour of sleep), WASO, number of stage shifts per hour, and percentage of the time spent in sleep stage 1, sleep stage 2, sleep stages 3 and 4, and REM sleep.

QOL

QOL was measured by the 36-Item Short Form Health Survey (SF-36), a self-report generic measure of health-related QOL that has been validated.²⁵ The SF-36 is composed of 8 multi-item subscales including physical functioning, role physical, bodily pain, general health, vitality, social functioning, role emotional, and mental health. The scores on the SF-36 can also be reported as a physical component summary and mental health component summary (MCS) that represent general physical and mental health status, respectively. The 8 scaled subscores are the weighted sums of the questions in each section. Each scale is directly transformed into a 0–100 scale on the assumption that each question carries equal weight. All scores used in this analysis were standardized, and lower scores indicated more disability.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics, version 25 (IBM Corp., Armonk, NY). Baseline characteristics and sleep measurements were reported for each subgroup. Sleep measurement variables with normal distributions were reported as mean ± standard deviation, and variables with nonnormal distributions were reported as median with first and third quartiles (Q₁, Q₃). Demographic, clinical, and sleep features were compared across groups. Categorical variables were tested using the chi-squared test or the Fisher exact test where appropriate. Because most of the continuous variables followed nonnormal distribution, nonparametric tests were used for comparisons. The comparisons between patients with and without insomnia in the 2 major groups were tested using the Mann-Whitney U test. Differences across the 4 subgroups were tested using the Kruskal-Wallis H test. If the global test was significant, then pairwise comparisons were conducted using Dunn’s test, and results were reported only for primary targets on the comparison between the NSC and IS groups and between the OSA and IS + OSA groups. For the sleep study night, 2 regression models using sleep discrepancy as a dependent variable were created to understand the factors (age, sex, MCS scores, AHI, insomnia, and objective sleep measurements) associated with sleep outcome measures in the studied population. Sleep discrepancy in this study was defined as the difference between self-reported measurement and objective measurement of TST (sTST–oTST) and of SOL (sSOL–oSOL). Positive values indicated overestimation on time, and negative values indicated underestimation on time.

We also created models with interaction between insomnia and the severity of apnea (indicated by AHI) but reported the results from models without the interaction term because the interactions were not significant. In addition, we created regression models using physical component summary and MCS scores as outcome measures separately to understand the chronic impact of the presence of insomnia. P values < .05 indicated significant difference.

RESULTS

Demographic characteristics

Among the eligible participants, 422 patients with no insomnia (n = 126 in the NSC group and n = 296 in the OSA group) and 216 patients with insomnia (n = 73 in the IS group and n = 143 in the IS + OSA group) were identified (Table 1). The only demographic characteristics with significant differences between groups were sex and hypertension. Patients with insomnia tended to be female compared with patients without insomnia. Male patients were less predominant in groups without sleep apnea (NSC and IS groups). Hypertension was more prevalent among participants with insomnia compared to those without insomnia (P = .027), and the highest prevalence was observed in patients with combined insomnia and OSA. No significant differences were found between groups or between subgroups with respect to age, BMI, race, ethnicity, education level, marital status, diabetes, or ESS score.

Table 1.

Demographic characteristics.

