INSOMNIA

Symptom Reports in Severe Chronic Insomnia Douglas E. Moul MD, MPH, Eric A. Nofzinger MD, Paul A. Pilkonis PhD, Patricia R. Houck MSH, Jean M. Miewald BA, and Daniel J. Buysse MD From the Department of Psychiatry, University of Pittsburgh School of Medicine, Western Psychiatric Institute and Clinic, Pittsburgh, Pennsylvania

Study Objectives: To describe patterns and severities of the daytime and nighttime symptoms of chronic insomnia patients Design: Exploratory chart review from clinicians’ evaluation summaries, a self-report screening instrument, the Pittsburgh Sleep Quality Index, the Beck Depression Inventory, the Epworth Sleepiness Scale, and the Hopkins Symptom Checklist-90 (HSCL90). Setting: A regional sleep disorders referral clinic Patients or Participants: 94 patients with chronic insomnia (DSM-IV code 307.42), classified into the subgroups “Primary Insomnia,” “Depression-Related,” “Anxiety-Related,” and “Other” Interventions: N/A Measurements and Results: Frequent symptoms occurred not only in nocturnal domains (e.g., sleep disturbances, environmental sensitivity), but also in daytime domains (e.g., cognitive difficulties, sleepiness). Compared to primary insomnia patients, those with depression-related insomnia endorsed more severe symptoms. All subgroups endorsed a

generally similar symptom profile when single symptoms were considered in isolation. When considered conjointly, severe symptoms typical of depression and generalized social alienation had a high negative predictive value for primary insomnia. The number of severe symptoms on the HSCL90 was related to fewer sleep hours in the nonprimary insomnia subgroup but not in the primary insomnia subgroup. Conclusions: Patients with chronic insomnia report significant daytime as well as nighttime symptoms. Depression-related and primary insomnias were separable only by some highly characteristic symptoms of depression. Diagnostic subgroups of insomnia patients may vary in how their overall distress relates to diminished self-reported sleep. Nighttime and daytime symptoms need to be assessed together when measuring insomnia severity. Key words: Primary insomnia; depression; symptoms; group differences; exploratory analysis; diagnosis; daytime

linked to the number of hours of sleep than to the minutes of sleep latency.15-17 Since sleep interferes with memory consolidation,18 self-reports about sleep are prone to inaccuracies, compared to polysomnographic measures of sleep.19-29 Other sources of distress (e.g., pain, depression) may also affect the accuracy of self-reports about sleep.30 Despite the inaccuracies in selfreports, polysomnographic studies have demonstrated that insomnia patients get less, poorer quality, and more inconsistent sleep compared to normal sleepers.20,27,30-33 Poor sleep is a reason for insomnia complaints, even if it may not be the ultimate cause.34 Whatever the relationships between subjective and objective sleep descriptions, clinicians are asked to provide treatments for chronic insomnia as it affects nocturnal and daytime functioning and quality of life. Daytime symptoms are associated with insomnia, and are frequently reported by insomnia patients.8,15,25,30,35-52 These include fatigue, decreased alertness, being unrefreshed, sleepiness, inability to nap, irritability, tension, hyperarousal, depressed mood, impaired memory functioning, decreased ability to concentrate, social aversion, anergia, disturbed cognitions about sleep, and disabilities in work and social life. Since mere dissatisfaction with sleep appears not to be a sufficient cause for seeking medical care, these daytime symptoms are likely to be salient reasons for seeking care. With this heterogeneity of symptoms, it may be that insomnia patients are heterogeneous in their symptom profiles and goals for treatment, as suggested by prior analytic studies in naturalistic samples.1,4,5,38,40,52-60 This heterogeneity is assumed both by the DSM-IV35 and ICSD36 diagnostic systems, where depression-related and anxiety-related causes of insomnia are distinguished from primary insomnia and psychophysiologic insomnia, respectively. Distinctions between insomnia diagnoses are based upon beliefs about the relative presence of symptoms and their temporal course. Thus, a com-

