Cognitive processes and their association with persistence and remission of insomnia: Findings from a longitudinal study in the general population

Behaviour Research and Therapy 54 (2014) 38e48

Contents lists available at ScienceDirect

Behaviour Research and Therapy journal homepage: www.elsevier.com/locate/brat

Cognitive processes and their association with persistence and remission of insomnia: Findings from a longitudinal study in the general population Annika Norell-Clarke a, *, Markus Jansson-Fröjmark a, b, Maria Tillfors a, Allison G. Harvey c, Steven J. Linton a a b c

School of Law, Psychology, and Social Work, Örebro University, SE-701 82 Örebro, Sweden Department of Psychology, Stockholm University, SE-106 91 Stockholm, Sweden Department of Psychology, University of California, 3210 Tolman Hall, Berkeley, CA 94720-1650, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 May 2013 Received in revised form 16 January 2014 Accepted 17 January 2014

Aim: Insomnia is a common health problem that affects about 10% of the population. The purpose of this investigation was to examine the association between cognitive processes and the persistence and remission from insomnia in the general population. Methods: In a longitudinal design, 2333 participants completed a survey on night time and daytime symptoms, and cognitive processes. Follow-up surveys were sent out six months and 18 months after the first assessment. Participants were categorised as having persistent insomnia, being in remission from insomnia or being a normal sleeper. Results: Cognitive processes distinguished between people with persistent insomnia and normal sleepers. Specifically, worry, dysfunctional beliefs, somatic arousal, selective attention and monitoring, and safety behaviours increased the likelihood of reporting persistent insomnia rather than normal sleep. For people with insomnia, more worry about sleep at baseline predicted persistent insomnia but not remission later on. Lower selective attention and monitoring, and use of safety behaviours over time increased the likelihood of remission from insomnia. In general, these results remained, when psychiatric symptoms and medical complaints were added to the models. Conclusions: The findings support that certain cognitive processes may be associated with persistence and remission of insomnia. Clinical implications are discussed. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Insomnia Worry Safety behaviours Selective attention Dysfunctional beliefs Physiological arousal

Insomnia is a common sleep disorder that affects 9e12% of the population worldwide, with up to a third of the population suffering from insomnia symptoms (Ford & Kamerow, 1989; Ohayon & Reynolds III, 2009). Insomnia is defined by difficulties initiating or maintaining sleep, or early morning awakenings with inability to fall asleep. The sleep disturbance causes distress or an impaired ability to function in important areas, such as social or work related contexts (American Psychiatric Association, 2000, 2013). Insomnia can be acute or persistent (Ellis, Gehrman, Espie, Riemann, & Perlis, 2012). Whilst a recent definition of acute insomnia emphasises that it is caused by life events or distress at the current situation (Ellis et al., 2012), the literature on persistent

* Corresponding author. Tel.: þ46 1930 1272; fax: þ46 1930 3484. E-mail address: [email protected] (A. Norell-Clarke). 0005-7967/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.brat.2014.01.002

insomnia describes perpetuating factors more directly related to sleep and the sleep situation, such as sleep habits, heightened arousal, and dysfunctional beliefs about sleep (e.g. Buysse, Germain, Hall, Monk, & Nofzinger, 2011; Espie, 2002; Harvey, 2002; Lundh & Broman, 2000; Morin, 1993; Perlis, Giles, Mendelson, Bootzin, & Wyatt, 1997). Insomnia is associated with high costs through healthcare appointments and sleep enhancing drugs (Daley, LeBlanc, Grégoire, & Savard, 2009), as well as work absenteeism (Linton & Bryngelsson, 2000) and an increased prevalence of vehicle accidents at work and after work (Léger, Massuel, Metlaine, & The SISYPHE Study Group, 2006). Insomnia also constitutes a well-established increased risk of developing depression for both adults (Ford & Kamerow, 1989) and adolescents (Breslau, Roth, Rosenthal, & Andreski, 1996). Insomnia is thus associated with severe consequences for the individual as well as high costs for society. Hence it is important to investigate possible maintaining processes.