	No Insomnia		Insomnia		No Insomnia vs Insomniaa	Subgroup Comparisons
	NSC (n = 126)	OSA (n = 296)	IS (n = 73)	IS + OSA (n = 143)		Subgroupsb	NSC vs ISc	OSA vs IS + OSAc
Age (y)	58.7 ± 11.1	64.6 ± 11.0	63.6 ± 11.9	65.3 ± 10.8	0.060	< 0.001	0.028	0.988
Male (%)	48 (38.1%)	189 (63.9%)	13 (17.8%)	65 (45.5%)	< 0.001	< 0.001	0.003	< 0.001
BMI	25.6 ± 4	28.9 ± 4.9	26.2 ± 3.5	29.3 ± 5.3	0.236	< 0.001	0.846	0.949
Race					0.218	0.052	—	—
White	96 (76.2%)	242 (81.8%)	58 (79.5%)	126 (88.1%)
Black	14 (11.1%)	27 (9.1%)	3 (4.1%)	10 (7.0%)
Other	16 (12.7%)	27 (9.1%)	12 (16.4%)	7 (4.9%)
Ethnicity					0.564	0.023	0.115	0.415
Hispanic or Latino	10 (7.9%)	18 (6.1%)	11 (15.1%)	6 (4.2%)
Other	116 (92.1%)	278 (93.9%)	62 (84.9%)	137 (95.8%)
Education level (y)					0.290	0.193	—	—
< 10	4 (3.9%)	20 (7.4%)	3 (5.1%)	12 (9.0%)
11–15	49 (48.0%)	146 (54.1%)	31 (52.5%)	83 (62.4%)
16–20	46 (45.1%)	92 (34.1%)	23 (39.0%)	34 (25.6%)
> 20	3 (2.9%)	12 (4.4%)	2 (3.4%)	4 (3.0%)
Marital status					0.159	0.116	—	—
Single	6 (4.8%)	5 (1.7%)	3 (4.1%)	2 (1.4%)
Married	95 (75.4%)	246 (83.1%)	51 (69.9%)	108 (75.5%)
Widowed/divorced/separated	23 (18.3%)	40 (13.5%)	18 (24.7%)	29 (20.3%)
Unknown/refused	2 (1.6%)	5 (1.7%)	1 (1.4%)	4 (2.8%)
Hypertension	46 (36.5%)	132 (44.6%)	34 (46.6%)	77 (53.8%)	0.027	0.041	0.163	0.069
Diabetes	9 (7.4%)	27 (9.7%)	2 (2.8%)	14 (10.2%)	0.584	0.249	—	—
ESS score	6.0 (3.0, 10.0)	7.0 (4.0, 10.0)	5.5 (3.8, 9.3)	7.0 (4.3, 11.0)	0.467	0.034	0.999	0.516

^aComparisons between patients with and without insomnia. ^bComparisons among the 4 subgroups. ^cPairwise comparisons if global test was significant. BMI = body mass index, ESS = Epworth Sleepiness Scale, IS = insomnia group, NSC = normal sleep control patient group, OSA = obstructive sleep apnea group.

Self-reported sleep measurements

The groups without insomnia (NSC and OSA groups) had longer sTST and shorter sSOL than those with insomnia (IS and IS + OSA groups) (both P < .001; Table 2). Subgroup comparisons indicated that insomnia was associated with significantly lower sTST and significantly longer latency among patients with OSA (OSA vs IS + OSA groups), suggesting worsened self-reported sleep with comorbid insomnia and OSA.

Table 2.

Self-reported and objective sleep measurements.