INTRODUCTION INSOMNIA SUFFERERS SEEK TREATMENTS FOR INSOMNIA MAINLY BECAUSE OF PERCEIVED DISTRESS OR IMPAIRMENT RATHER THAN HOW MUCH SLEEP THEY GET.1 Many people have “inadequate” sleep but do not seek care.2-4 Being dissatisfied with one’s sleep is not the same as having a sleep complaint that might occasion actual care seeking.4,5 Insomnia complaints are indeed associated with increased medical utilization.6-9 Since insomnia is quite prevalent,2,8,10-12 and may be causally linked to other disorders,8,13,14 further investigation into the nature of insomnia complaints is likely to be important for improving medical care and diagnostic specificity. Because of their potential impact on treatment costs, clinically referred chronic-insomnia patients are an especially important population to study. Detailed descriptions of their symptoms are essential in understanding the factors related to their medical utilization. For clinicians addressing an insomnia complaint, symptom queries follow several lines, often not integrated into a common metric or even into convenient symptom clusters. Sleep symptoms such as the self-reported number of hours of sleep are assessed in patients presenting with sleep complaints. Among distressed poor sleepers, the degree of distress may be more Disclosure Statement NIH Grants AG00972, AG15138, MH30915, MH24652, and MH16804. Submitted for publication August 2001 Accepted for publication March 2002 Address correspondence to: Douglas E. Moul, M.D., M.P.H., Sleep and Chronobiology Center, Western Psychiatric Institute and Clinic, 3811 O’Hara Street, Pittsburgh, PA 15213; Tel: 412.624.5281; Fax: 412.624.2841; E-mail: [email protected] Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

548

Symptom Reports in Severe Chronic Insomnia—Moul et al

prehensive, integrated description of patients’ nighttime and daytime symptom patterns may help to advance our understanding of insomnia complaints generally, and possibly of diverse subtypes more specifically. Describing how symptoms cluster together may assist clinicians in organizing productive lines of clinical inquiry. Such a descriptive effort would also provide an information source for constructing better insomnia rating scales that integrate the various symptom domains queried in an assessment of severity. This chart review of severe chronic insomnia patients was conducted with existing diagnostic categories35,36 that were used in delivering clinical care to these patients with a presenting chief complaint of chronic insomnia. We sought in particular to determine whether primary insomnia could be distinguished symptomatically from other insomnia-related diagnoses. For improving clinical practice, we believe there is a need to review the symptoms of clinical populations and to do so with minimal theoretical presumptions. The specific aims of the study were: 1) to describe the symptoms and symptom patterns that evaluating clinicians reported in their clinical summaries; 2) To describe the types and patterns of patient-endorsed symptoms on a sleep disorders screening questionnaire; 3) to describe the types and patterns of symptoms endorsed on several self-report questionnaires relevant to evaluating insomnia patients; and 4) To explore relationships in the observed data that may assist in distinguishing diagnostic subtypes or assist in understanding patients’ distress.

Patients completed a self-report sleep history questionnaire, called the Survey of Sleep (SOS) (available upon request), prior to clinical evaluation. The SOS helps to assess a variety of sleep disorders, symptoms, and impairments. All items that could be considered as symptoms or impairments were included in this analysis. Symptom frequencies and percentages were computed. Patients also completed several standard, well-validated selfreport questionnaires that focus on symptoms that may accompany a sleep complaint. These included the Pittsburgh Sleep Quality Index (PSQI),37 the Epworth Sleepiness Scale (ESS),61 an augmented version of the Beck Depression Inventory Version IA (BDI),62 and the Hopkins Symptom Checklist-90 (HSCL90).63 The PSQI evaluates sleep quality, with a global score greater than 5 reflecting poor sleep quality. The ESS measures subjective sleepiness, posed as scaled questions about dozing in various situations. An ESS score greater than 9 is considered potentially pathologic.61 The BDI is particularly relevant in assessing insomnia patients because depression is well known to affect sleep. For local clinical work, additional items were added to the BDI Version IA for weight gain (Item 22, scaled 0-2), increased sleep (Item 23, scaled 0-2), and daytime drowsiness (Item 24, scaled 0-3). The HSCL90 is a general screening instrument for assessing a broad spectrum of psychiatric symptoms, including depression, anxiety, and paranoia, among others.

METHODS

Treatment of Variables

This study was approved by the University of Pittsburgh Institutional Review Board.

