Insomnia Severity and Degree of Dysfunction: What Is to Be Learned When These Domains are Discordant?

Abstract

Objective/Background:

Illness severity and resultant dysfunction are often linearly related and tightly coupled (concordant). Some percentage of individuals, however, exhibit discordant associations (high illness severity and low dysfunction [HL] or low illness severity and high dysfunction [LH]). In the present study, a sample of subjects with insomnia complaints were evaluated to determine what percentage of subjects exhibited discordant associations.

Participants:

Archival data were drawn from a community-based sample (n = 4,680; 61.8% female; Ages 18–105).

Methods:

Median splits were calculated for illness severity and daytime dysfunction and each individual was typed as High (H) or Low (L) for the concordant (HH and LL) and discordant domains (HL and LH).

Results:

Given this typology, 61% were classified as concordant and 39% were classified as discordant. Of these, 38% were sub-typed as HH, 23% as LL, 26% as LH, and 13% as HL.

Conclusions:

We propose that some of the discordance may be ascribable to a mismatch between sleep need and sleep ability. Those “who need a lot, may suffer a lot, in the face of only a little (LH)”, whereas those “who need a little, may suffer only a little, in the face of a lot (HL)”.

Introduction

In clinical practice and research, it is nearly axiomatic that illness positivity and/or severity and resultant dysfunction be linearly related and tightly coupled (concordant). This said, it is also commonly observed that some percentage of individuals exhibit discordant associations between pathophysiologic measures (and/or reported or observed symptoms) and patient-reported consequences. Such observations have been made for a variety of disorders such as chronic pain, cardiovascular disease, infectious disease, and insomnia disorder (Fichten et al., 1995; Greco et al., 2015; Mizumoto et al., 2020; Pan & Jones, 2018). Taking into account such discordances may provide clues regarding the factors that moderate or mediate disease expression and/or individual suffering.

In the context of sleep, there are instances where the sleep continuity disturbance (i.e., insomnia) and the related consequences do not align. The concept that individuals can experience poor sleep but not experience distress, dissatisfaction, or dysfunction was first described by Fichten and colleagues in 1995. In this study, 396 subjects between the ages of 55 and 88 completed a variety of sleep measures (sleep history, sleep behaviors, sleep distress, and daytime sleepiness), psychological adjustment (arousal, anxiety, worry, and life satisfaction), and life activities (daytime activities and life events). The volunteers were then placed into one of three categories based on their reported sleep distress (i.e., the extent to which subjects were bothered by sleep disruptions): high distress poor sleepers, low distress poor sleepers, and good sleepers. Given these categorizations, high distress poor sleepers (concordant) were 23% of the sample and low distress poor sleepers (discordant) were 25% of the sample. While these findings do not highlight all four possible dimensions, they do highlight an unexpectedly high rate of discordance.

In 2017, Lichstein more comprehensively addressed the notion of an uncoupling between insomnia severity and the individual’s degree of complaint. In this paper, he reviews the aforementioned study, and five subsequent studies that adopt a 2 × 2 approach to the categorization of individuals with Insomnia Disorder (Lichstein, 2017). In this review, he presents average data from two studies for each dimension over 4 domains (Lichstein et al., 2003; McCrae et al., 2005). One dimension (referred to as poor sleep) is the percentage of subjects that meet and do not meet the sleep continuity definition for insomnia (≥31 min for sleep latency or wake after sleep onset for 3 or more days per week for 6 or more months). The other dimension (referred to as Insomnia Identity) is the percentage of subjects that endorse having insomnia (“claiming of” insomnia). In this study, the sample was found to be comprised largely of “coupled” as opposed to “uncoupled” individuals (73% and 28%). That is, 73% of the subjects were complaining poor sleepers (Insomnia) or noncomplaining good sleepers (normal sleep) (coupled) and 28% of the subjects were noncomplaining poor sleepers or complaining good sleepers (uncoupled). Of those that were typed as uncoupled, 17% were non-complaining poor sleepers and 11% were complaining good sleepers. Lichstein’s focus regarding these data was on the “complaining good sleepers” (i.e., individuals who present a discordance between their “good” sleep status and report of insomnia) which is one of two types of individuals that may present with “insomnia identity” (the other being complaining poor sleepers). In the case of those that exhibit discordance, Lichstein suggested that this form of insomnia identity, or “the conviction that one has insomnia” may be related to skewed sleep expectations, non-restorative sleep, and/or personality “bias” (Lichstein, 2017, p. 1). Skewed sleep expectations refer to what the individual believes to be normal and good sleep (e.g., the belief that most people fall asleep in 2 minutes, therefore a 15-min sleep latency is severe sleep onset insomnia). Non-restorative sleep refers to the phenomenon of objective sleep that is perceived as sleep by the individual but, upon awakening, is experienced as if one had not slept at all (Stone et al., 2008). Finally, there is the issue of personality. Lichstein suggests that insomnia identity may be related to a negative perceptual bias and/or the tendency toward negative reporting.