	No Insomnia		Insomnia		No Insomnia vs Insomniaa	Subgroup Comparisons
	NSC (n = 126)	OSA (n = 296)	IS (n = 73)	IS + OSA (n = 143)		Subgroupsb	NSC vs ISc	OSA vs IS + OSAc
Self-reported measures
Difficulty falling asleep	31 (25.0%)	84 (29.5%)	24 (33.3%)	62 (44.6%)	0.001	0.003	0.211	0.002
Sleep length	3.1 ± 1.0	3.3 ± 1.0	2.8 ± 1.2	2.6 ± 1.2	< 0.001	< 0.001	0.553	< 0.001
Sleep depth	3.3 ± 1.1	3.3 ± 1.0	3.0 ± 1.2	2.6 ± 1.2	< 0.001	< 0.001	0.551	< 0.001
Restful sleep	3.2 ± 1.2	3.2 ± 1.1	2.9 ± 1.2	2.6 ± 1.3	< 0.001	< 0.001	0.183	< 0.001
sTST (min)	402.1 ± 64.7	404.2 ± 69.8	366.7 ± 105.0	356.2 ± 98.4	< 0.001	< 0.001	0.064	< 0.001
sSOL (min)	15 (5, 20)	15 (10, 30)	20 (10, 30)	30 (15, 45)	< 0.001	< 0.001	0.106	< 0.001
Objective measures
oTST (min)	368.6 ± 46.0	357.6 ± 53.0	351.0 ± 66.5	352.0 ± 57.9	0.086	0.072	—	—
oSOL (min)	15.25 (9.5, 24.75)	15.0 (10.0, 26.0)	17 (8.5, 30)	19 (9, 31.5)	0.091	0.348	—	—
Sleep efficiency (%)	86.4 (80.7, 91.2)	84.3 (77.2, 89.2)	83.4 (75.8, 90.4)	82.9 (75.3, 87.8)	0.003	0.001	0.107	0.086
Stage 1 sleep (%)	4.4 ± 2.7	5.3 ± 3.9	4.7 ± 4.2	5.3 ± 3.8	0.650	0.116	—	—
Stage 2 sleep (%)	55.1 ± 11.3	57.6 ± 11.5	53.1 ± 13.5	56.9 ± 11.5	0.128	0.012	0.859	0.995
Stages 3 and 4 sleep (%)	19.8 ± 11.4	17.1 ± 11.7	23.0 ± 12.9	18.9 ± 11.2	0.008	0.001	0.587	0.119
REM sleep (%)	20.6 ± 6.0	20.0 ± 6.1	19.3 ± 6.4	19.0 ± 6.1	0.030	0.115	—	—
REM sleep latency (min)	70 (52.5, 97.25)	69 (51, 99.5)	70.5 (50.75, 113.25)	77.5 (75.75, 117.25)	0.028	0.104	—	—
AHI (events/h)	2.6 (1.4, 3.9)	14.6 (9.5, 24.9)	2.6 (1.3, 4.0)	14.6 (8.4, 23.3)	0.261	< 0.001	1.000	1.000
AI (events/h)	13.1 (10.0, 17.0)	18.8 (13.0, 25.6)	13.8 (9.3, 17.8)	17.8 (12.6, 23.4)	0.222	< 0.001	1.000	< 0.001
WASO (min)	37 (20, 66)	45.5 (25.625, 71)	45.5 (28.75, 77.5)	50 (31.5, 73)	0.021	0.011	0.572	0.485
Stage shifts (/h)	6.6 (5.5, 8.2)	7.3 (5.6, 9.0)	6.7 (5.3, 7.9)	7.0 (5.4, 8.8)	0.305	0.103	—	—
Differences
sTST–oTST (min)	23.9 ± 86.0	33.0 ± 100.7	10.7 ± 103.5	–5.8 ± 106.1	< 0.001	< 0.001	0.927	0.002
sSOL–oSOL (min)	0.6 ± 27.9	–1.8 ± 28.3	7.8 ± 36.8	15.0 ± 56.8	< 0.001	< 0.001	0.614	0.006

^aComparisons between patients with and without insomnia. ^bComparisons among the 4 subgroups. ^cPairwise comparisons if global test was significant. AHI = apnea-hypopnea index, AI = arousal index, IS = insomnia group, NSC = normal sleep control patient group, OSA = obstructive sleep apnea group, oSOL = objectively reported sleep onset latency, oTST = objectively reported total sleep time, REM = rapid eye movement, SOL = sleep onset latency, sSOL = self-reported sleep onset latency, sTST = self-reported total sleep time, TST = total sleep time, WASO = wake after sleep onset.

Compared with the NSC group, the percentage of patients who reported having difficulty falling asleep was higher in the OSA group, even higher in the IS group, and highest in the combination group (IS + OSA group). Significant differences were also observed between patients with and without insomnia on each of the self-reported measures (Table 2).

Objective sleep measurements

Differences in oTST and oSOL were not as consistently evident as those from self-reported measurements (Table 2). Compared to the sleep efficiency of the NSC group (86.4%), lower sleep efficiency was found in the OSA (84.3%), IS (83.4%), and IS + OSA (82.9%) groups. Differences between the NSC group vs the IS group and the OSA group vs the IS + OSA were not significant on objective sleep measurements, and only some differences were observed by adding insomnia. Significant differences were observed on the percentages of stages 3 and 4 sleep, REM sleep, REM latency, and WASO between patients with and without insomnia.

Subjective-objective sleep discrepancy

Sleep discrepancy was observed in all the defined subgroups and even in the NSC group (Table 2). Subgroup comparisons indicated significant differences in sTST–oTST between the OSA and IS + OSA groups (P = .002), with overestimation of TST in the OSA group and underestimation of TST in the IS + OSA group. Significant differences were also found in sSOL–oSOL between the OSA and IS + OSA groups (P = .006) and between the NSC and IS + OSA groups (P = .046), with significant SOL overestimation in patients with insomnia and an even greater discrepancy in the comorbid IS + OSA group.

Based on the response to the morning survey question regarding difficulty falling asleep in the study night, we observed significant differences between patients with and without insomnia. In addition, in each of the 2 groups of patients with and without insomnia, a larger degree of subjective-objective discrepancy was noticed in patients who believed they had difficulty falling asleep (Figure 1). Whereas patients who reported no difficulty falling asleep slightly underestimated their SOL, those who reported difficulty substantially overestimated their SOL.