The PSQI, ESS, and HSCL90 scoring was computed as previously described.37,61,63 Scores from the BDI were also computed in the same manner as Version IA of the BDI62 scoring but without the sleep-related items 16, 23, and 24. This score was called “BeckNS.” For analysis of potentially discriminating HSCL90 items, items endorsed as “Quite a bit” or “Extremely” bothersome were classified as severe. This cutpoint was selected to represent the distinction between a sleep dissatisfaction and a frank complaint. The total number of severe symptoms was log-transformed for use in regression models. The HSCL90 items selected as potential discriminators between the PI and DR subgroups were required to have an odds ratio of 8 or greater and a 15% or more difference (DR > PI) in severe symptom endorsement. These criteria were chosen after an inspection of symptom response distributions suggested that they may be diagnostically informative. However, to broaden the scope of the comparison further, the non-PI groups were collapsed together for this comparison. The selected items may also be clinically important in making diagnostic discriminations between PI and non-PI subgroups.

Questionnaires

Clinical Evaluations The clinical sample of adult (age > 19 years) patients was drawn from a computerized list of patients who had been clinically referred for evaluation of a problem with sleep or wakefulness. All patients’ diagnoses had been computerized using the DSM-IV. Those 94 patients given the code “307.42” between 1997 and 2000 were identified by computer search and included in the analysis. Due to the way the DSM-IV is constructed, this “307.42” code includes both primary insomnia as well as insomnias related to other Axis I or Axis II disorders. With review of the typed clinical summaries, primary insomnia (PI) patients were distinguished from those with depression-related (DR), anxiety-related (AR), and other (O) patients. Age, sex, evaluating clinicians, and final diagnoses were obtained for each patient. Demographic and other sample characteristics are shown in Table 1. To explore clinicians’ described symptomatologies, the dictated summaries were individually reviewed by the first author. As a new symptom was mentioned that could be considered distinct and possibly generalizable across patients, a new symptom category was created and evaluated in subsequent dictations. In order to avoid making the analysis completely individualized, some words were considered to be clinical synonyms (e.g., “tired” and “fatigued”). No semantic theory was consciously imposed upon the data. Fifty-six symptoms resulted from these procedures. Symptom frequency and percentages were computed on the completed data.

Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

Statistical Methods The general statistical approach was exploratory and in general utilized observational study techniques. Statistical analysis was done with S-PLUS 2000 (MathSoft, Inc., Seattle, WA). Fisher’s exact test was used to compare categorical data from two-dimensional tables. For stratified 2 x C exact tests, the exact permutation test from StatXact 4 for Windows (Cytel Software Corporation, Cambridge, MA) was used. Group differences in continuous measures were explored with Welch modified two549

Symptom Reports in Severe Chronic Insomnia—Moul et al

Table 1—Sample Characteristics Number (% completed) or Mean (SD) Demographics (N = 94) Females Age

54 (57%) 46.4 (13.8)

Clinician Evaluation Summary Data (N = 94) Primary Insomnia (PI) Depression-Related (DR) Anxiety- Related (AR) Other (O) Total Symptoms Mentioned

50 (53%) 31 (33%) 6 (6%) 7 (7%) 12.8 (5.1)

Survey of Sleep (SOS) Self-Report Available (N = 83) Sleep problem present for more than 1 year "Severe Problem" with sleep Use a sleeping pill nightly Sleep problems occur nightly Take daily naps Total symptom endorsement count Initial sleep latency (minutes) Number of awakenings Sleep latency after awakenings (minutes) Time in Bed (hours) Hours of sleep Sleep efficiency Pittsburgh Sleep Quality Index (PSQI) Completed (N = 70) Total Score > 5 Total Score Beck Depression Inventory (BDI)a Completed (N = 79) Total BeckNSb Score Epworth Sleepiness Scale (ESS) Completed (N = 79) Total Score > 9 Total score Hopkins Symptom Checklist 90 (HSCL90) Completed (N = 79) Number of Severe Endorsementsc

66 (83%) 65 (81%) 72 (89%) 64 (79%) 9 (11%) 20.5 (8.3) 53.3 (51.8) 2.7 (1.7) 56 (64.7) 8.2 (1.9) 5 (1.7) 64% (22.5)