Both Fichten and Lichstein’s work serves to highlight the value of studying the concordance and discordance between symptom severity, perceived/reported distress, and/or the patient’s identification with their disorder/disease (i.e., I have or don’t have insomnia). The present exploratory study revisits this issue in a large data set with the aims of quantifying: 1) what percent of individuals with sleep complaints exhibit concordances and discordances with respect to illness severity and daytime dysfunction (impairment) and is the concordant dimension larger than the discordant dimension; 2) what percent of subjects that are categorized as discordant fall into the “high-severity and low-dysfunction” (HL) and “low-severity and high-dysfunction” (LH) domains and are these higher or lower than expected; 3) are the discordant domains (LH and HL) equally populated; and 4) do the identified groups vary by demographics. It was hypothesized that the majority of subjects would exhibit concordant profiles and that the small percentage of individuals who exhibited discordant profiles would be equally distributed to (HL) and (LH) domains. Finally, each domain was profiled with respect to demographics, shiftwork, anxiety, depression, and chronic pain so as to be mindful of what other ways the groups may differ and how these differences may or may not serve as potential mediator/moderators.

Methods

Data source

The data for the present study were drawn from an archive comprised screening surveys. Individuals responding to insomnia research advertisements were directed to an on-line website to determine their eligibility for ongoing research (www.sleeplessinphilly.com). The website had several pages (a welcome page, a “why participate in research” page, and two consent forms). Consent was required to access the survey page. Data from 2011 to 2019 were used for the present study.

Subjects

Recruitment efforts targeted individuals with sleep continuity disturbance (i.e., insomnia). A stable sleep pattern and/or a diagnosis of insomnia disorder was not a requirement for participation in the survey. The following inclusion criteria were used for the present archival data review: age 18 or older, consented for their submissions to be used as part of research, submitted a completed survey, and lived in Philadelphia or its surrounding suburbs. Subjects were not excluded from the dataset for any comorbid sleep disorder, medical condition, psychological condition, or substance use.

Measures

Individual data were collected at one time point. The survey was comprised of 103 questions and was intended to: quantify the individuals’ level of sleep continuity disturbance (severity, frequency, & chronicity), to assess for known sleep disorders, medical and psychiatric illness, to document medication use, and to collect demographic data (age, sex, race and ethnicity, education level, occupation, typical hours at work, etc.). Of age, subjects were categorized as young adult (18–29 yrs.), adult (30–44 yrs.), middle age (45–64 yrs.), or older adult (65–105 yrs.).

Sleep continuity disturbance items consisted of questions that asked about the length of time it takes the subject to fall asleep (i.e., sleep latency [SL]), time awake in the middle of the night (i.e., wake after sleep onset [WASO]), and time awake earlier than intended (i.e., early morning awakenings [EMA]). See Table 1 for the specific sleep continuity disturbance questions. For the present study, a global metric of sleep continuity disturbance (total wake time [TWT]) was calculated by summing (in minutes) an individual’s self-reported sleep latency, wake after sleep onset, and early morning awakenings [i.e., TWT = SL + WASO + EMA]. This metric has been used in prior analyses (Carney et al., 2017; Ong et al., 2014; Ravyts et al., 2019) and was used in preference to sleep efficiency (SE%) as it provides a measure of absolute illness severity (total amount of nocturnal wakefulness).