Figure 1. Sleep discrepancy in patients with and without difficulty falling asleep.

oSOL = objectively reported sleep onset latency, sSOL = self-reported sleep onset latency. The asterisk and open circles represent potential outliers.

In regression models with adjustment for age, sex, mental health dimensions (indicated by MCS scores), and AHI, subjective-objective discrepancy on TST was negatively correlated with the presence of insomnia (β = –34.4; 95% confidence interval, –52.6 to –16.2; P < .001) and oTST (β = –0.39; 95% confidence interval, –0.53 to –0.24; P < .001). Similarly, with adjustment of the same variables, discrepancy on SOL was positively correlated with the presence of insomnia (β = 14.7; 95% confidence interval, 8.6–20.7; P < .001) and negatively correlated with oSOL (β = –0.57; 95% confidence interval, –0.69 to –0.46; P < .001).

Insomnia and QOL

Significant differences were observed for all 8 subscales of the SF-36 (P < .001) in the comparisons between patients with and without insomnia and in the subgroup comparisons of interest (Table 3). These results indicated that adding insomnia components led to significant lowered scores on all QOL dimensions. For overall physical health measured by physical component summary scores, compared to the NSC group, lower scores were observed in the OSA and IS groups but the lowest scores were in the IS + OSA group. For overall mental health measured by MCS scores, no significant differences were seen between subgroups with and without OSA, but differences were significant between subgroups with and without insomnia (Table 3). Regression models (Table 4) indicated that age, sex, the presence of insomnia, and daytime sleepiness were significant correlates for physical health (physical component summary), whereas age, the presence of insomnia, daytime sleepiness, and education level were significant predictors for mental health (MCS).

Table 3.

QOL measurements from SF-36 questionnaire.

	No Insomnia		Insomnia		No Insomnia vs Insomniaa	Subgroup Comparisons
	NSC (n = 126)	OSA (n = 296)	IS (n = 73)	IS + OSA (n = 143)		Subgroupsb	NSC vs ISc	OSA vs IS + OSAc
Eight dimensions
Physical functioning	89.6 ± 15.0	81.2 ± 22.9	76.1 ± 23.4	70.9 ± 25.8	< 0.001	< 0.001	< 0.001	0.001
Role physical	90.3 ± 25.5	81.9 ± 33.4	61.0 ± 39.4	64.2 ± 42.7	< 0.001	< 0.001	< 0.001	< 0.001
Bodily pain	81.7 ± 18.7	76.1 ± 21.8	66.2 ± 22.6	63.1 ± 26.7	< 0.001	< 0.001	< 0.001	< 0.001
General health	79.4 ± 18.0	76.2 ± 17.2	65.7 ± 21.7	66.3 ± 21.9	< 0.001	< 0.001	< 0.001	< 0.001
Vitality	72.8 ± 17.7	68.8 ± 17.9	56.3 ± 19.3	52.5 ± 24.5	< 0.001	< 0.001	< 0.001	< 0.001
Social functioning	95.6 ± 10.8	93.3 ± 14.1	82.1 ± 23.0	81.3 ± 23.2	< 0.001	< 0.001	< 0.001	< 0.001
Role emotional	97.7 ± 9.5	95.4 ± 14.4	87.4 ± 23.2	91.6 ± 19.4	0.001	< 0.001	0.010	0.324
Mental health	85.3 ± 11.8	84.4 ± 12.5	71.7 ± 15.2	72.4 ± 19.3	< 0.001	< 0.001	< 0.001	< 0.001
Overall summary scores
Physical component score	51.7 ± 6.9	48.4 ± 9.0	45.0 ± 10.4	43.4 ± 11.8	< 0.001	< 0.001	< 0.001	< 0.001
Mental component score	56.0 ± 6.1	56.1 ± 6.4	49.7 ± 9.4	50.3 ± 10.4	< 0.001	< 0.001	< 0.001	< 0.001

^aComparisons between patients with and without insomnia. ^bComparisons among the 4 subgroups. ^cPairwise comparisons if global test was significant. IS = insomnia group, NSC = normal sleep control patient group, OSA = obstructive sleep apnea group, QOL = quality of life, SF-36 = 36-Item Short Form Health Survey.

Table 4.

Regression analysis on SF-36–measured QOL.