53 (100%) 13.1 (3.4)

12.5 (9.0)

16 (20%) 5.6 (4.5)

11.9 (15.3)

aAugmented cThose

24-item Beck Depression Inventory (see Methods); bAugmented Beck Depression Inventory without sleep-related items; symptoms endorsed as "Quite a bit" or "Extremely" bothersome

sample t-tests and one-way and two-way (group by gender) ANOVAs. When distributions were not normal, Kruskal-Wallis and Wilcoxon rank-sum tests were used. Exploratory factor analysis was conducted on the clinicianreported symptoms mentioned in more than 10% of the patients. For the SOS symptoms, the 40 most-frequent symptoms were used (which corresponded to an 18% or greater rate of endorsement). In both cases binomial data for symptoms (i.e., present versus absent) were used. These cut-offs were adopted based upon compromises between the availability of responses and the need for parsimony. These analyses were done using the principal axis factoring method with varimax rotation and an entry loading set for individual items at 0.4. Five factors were selected as a compromise between inclusiveness and parsimony. A receiver-operator-characteristic analysis was conducted using the selected HSCL90 items (see Table 5) to differentiate the Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

PI from the non-PI patients. In addition, discriminant analysis of the non-PI score variables was conducted, to cross-check the performance of the variables. Multivariate linear least-squares regressions were individually developed in a progressive fashion to investigate relationships between the number of severe HSCL90 symptoms (log transformed) versus other continuous (e.g., age, subjective sleep latency and number of hours of sleep) and binomial (e.g., gender, PI vs non-PI grouping) variables. Variables identified as significant in inspections of simpler regressions were included in regressions with more covariates. Pearson correlations were computed selectively and tested with two-sided t-tests when tested against the null hypothesis.64 Statistical significance was set at less than or equal to 0.05. All statistical tests were interpreted as exploratory rather than probative,65 so corrections for multiple comparisons were not 550

Symptom Reports in Severe Chronic Insomnia—Moul et al

Table 2—Most Frequent Clinician-reported Symptoms

Symptom or Problem

Primary Insomnia) (PI) N = 50

Diagnostic Subgroups Depression Anxiety Related Related (AR) (DR) N = 31 N=6

Other (O)

Total

N=7

N = 94

NIGHTTIME Long sleep latency Numerous awakenings Difficulty returning to sleep if awakened Uses a sleeping pill frequently Insufficient sleep hours Nighttime restlessness Awakens too early Light sleep Mind can't stop at bedtime Life event makes insomnia worse Frustration with sleep Annoyed by environmental stimuli

68% 64% 62% 48% 42% 36% 32% 36% 28% 30% 30% 32%

55% 48% 45% 39% 45% 45% 26% 26% 39% 39% 32% 23%

83% 50% 67% 83% 17% 33% 67% 33% 33% 33% 33% 33%

71% 86% 57% 29% 57% 71% 71% 29% 14% 0% 14% 43%

65% 60% 56% 46% 43% 41% 35% 32% 31% 31% 30% 30%

72% 38% 48% 40% 40% 38% 26% 24% 26% 30% 32% 34%

55% 65% 55% 52% 42% 35% 48% 45% 45% 39% 29% 35%

67% 83% 50% 50% 50% 33% 33% 83% 50% 17% 0% 17%

57% 86% 14% 43% 57% 71% 57% 29% 43% 14% 57% 0%

65% 53% 48% 45% 43% 39% 36% 35% 35% 31% 31% 31%

DAYTIME Daytime bodily fatigue Depressiona Difficulty concentrating General worrying Loss of energy Irritability Feels tense General anxiety Low motivation Daytime sleepiness Loss of productivity Napping during daytime aDR

> PI significant by Fisher's exact test (p = 0.02)

employed.