Table 1.

Relevant questions of sleep continuity disturbance and sleep related daytime dysfunction.

Questions of Sleep Continuity Disturbance
1. On a typical night, how many minutes does it take you to fall asleep? _______
2. On a typical night, how long are you awake altogether during awakenings? _______
3. Do you sometimes wake up before you intend to in the morning (circle one)? Yes/No
Typically, how many minutes before? ______
Questions of Sleep Related Daytime Dysfunction
1.	Do you have difficulty with attention or concentration during theday?	Yes/No
2.	Are you irritable during theday?	Yes/No
3.	Do you feel rested when you wake up?	Yes/No*
4.	Do you feel your daytime function is impaired due to trouble sleeping?	Yes/No
5	Do you feel fatigued during the day?	Yes/No
6.	Do you feel sleepy during the day?	Yes/No

^*Item was reversed scored, where “yes” was counted as a 0 and “no” counted as 1.

Subjects were also asked about sleep-related daytime impairment (symptoms of dysfunction during the daytime). These symptoms included the incidence of: impaired daytime function due to trouble sleeping; concentration difficulties; irritability; feeling unrested upon waking; and the experience of daytime fatigue and sleepiness. Subjects were asked to indicate “Yes” or “No” for each item (response values of 1 and 0, respectively). Sleep-related daytime dysfunction (SRDD) was quantified as a total score with each “yes” endorsement of a symptom counting as a 1. These items were summed for a total score ranging from 0 to 6. Specifically, subjects who reported no sleep-related daytime dysfunction received a score of 0 and subjects endorsing all six symptoms yielded a score of 6. See Table 1 for specific questions. Note, the question, “Do you feel rested upon waking?” was reverse scored.

Data management and statistical analyses

Median splits were calculated for TWT and sleep-related daytime dysfunction. TWT was totaled and the median length of TWT was determined to be 125 minutes. As such, individuals who experienced a TWT of 125 minutes or greater were categorized as having a HIGH amount of sleep continuity disturbance. Individuals who experienced less than 125 minutes of sleep continuity disturbance were categorized as having LOW sleep continuity disturbance. [Note: please recall the sample is comprised of individuals that are responding to research advertisements about insomnia (problems falling and staying asleep). Thus, “low” is actually relatively low for those that have sleep continuity disturbance]. A median split was also used to determine the high and low classifications of the sleep-related daytime dysfunction (SRDD) scaled score. The median value was 4, perhaps slightly higher than the expected “3” but again, a reflection of the recruited subjects. From this, subjects were considered to have a LOW amount of dysfunction if they scored between 0 and 4 and a HIGH amount of dysfunction if they scored greater than 4.

Based on these two thresholds, individuals could be typed as High severity and High dysfunction (HH); High severity and Low dysfunction (HL); Low severity and High dysfunction (LH); or Low severity and Low dysfunction (LL). Descriptive statistics are presented for each domain and category. A Chi-square analysis was conducted to determine if the distribution of subjects significantly differed from expected values.

In order to assess for group differences between the domains (HH, LL, HL, and LH) with respect to sex, age (years and/or age groups), race, BMI, education, shift work, chronic pain, and current depression and anxiety status as potential third variables, statistical analyses were conducted using one-way ANOVAs, contingency analyses, and Chi-Square tests.

Results

Subjects (n = 4,680) were between the ages of 18 and 105 years (${\overline{x}}_{age}=40.0$; SD = 15.1), and included primarily women (61.8%), who were white (63.5%), and comprised of about 35% young adults, 27% adults, 32% middle-aged adults, and 7% older adults. See Table 2.

Table 2.

Demographics.