A	Coefficients		95% CI for β		t	P
	β	Standard Error	Lower Bound	Upper Bound
Intercept	65.177	3.723	57.863	72.492	17.507	< .001
Age	–0.200	0.039	–0.277	–0.123	–5.117	< .001
Sex (male = 1)	–2.172	0.855	–3.851	–0.493	–2.541	.011
AHI	–0.050	0.028	–0.106	0.006	–1.769	.077
Insomnia	–3.961	0.878	–5.687	–2.236	–4.510	.000
ESS	–0.295	0.095	–0.481	–0.109	–3.109	.002
Education levels	1.252	0.648	–0.021	2.525	1.932	.054
B	Coefficients		95% CI for β		t	P
	β	Standard Error	Lower Bound	Upper Bound
Intercept	48.784	3.074	42.746	54.823	15.872	< .001
Age	0.077	0.032	0.013	0.140	2.376	.018
Sex (male = 1)	0.585	0.706	–0.801	1.971	0.829	.408
AHI	0.021	0.023	–0.025	0.067	0.909	.364
Insomnia	–5.040	0.725	–6.464	–3.615	–6.950	< .001
ESS	–0.197	0.078	–0.351	–0.043	–2.513	.012
Education levels	1.077	0.535	0.026	2.128	2.013	.045

(A) Outcome: physical component score. (B) Outcome: mental component score. AHI = apnea-hypopnea index, CI = confidence interval; ESS = Epworth Sleepiness Scale, QOL = quality of life, SF-36 = 36-Item Short Form Health Survey.

DISCUSSION

Our analysis showed that subjective-objective sleep discrepancy (aka, sleep misperception) existed in patients in each of the 4 subgroups: IS, OSA, IS + OSA, and even in NSC. Patients with self-reported insomnia complaints had misperception or bias regarding their sleep difficulties, with the tendency to overestimate the time they spent falling asleep and to underestimate their TST. On the other hand, patients who reported no difficulty falling asleep slightly underestimated their SOL. With OSA alone, patients interestingly exhibited TST overestimation. However, in those with comorbid insomnia and OSA, we saw the most underestimation of sleep time and overestimation of SOL, suggesting some nuanced relationships and worsened self-reported sleep with comorbid disease. Notably, insomnia, with or without OSA, had a significantly negative impact on QOL in both the physical and mental components.

Although insomnia was associated with significantly lower sTST and significantly longer sSOL, the differences were not significant in the corresponding objective parameters (oTST or oSOL). These findings are in line with previously reported results.²⁶ A new finding is that the IS + OSA group showed the most significant underestimation of TST and overestimation of SOL. Underestimation of sleep duration is important because it carries an overall negative effect psychologically. Interestingly, we found the greatest level of TST overestimation in the OSA group, which may result from the higher sleepiness level, lack of deep sleep, sleep fragmentation, and exertions of breathing during the night.^27,28 This finding, coupled with the slight underestimation of TST in the IS group, suggests that insomnia and OSA pull the subjective-objective sleep discrepancy in opposite directions. Similarly, the OSA group tended to underestimate SOL, whereas the IS group tended to overestimate SOL. The greatest level of SOL overestimation was observed in the IS + OSA group, which correlated with the reported increased sleep difficulties in the comorbid insomnia and OSA population. Furthermore, we also found discrepancy in normal sleepers, with an overestimation of TST. Because self-described sleep quality is a core evaluation in clinical settings, clinicians need to be aware of the potential subjective-objective discrepancy patterns in different patient populations. Complicating the interpretation of data within this field is the recognition that recall of sleep in self-reports can be biased.¹⁶ Both the underestimation and overestimation of sleep variables should be well acknowledged for home sleep apnea testing that uses self-reported sleep measures.

Years ago, sleep discrepancy was explained by a “prodromic or transitional state of sleep dysfunction” between normal sleep and objectively reported insomnia.²⁹ Recently, new findings suggest that the overestimation of SOL may be associated with altered brain activity during the process of sleep onset and sleep.³⁰ Different hypotheses have been proposed to explain the etiology/mechanism, including different personality traits,^29,31 cognitive and psychological factors (eg, depression,^32,33 anxiety,³⁴ bipolar disorder,³⁵ memory³⁶), sleep macrostructure^18,20 (eg, frequent awakenings, abnormal REM sleep, duration or stage of sleep), sleep microstructure (eg, EEG power spectral changes,^37–39 sleep spindles characteristics,⁴⁰ increased arousal,³⁷ cyclic alternating pattern^41,42), and other physiological changes (eg, altered regional glucose metabolism³⁰). In our analysis, although we found significant differences in deep sleep (stages 3 and 4 sleep), REM sleep, and WASO between patients with and without insomnia, we did not find that these or other sleep macrostructure parameters significantly contributed to sleep discrepancy.