adjusted for diagnostic distributions with stratified exact tests, appeared to support a bias being present only for the symptom of heightened daytime alertness (p = 0.03). In reviewing symptoms, it was evident that the AR and O groups were too few to allow even for exploratory comparisons. However, symptom-wise comparisons were made for all symptoms between the PI and DR groups. Only depressed mood or affect was statistically supported as different between the PI and DR subgroups using Fisher’s exact test (p = 0.02). To illustrate the more common symptoms, clinician-reported symptoms noted in more than 30% of the clinical summaries are shown in Table 2. (This top 70% cut-off was selected only to reduce this table’s complexity and is not related to cut-offs selected for statistical tests.) The exploratory factor analysis of clinician-reported symptoms (see Table 3) produced factors we interpreted as 1) cognitive difficulty and worry (e.g., difficulty concentrating, general worry, poor attention and memory), 2) tension and restless sleep (e.g., feels tense, nighttime restlessness, nightmares), 3) sleep anxiety and frustration (e.g., anxiety, frustration, and worry about sleep), 4) daytime sleepiness (e.g., daytime sleepiness and nap-

RESULTS Clinical features of the entire sample are presented in Table 1. There were no diagnostic subgroup differences in gender (Fisher’s Exact Test, p=0.48) or in age (F=0.64; df=3,90; p=0.59). 1. Clinician-reported Symptoms A one-way ANOVA of symptom count indicated that the seven reporting clinicians varied in the number of symptoms they reported in their clinical summaries (F=2.48; df=6, 87; p=0.03). However, only one clinician had a mean symptom count lower than 9. There were no differences in diagnostic frequency of insomnia subgroups between the senior supervising clinicians who had finalized all diagnoses (Fisher’s, p = 0.10). Clinicians’ total symptom count was not correlated with patient age (r=-.03), gender (t=1.19; df=81; p=0.24), or diagnostic subgrouping (F=1.02; df=3,90; p=0.39). The seven reporting clinicians’ reports provided no evidence for reporting biases, when adjustments for diagnostic distributions were made with stratified exact tests. The reports signed by the supervising clinicians, when Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

551

Symptom Reports in Severe Chronic Insomnia—Moul et al

Table 3—Exploratory Factor Analysis of Clinician-reported Symptoms

Symptom Difficulty concentrating Poor attention Poor memory General worrying Feels tense Nightmares Nighttime restlessness Frustration about sleep Worries about sleep Anxiety about sleep Napping during the daytime Daytime sleepiness Difficulty returning to sleep if awakened Loss of productivity Just awakens for no reason Irritability Numerous awakenings Variance associated with factor †Factor

1

2

Factor† 3

4

0.76 0.57 0.55 0.47 0.56 0.54 0.47 0.64 0.49 0.41 0.63 0.43

6.0%

5.8%

4.4%

4.4%

0.49 0.45 0.43 0.43 0.40 4.0%

h2§ 0.62 0.34 0.35 0.34 0.39 0.39 0.24 0.46 0.29 0.39 0.45 0.26 0.32 0.26 0.24 0.28 0.25

loadings selected if greater than 0.4; §Commonality (h2) variance associated with factor loadings.

Table 4—Most Frequently Endorsed Symptoms from the Survey of Sleepa Diagnostic Subclasses Primary Depression Anxiety Insomnia (PI) Related (DR) Related (AR) Symptom or Problem N = 45 N =28 N=5 NIGHTTIME Difficulty staying asleep 80% 71% 100% Wakes for no particular reason 76% 71% 100% Wakes early and unable to fall back asleep 78% 64% 80% Difficulty falling asleep 73% 68% 60% Anxious, tense, or worrying at bedtime 60% 86% 100% Frequent tossing and turning 73% 79% 20% Need to urinate during the night 56% 64% 80% Physically tense at bedtime 42% 64% 80% Awakened by noises 40% 46% 60% Regular nightly awakenings 40% 54% 0% Large body jerks when falling asleep 38% 50% 40% Feeling too hot or too cold 38% 46% 60% Awakened by bedpartner 33% 54% 60% Snoring 38% 36% 60% Evening restless legs sensations 31% 43% 20% Heartburn or other burning in chest, stomach 38% 25% 40% Awakened by light 24% 36% 80% DAYTIME Interference with ability to function 91% 93% 100% Fatigued feeling during the day 84% 96% 60% Any aspect of daytime function that suffers 82% 93% 80% Morning low energy 71% 64% 80% At least some daytime sleepiness 58% 86% 40% Morning sleepiness 53% 68% 60% Difficulty awakening in the morning 49% 64% 20% Morning irritability 38% 57% 40% Excessive daytime sleepiness 31% 54% 40% Difficulty scheduling sleep and waking 31% 50% 20% Feeling anxious in the morning 27% 43% 60% aA