Demographics (n = 4,680)
		N	%
Sex
	Female	2982	61.8%
	Male	1788	38.2%
Age Group
	Young Adult (18–29 years)	1636	35.0%
	Adult (30–44 years)	1251	26.7%
	Middle Age Adult (45–64 years)	1480	31.6%
	Older Adult (65–105 years)	313	6.7%
Race
	White	2974	63.5%
	Black/African American	1377	29.4%
	American Indian/Alaska Native	34	0.7%
	Asian	174	3.7%
	Native Hawaiian or Other Pacific Islander	11	0.2%
	Race not indicated	110	2.4%
Ethnicity
	Hispanic or Latinx	245	5.2%
	Not Hispanic or Latinx	3809	81.4%

The majority of subjects were classified within the concordant category (61%) with about 38% typed as HH and 23% typed as LL. The discordant category was comprised of about 39% of the sample with about 26% LH and 13% typed as HL. See Figure 1.

Figure 1.

Matrix of sleep groups (Chi-Square p < .001).

When assessing groups [HH (high severity, high dysfunction), LL (low severity, low dysfunction), LH (low severity, high dysfunction), HL(high severity, low dysfunction)] for differences with respect to age (years) and BMI, it was found that the HL group was significantly older than the other groups. BMI differences were also detected, but the magnitude was small and not likely to be clinically relevant (BMI differences of ~ 1). Note: It was found that the group variances were not equal but both the adjusted and unadjusted p values were significant. See Table 3.

Table 3.

Analysis of variance results for age and BMI.

	Concordant		Discordant
	HH	LL	HL	LH	p-value*
Age (years)	40.7 (14.2)	37.8 (15.3)	46.1 (16.2)	37.8 (14.5)	< .001
BMI	28.2 (7.4)	27.0 (6.9)	28.1 (6.5)	28.1 (7.3)	< .001

When assessing groups (HH, LL, LH, HL) for differences with respect to sex, race, education, shift work, chronic pain, and current depression and anxiety status, it was found that the HH group was more educated, suffered more chronic pain, and more likely to be currently depressed or anxious. See Table 4.

Table 4.

Group differences.

	Concordant			Discordant
	Total	HH	LL	Total	HL	LH	p-value
Sex							.142
Female	61.2%	39.2%	22.0%	38.8%	13.2%	25.6%
Male	60.3%	36.1%	24.2%	39.7%	13.4%	26.3%
Race							< .001
Black/African American	63.3%	40.3%	23.0%	36.7%	18.3%	18.4%
White	59.3%	37.1%	22.2%	40.7%	11.4%	29.3%
Hispanic or Latinx	60.8%	35.9%	24.9%	39.2%	11.0%	28.2%	.149
College Degree or Higher	60.0%	34.2%	25.8%	40.0%	12.1%	27.9%	< .001
Rotating Shifts	61.4%	42.5%	18.9%	38.6%	14.2%	24.4%	.035
Chronic Pain	61.9%	45.1%	16.8%	38.1%	12.5%	25.6%	< .001
Psych Dx							< .001
Depression (current)	65.9%	47.2%	18.7%	34.1%	12.9%	21.2%
Anxiety (current)	65.0%	46.5%	18.5%	35%	12.5%	22.5%

Discussion

There are numerous examples throughout the clinical research literature documenting the phenomenon where illness positivity and/or severity is discordant from self-reported suffering (i.e., consequences/symptoms of illness). Within behavioral sleep medicine, several investigators have documented this phenomenon with respect to insomnia (Fichten et al., 1995; Lichstein et al., 2003; McCrae et al., 2005). The present study sought to revisit this issue in a large community-based sample of individuals who endorsed sleep difficulties (n = 4,680). Median splits were calculated for illness severity and daytime dysfunction and each individual was typed as High or Low for each domain. Thus, each individual and the groups themselves were empirically defined based on the current dataset (i.e., HH, HL, LH, LL). The questions for the present study were: 1) is the concordant dimension more populated than the discordant dimension (i.e., percent of subjects HH & LL > HL & LH); 2) are any of the cells populated at a percentage that was higher or lower than expected (i.e., null hypothesis would be 25% for each of the four cells); 3) is the discordant domain equally populated for LH and HL; and 4) are the defined groups different with respect to demographic measures. It was found the cells were not evenly populated. As would be expected, more subjects were found to be concordant than discordant (61% vs. 39%). The highest cell frequency was for HH (38%) and the lowest was for HL (13%). With respect to group demographics, HL group was older than the other groups and the HH group was more educated, suffered more chronic pain, and had higher BMI, depression, and anxiety scores.