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

Cognitive processes of insomnia Many attempts have been made to explain what drives chronic insomnia (e.g. Buysse et al., 2011; Espie, 2002; Harvey, 2002; Lundh & Broman, 2000; Morin, 1993; Perlis et al., 1997). The following cognitive processes have been suggested across various models of insomnia: worry, dysfunctional beliefs, arousal, selective attention and monitoring, and safety behaviours (maladaptive habits). Worry is theorised to maintain insomnia; both worry specifically related to sleep problems or consequences of poor sleep as well as more general worries (Espie, 2002; Harvey, 2005; Morin, Stone, Trinkle, Mercer, & Remsberg, 1993). Worry is thought to disturb sleep by triggering arousal, and to be maintained by dysfunctional beliefs and selective attention (discovering more reasons to be worried) (Harvey, 2005). Studies have shown that it is possible to increase sleep latency by inducing worry (Gross & Borkovec, 1982), or shorten it by targeting worry through an intervention (Carney & Waters, 2006). Dysfunctional beliefs about sleep include unhelpful beliefs about the amount of sleep needed every night, or fearful ideas about what will happen to one’s health and ability to function in different areas in life, if insomnia persists. Dysfunctional beliefs are believed to trigger worry about sleep, both during day and night (Morin et al., 1993), and motivate the use of safety behaviours in an attempt to avoid feared outcomes (Harvey, 2005). People with insomnia have more dysfunctional beliefs about sleep than normal sleepers (Carney et al., 2010). The results of a longitudinal study indicated that dysfunctional beliefs about sleep were related to persistent insomnia and poor sleep over time (Jansson & Linton, 2007) and a cross-sectional study showed that dysfunctional beliefs were positively associated with the use of safety behaviours (Woodley & Smith, 2006). A review of mediating factors in insomnia treatment showed that cognitive behaviour therapy for insomnia (CBT-I) was consistently associated with reductions in dysfunctional beliefs, and that these reductions were associated with improvements on both subjective and objective sleep outcomes (Schwartz & Carney, 2012). Arousal is a core feature in many insomnia models, and it has been conceptualised as somatic, cognitive and cortical (e.g. Buysse et al., 2011; Espie, 2002; Harvey, 2002; Lundh & Broman, 2000; Morin, 1993; Perlis et al., 1997). Hyperarousal models of insomnia propose that conditioned cognitive and somatic arousal are perpetuating factors of insomnia (Perlis et al., 1997; Riemann et al., 2010). The bed, the bedroom or bedtime rituals are conditioned to arousal from the unhelpful practice of spending excessive time in bed while awake. Cognitive arousal is experienced as increased cognitive activity (e.g. a racing mind). Somatic arousal could elicit symptoms similar to those of a “fight or flight” response from the sympathetic nervous system, for example tense muscles, rapid heartbeat, and a restless/nervous sensation in the body. There are also studies supporting local CNS activation: elevated cortical arousal during sleep, and this is believed to set the stage for sleepstate misperception through increased sensory processing (hearing noises), information processing (being able to think), and memory encoding (Drummond, Smith, Orff, Chengazi, & Perlis, 2004; Perlis et al., 1997; Riemann et al., 2010). These phenomena, which normally would be inhibited during sleep, give the impression of being awake rather than being asleep. Arousal has been linked to the maintenance of insomnia in longitudinal studies (Jansson & Linton, 2007; Jansson-Fröjmark, Lundquist, Lundquist, & Linton, 2008) and higher arousability is a predictor of insomnia incidence (LeBlanc et al., 2009). Although arousal may strictly speaking not be a cognitive concept, it has been associated with cognitive processes in cognitive models of insomnia and warrants investigation together with cognitive processes (Espie, 2002; Harvey, 2002).