5

Total Other (O) N=5

N = 83

100% 60% 60% 80% 60% 60% 40% 20% 60% 60% 40% 40% 40% 60% 20% 20% 20%

80% 75% 72% 71% 71% 71% 59% 51% 45% 43% 42% 42% 42% 40% 34% 33% 31%

100% 100% 100% 100% 0% 60% 20% 60% 0% 40% 20%

93% 88% 87% 71% 63% 59% 51% 46% 37% 37% 34%

local self-report sleep questionnaire used in clinical evaluations

Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

552

Symptom Reports in Severe Chronic Insomnia—Moul et al

Table 5—Clinical Ratings in Primary Insomnia (PI) and Depression-related (DR) Insomnia

Survey of Sleep Total Symptom Count

Primary Insomnia Female Maleb 15.6 20.5 (7.4) (6.9)

Pittsburgh Sleep Quality Index

11.9 (4.0)

14.4 (2.6)

13.1 (3.0)

13.0 (3.7)

F = 0.27 p = 0.60

F = 1.91 p = 0.17

Beck Depression Inventory without Sleep Items

7.8 (5.6)

11.2 (8.3)

13.7 (5.6)

17.0 (11.6)

F = 8.33 p = 0.0053

F = 2.94 p = 0.091

Epworth Sleepiness Scale

3.7 (3.4)

4.8 (4.6)

6.3 (5.5)

6.5 (5.3)

F = 4.09 p = 0.046

F = 0.69 p = 0.41

aNo

Depression-Related Male Female 20.2 25.8 (5.1) (8.9)

ANOVAa (F, p) Group Gender F = 9.50 F = 8.09 p = 0.0029 P = 0.0058

Group by gender interaction ANOVA was significant; bValues are mean (SD)

Table 6—Exploratory Factor Analysis of Symptoms from the Survey of Sleep

Factor† Symptom Physically tense at bedtime Evening restless legs sensations Difficulty scheduling sleeping & awakening Difficulty falling asleep Morning depressed mood Morning irritability Daytime anxiousness Feet jerk at night Daytime fatigue Awakened by dreams Interference with ability to function Regular nocturnal awakenings Bothered by nightmares Difficulty staying asleep Awakens for no particular reason Awakens early & unable to fall back asleep Excessive daytime sleepiness At least some daytime sleepiness Difficulty awakening in the morning Morning sleepiness Awakened by bedpartner Awakened by light Awakened by noises Variance associated with factor

†Factor

1

2

3

4

5

0.74 0.55 0.48 0.48 0.47 0.44 0.42 0.42 0.59 0.49 0.45 0.43 0.40 0.55 0.51 0.46 0.63 0.55 0.49 0.49

8.6%

5.4%

5.0%

5.0%

0.63 0.62 0.46 4.7%

h2§ 0.56 0.38 0.36 0.48 0.30 0.31 0.30 0.30 0.47 0.28 0.30 0.27 0.32 0.32 0.28 0.26 0.42 0.35 0.37 0.29 0.45 0.46 0.27

loadings selected if greater than 0.4; §Commonality (h2) variance associated with factor loadings.

ping), and 5) sleep disturbances and irritability (e.g., numerous awakenings, difficulty returning to sleep, irritability, awakening for no reason). 2. Symptoms from the survey of sleep The SOS symptoms endorsed by more than 30% of patients are displayed in Table 4 (all symptoms available upon request). From the available 63 SOS items, patients endorsed a mean of 20.5 symptoms. Women endorsed more symptoms than men Vol. 25, No. 5, 2002 DownloadedSLEEP, from https://academic.oup.com/sleep/article-abstract/25/5/548/2750158 by guest on 07 June 2018