These findings differ from Lichstein and colleagues’ observations (Lichstein, 2017; Lichstein et al., 2003), with nearly double the representation into the discordant dimension. This is likely the case because of conceptual/methodological differences with respect to the two dimensions evaluated. Lichstein framed these dimensions in terms of 1) “Poor Sleep” (assessed with sleep diaries) and 2) “Sleep Complaint” (assessed by whether the subject self-identified as having insomnia or not). While the first dimension is conceptually similar across studies, the measurement of illness severity differs. For example, Lichstein uses sleep diary data while the present study used retrospective self-report data (survey items). Lichstein focused on SL and WASO data to determine if someone was a poor sleeper (≥ 31 min for SL or WASO for 3 or more days per week for 6 or more months). The present study utilized a composite variable for nocturnal wakefulness (TWT [SL+WASO+EMA]). Regarding the second dimension, Lichstein focused on sleep complaint as defined by whether the subject self-identifies as having insomnia while the present study focuses on the sum of multiple single-item measures of daytime dysfunction. Lastly, Lichstein’s sample was comprised of good and poor sleepers while the present study was populated only with individuals with sleep difficulties. Given these (and other) differences, it is not surprising to find that the observed concordance and discordance rates were different. In the final analyses, it may be most productive to take into account all three constructs. For example, how well does illness severity and the intensity of daytime consequences predict “insomnia complaint”.

Lichstein (2017) offers a comprehensive set of reasons for the observed discordance but primarily focuses on insomnia identity as the explanation for “complaining good sleepers” (similar to the present study’s “Low severity, High dysfunction”; LH). In the present study, the observed discordances may be explained by a mismatch between sleep ability, need, and opportunity. For example, individuals who endorse low severity and high consequence (LH) may have a greater sleep need, a need that is unmet owing to reduced sleep ability (insomnia disorder) and/or reduced sleep opportunity (insufficient sleep disorder). It is also likely that individuals with insomnia identity comprise a portion of this group, not because of unmet sleep need, but simply owing to factors that cause them to identify their daytime dysfunction as the result of sleep disturbance as opposed to other possible attributions. Conversely, individuals who endorse high severity and low consequence (HL) may simply have a lesser sleep need. That is, while HL individuals may have poor sleep ability, they nonetheless meet their sleep need given their opportunity. This possibility seems to be likely for older adults and in our sample as the HL group was found to be older. This said, age may be a proxy measure for increased cumulative morbidity and that this (in addition to sleep need) may also moderate the association between sleep ability and daytime dysfunction. Interestingly, none of the “profile” factors assessed were found to be significantly higher or lower in the LH group, suggesting that the observed discordance was not related to common confounds (3^rd variables) like age, sex, or race (or factors associated with these variables).

The conceptualization of discordance being related to “sleep need”, while embracing a factor that some consider unmeasurable, has the value of being parsimonious. This said, the other factors that were proffered as underlying insomnia identity (skewed sleep expectations, non-restorative sleep, negative perceptual bias and/or the tendency toward negative reporting) are also likely to be contributory. It is our perspective that a substantial percentage of the variance for the whole of the discordant dimension is ascribable to sleep need. Put simply, those who need a lot, may suffer a lot, in the face of only a little (LH), whereas those who need a little, may suffer only a little, in the face of a lot (HL).