Despite the subjective-objective discrepancy in sleep estimation, we believe both objective and self-reported domains of sleep are important. Our results showed that patients with self-reported insomnia, regardless of OSA, experienced a significantly negative impact on QOL in both physical and mental components. Addressing sleep misperception could be uniquely beneficial for long-term patient outcomes. Sleep misperception in insomnia tends toward beliefs of difficulty sleeping and insufficient sleep, and these beliefs may subsequently lead to increased levels of anxiety and false expectations for better sleep, which may actually hinder the propensity to sleep. Chronically, these developments lead to a vicious cycle and a fixated misbelief about sleep, which exacerbates insomnia symptoms.^15,18 Patients with such insomnia are considered challenging to treat in clinical practice.

Existing studies have disadvantages regarding further understanding the subjective-objective sleep discrepancy because of limitations inherent in the 2 main diagnostic approaches of PSG or actigraphy. PSG has multiple sensors and can collect rich, more reliable physiological signals. However, it is usually used for a single night and often generates discomfort and unusual sleep conditions, which may not be representative for the patient. Actigraphy is less intrusive, causes less change to routine sleep, and can be used on multiple nights; however, it does not collect sleep data with high accuracy. To better understand sleep discrepancy, more advanced wearable devices with reliable data collection allowing nonintrusive, multiple-night measurements may make large-scale studies possible to better advance our understanding of this area.

There are limitations to the current analyses. To limit heterogeneity, we excluded patients with use of alcohol, caffeine, nicotine, or other substances before sleep and participants reporting an intermediate level of sleep complaints. We were also not able to comment on the impact of individual sleep habits, bedtime routines, or different sleep environments. In general, we were limited by our use of a publicly available dataset. Future studies might consider techniques that are more advanced. For example, it would be informative to include data from multiple-night measurements of sleep. In addition to the commonly used measurements (TST and SOL), a larger spectrum of objective parameters (eg, sleep efficiency, WASO, arousals, and percentage of time spent in a particular sleep stage), self-report parameters (eg, subjective perception of sleep quality), and more sophisticated PSG-based sleep analyses (eg, fine markers of sleep instability/discontinuity) would also be informative. To further inform neurophysiological mechanisms, future analyses should consider brain dynamic changes (eg, quantitative EEG analyses, EEG patterns during presleep wakefulness stage using spectral analysis or complexity analysis) or other physiological signals that may inform sleep quality/patterns (eg, heart rate variability and the autonomic nervous system). In addition, the clinical significance and implications of high night-to-night variability of sleep discrepancy and the role of prescribed opioid medications in sleep perception are also areas of consideration.

CONCLUSIONS

Subjective-objective sleep discrepancy exists in patients with insomnia, OSA, comorbid sleep disorder, and even in people with normal sleep, with both underestimation and overestimation of important sleep variables. Insomnia complaints seem to be the strongest correlate of sleep discrepancy, although comorbid insomnia and OSA may additionally exacerbate the discrepancy. Self-reported insomnia has a significantly negative impact on QOL in both physical and mental components.

DISCLOSURE STATEMENT

All authors have approved the final version of the manuscript. This study was funded by the National Institutes of Health (NCCIH T32AT000051, K24AT009465 and NHLBI 5T32HL007901-22) and was conducted with support from Harvard Catalyst, The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health Award UL 1TR002541). The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University, and its affiliated academic health care centers, or the National Institutes of Health. The authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

EEG

electroencephalogram

ESS

Epworth Sleepiness Scale

IS

insomnia (group)

MCS

mental health component summary (of SF-36)

NSC

normal sleep control patients

OSA

obstructive sleep apnea

oSOL

objectively measured sleep onset latency

oTST

objectively measured total sleep time

PSG

polysomnography

QOL

quality of life

REM

rapid eye movement

SF-36

36-Item Short Form Health Survey

SOL

sleep onset latency

sSOL

self-reported sleep onset latency

sTST

self-reported total sleep time

TST

total sleep time

WASO

wake after sleep onset