(22.5±8.3 vs. 17.8±7.6; t=2.66; df=77; p=0.01), but symptom count was not correlated with age(r=-0.09). Patients overwhelmingly endorsed severe problems with insomnia (Table 1), as reflected in marked distress, chronicity, and poor sleep parameters. “Feeling anxious or emotionally tense, or worrying about things at bedtime” and “Usually feeling sleepy during the daytime” were endorsed more frequently in the DR subgroup than in the PI subgroup (Fisher’s test, p=0.03 and p=0.02, respectively). 553

Symptom Reports in Severe Chronic Insomnia—Moul et al

Table 7—Hopkins Symptom Checklist Items that Separated Depression-Related from Primary Insomniaa Prevalence Comparisons Primary Depression Insomnia (PI) Related (DR) N = 42 N = 27 7% 48% 5% 33% 2% 35% 5% 30% 5% 32% 2% 30% 2% 26% 2% 23% 2% 23% 0% 30% 0% 22% 0% 26% 2% 19% 0% 19% 0% 15%

Selected Hopkins Checklist Items Feel Blue Nervousness Feel Lonely Excessive Restlessness No Interest Self-Conscious With Others Uneasy When Watched Cry Easily Feel Guilty Worthlessness Feeling Taken Advantage Of Never Close to Another Uneasy in Crowds Lonely Even with Others Feeling Watched aPercents

Observed Odds Ratio For DR vs. PI 12 10 21 8 9 17 14 12 12 >30 >30 >30 9 >30 >30

reflect symptoms that were "Quite a Bit" or "Extremely" bothersome.

Table 8—Linear Regression of Log Number of Severe Symptoms on the Hopkins Symptom Checklist Parameter Intercept Hours of sleep Group: PI vs Non-PI Gender Group x Hours Interaction Gender x Hours Interaction

ß value 2.51 -0.11 -1.11 -1.08

SE 0.4 0.07 0.38 0.39

t 6.34 -1.44 -2.91 -2.76

p < 0.001 0.115 0.005 0.007

0.14

0.07

1.98

0.052

0.17

0.07

2.29

0.025

test, p = 0.0003). BeckNS scores were in the range of moderate depression (see Table 1), with the DR subgroup having nonsignificantly higher scores (15.6 ± 9.4) than the PI subgroup (9.7 ± 7.4) (Wilcoxon Z = 1.91, p = 0.06). There was an overall trend-level association of BeckNS with gender (women > men) (Wilcoxon Z = 0.162, p = 0.1), but not with age. Since the bivariate analyses suggested possible influences of subgroup and gender, a 2-way ANOVA in which DR versus PI subgrouping and gender were factors was performed. This ANOVA (see Table 5) demonstrated that the DR group had higher BeckNS scores, adjusting for gender. The ESS scores were generally in the subclinical range (see Table 1), and uncorrelated with age or sex. This measure indicated that patients had little propensity to doze in inappropriate situations. The subgroups differed (Kruskal-Wallis χ2 = 9.70, df = 3, p = 0.021), but in the main comparison of interest, the DR subgroup did not have statistically higher ESS scores than the PI subgroup (Wilcoxon Z = 1.74, p = 0.08). However, after adjusting for gender in a 2-way ANOVA (see Table 5), the DR subgroup acknowledged more “tendency to doze” than the PI group, albeit within the subclinical range. The BeckNS scores were uncorrelated ( r = 0.26) with ESS scores (t = 1.78, df = 77, p > 0.05), and subgroups had similar correlations. On the HSCL90, this sample reported substantial distress (see Table 1). Diagnostic subgroups differed in number of severe responses (Kruskal-Wallis χ2= 8.55, df = 3, P = 0.036), especially between the DR subgroup (17.4 ± 17.8) and the PI subgroup (6.5 ± 6.9 )(Wilcoxon Z = 2.49, p = 0.013). Except for sleep disturbance items, SCL90 subscale scores were consistently lower in the PI subgroup compared to the DR subgroup (data not shown). Severe responses for items in Table 7 met the stipulated criteria (odds ratio ≥8 with increased prevalence ≥15%) for discriminating between DR and PI subgroups. These items (scored as yes/no for “severe” responses) were summed to give an ad hoc “Non-PI Index Score” as applied to all non-PI subgroups. With

Multiple R2=0.33; Overall F=6.35 (DF=5,66), p