Strengths & limitations

The primary strength of the study is also its primary limitation: it is a community survey study. The strength is, as a survey, it was possible to acquire a large and diverse sample. The weakness is, as a survey, the assessment was: entirely retrospective (i.e., subjective measures of severity and dysfunction include the limitation and possibility of recall bias); based on single items; acquired from a single geographic region; and (as with all research), based on only those that volunteered. Another potential limitation is the use of single-item measure of illness severity; moreover, one that is less common than others. Total Wake Time (TWT) was chosen as our measure of absolute illness severity. This measure was used instead of the common single measure of Sleep Efficiency (SE%) because it more completely captures the absolute magnitude of illness severity in terms of wakefulness during the nocturnal period (i.e., two individuals may have the same SE% [e.g., 70%] but vastly different TWTs.). It also should be noted that there is a reasonable precedent for the use of this variable (Carney et al., 2017; Ong et al., 2014; Ravyts et al., 2019). Finally, the cutoff for illness severity, while empirically determined based on the study cohort, are skewed toward high illness severity, leaving values <125 TWT to be classified as “low” sleep continuity disturbance. While “true” for our sample, such values likely fall within the mild and moderate ranges of illness for other samples.

Future directions (The need to take into account "sleep need")

We, as have others, are suggesting that the evaluation of discordance may be as informative as the evaluation of concordance, though it is usually the latter that is the subject of empirical study. The evaluation of discordance has been productive in the present case because it allowed for the generation of a “novel” idea about how such a phenomenon is possible. The idea being that by taking into account sleep need, one might explain how small amounts of sleep continuity disturbance result in substantial levels of daytime dysfunction, and vice versa (large amounts of sleep continuity disturbance result in minor levels of daytime dysfunction). The concept of sleep need is neither new nor novel, but calling for it to be accounted for in terms of the conceptualization of illness status is relatively new and novel. That is, while sleep opportunity and sleep ability are regularly conceptualized and operationalized in terms of average time allotted for sleep (i.e., sleep opportunity = time in bed [TIB]) and in terms of average total sleep time (sleep ability = average total sleep time [TST]) (Fallone et al., 2005; Oda et al., 2019; Perlis et al., 2006; Spielman et al., 1987) quantifying sleep need is not generally a focus of conceptualization, assessment, and/or treatment. This may be, not because of a lack of relevance, but because the construct has long been considered ineffable and unmeasurable. Perhaps what is needed is to frame “sleep need” in terms of “need for what”. Sleep is thought to sub-serve many functions ranging from the simple-and-obvious (sleep so that one is not sleepy) (Broman et al., 1996) to the everyday (enhancing cognitive and physical performance) (Van Dongen et al., 2003) to the less-simple-and-obvious (response to injury and infection, the regulation of metabolic activity, CNS waste clearance, etc.) (Grandner et al., 2015; Ibarra-Coronado et al., 2015; Irwin, 2019). These complexities not without standing, perhaps the first best way is the first way; define sleep need as its inverse analog: the intensity of daytime sleepiness and/or fatigue. An example of such an approach is embodied in Spielman’s Sleep Need Questionnaire. Ultimately, what may be best is to use this construct and/or measure to determine if it can (alone or in combination with measures of cumulative morbidity, non-restorative sleep [e.g., sleep fragmentation and/or alpha sleep], skewed sleep expectations, negative perceptual bias and/or the tendency toward negative reporting) predict the discordant categories of HL and LH and/or Lichstein’s category re: insomnia identity. Such an analysis may be particularly productive for older adults as they appear to more regularly exhibit High illness severity and Low daytime dysfunction (HL status) (Grandner et al., 2012).

Conclusion

While the present paper has been focused on sleep continuity disturbance (insomnia), the issues that lie at the heart of this paper are transdiagnostic, that is, regardless of which sleep disorder, we may learn as much from when our measures of cause and consequence are discordant as when they are concordant. For example, it may be productive to undertake a similar investigation to the present in patients with OSA. If similar discordances are found, it would be informative to evaluate how, when, and in whom higher AHI are associated with relatively lower measures of sleepiness. More, such studies may extend well beyond sleep medicine to medical and behavioral health in general and serve to teach us much about human suffering and resilience.