39

Selective attention and monitoring for threats refers to the narrowing of the focus of attention and scanning of anything that could be perceived as a threat to sleep and regards both external stimuli such as noises or monitoring the clock to keep track of time, and internal stimuli such as tensions of the body. Based on the assumption that normal sleep is automatic and effortless, selective attention is believed to disturb the transition between wake and sleep (Espie, Broomfield, MacMahon, Macphee, & Taylor, 2006). Experimental studies have found support for attentional bias in insomnia. For example, instructions to monitor a clock during the night increased sleep onset latency and worry for people with poor sleep and good sleep alike (Tang, Schmidt, & Harvey, 2007), and people with insomnia have displayed attentional bias towards sleep-related stimuli in several experimental paradigms (Marchetti, Biello, Broomfield, Macmahon, & Espie, 2006; Spiegelhalder, Espie, & Riemann, 2009; Woods, Marchetti, Biello, & Espie, 2009; Woods, Scheepers, Ross, Espie, & Biello, 2013). Safety behaviours are subtle behaviours people use in an attempt to avoid feared outcomes (Salkovskis, 1991). In the case of insomnia, common fears include the fear of not falling asleep or fear of negative consequences of poor sleep, such as failing at work or becoming ill due to sleeplessness. Safety behaviours can be overt and, for example, include going to bed very early to allow for plenty of time to fall asleep, cancelling appointments after a poor night as they are perceived as too energy consuming, and napping during daytime in order to feel more energised. Covert safety behaviours, such as attempts to suppress unwanted thoughts while trying to go to sleep, are also possible. Safety behaviours may be helpful in the short term but often have undesirable long-term effects, as they may interfere with the kind of regular sleep schedule that would promote healthy sleep (Morin, 1993). Another disadvantage of safety behaviours is that they prevent dysfunctional beliefs from being tested and corrected. To summarise, according to several insomnia models, there are maintaining cognitive processes that hinder normal sleep and worsen sleep problems by creating vicious cycles. It is already known from cross-sectional studies that people with insomnia experience cognitive processes to a larger degree than people without insomnia. For example, people with insomnia experienced worry, dysfunctional beliefs, arousal, selective attention and monitoring, and safety behaviours to a significantly larger extent, compared with people with poor sleep, who in turn scored higher than normal sleepers (Jansson-Fröjmark, Harvey, Norell-Clarke, & Linton, 2012). However, a gap in current knowledge is the association between insomnia and cognitive processes over time. A little over ten years ago it was noted that almost all epidemiological studies of insomnia in the general population were cross-sectional, although exceptions should be noted (e. g. The Zurich Study: Vollrath, Wicki, & Angst, 1989), and that much was unknown regarding the development of insomnia (Ohayon, 2002). In The Zurich Study (Vollrath et al., 1989), insomnia increased the risk of future insomnia, and longitudinal research since then has also pointed towards the chronic nature of insomnia, finding insomnia episodes to be a risk factor for future insomnia and that insomnia complaints often persist over time (LeBlanc et al., 2009; Morin et al., 2009). Prospective studies, investigating differences between those who develop chronic insomnia and those who do not, have found that several aspects of physical and mental health predicted insomnia (LeBlanc et al., 2009; Singareddy et al., 2012). Premorbid reports of people who would later develop insomnia showed that they reported poorer general health, more pain, higher arousability, and more depression and anxiety (LeBlanc et al., 2009). Data from the Penn State Sleep Cohort showed that those who reported poor mental health were more likely to develop chronic insomnia (Singareddy et al., 2012). The longitudinal studies

40

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

have made significant contributions towards a greater understanding of the aetiology of insomnia. Although previous studies have investigated the longitudinal association between insomnia and mental health, focus has mostly been on psychiatric diagnoses or domains of general psychological functioning rather than on insomnia-specific concepts. Little is known about the longitudinal association between cognitive insomnia processes, such as worry about sleep, and insomnia. Also, previous research of cognitive processes associated with insomnia has mostly focused on one process at a time, and rarely prospectively in the population. Moreover, samples are often comprised of help-seeking individuals. Hence we cannot know if the results are generalised to the general population, as many individuals with sleep problems never seek help (Ancoli-Israel & Roth, 1999; Morin, LeBlanc, Daley, Gregoire, & Mérette, 2006). The present study was designed to begin the process of overcoming these limitations. Aim of the study The overall purpose of this investigation was to examine cognitive processes implicated in insomnia; worry, dysfunctional beliefs about sleep, arousal, selective attention and monitoring, and safety behaviours, and their association to insomnia persistence and remission. A longitudinal design with three measuring points was employed: baseline, six months later and 18 months later. On the basis that the cognitive processes are proposed to maintain insomnia (Espie, 2002; Harvey, 2002), three hypotheses were tested: 1. The cognitive processes at baseline can discriminate between people with persistent insomnia and people with persistent normal sleep.

Participants A random sample of 5000 people, ages 18 to 70, from two Swedish counties, was contacted. The random sample was obtained from the national register, in which all residents are listed, and the randomisation was conducted by Statistics Sweden, the national statistics department. 2333 people (47.1% of those that were eligible) returned the survey at time 1, 1887 at time 2, 6 months later, and 1795 returned the survey at time 3, 18 months after time 1. The mean age was 47.9 years, and 56.6% of the participants were women. As for marital status, 12.8% reported being single, 80.1% being cohabitant or married or having a partner, 4.0% being divorced, and 1.6% being widowed. Regarding vocational status, 73.1% were employed (full or part-time) or students and 26.2% were unemployed, on sick leave, on pension or other status. Concerning educational level, 25.4% had compulsory school as their highest level of education, 43.3% high school, and 29.7% college or university. Most, 91.3%, were born in Sweden. Comparisons with register data showed that our sample was representative for the Swedish population regarding age, gender, relationship status, occupational status, educational level, and reported sleep disturbance (JanssonFröjmark et al., 2012). Demographic data for people with insomnia and normal sleepers respectively are available in Table 1. The insomnia group consisted of significantly more women, more divorced, widowed or single people, more people outside the work force, and more people born outside Sweden. Attrition analyses showed no differences for gender, sleep disturbance, or insomnia severity although non-responders were more likely to be younger than responders. The attrition analyses have been described in detail elsewhere (Jansson-Fröjmark et al., 2012). Measures and classifications

The second and third hypotheses pertain to people with insomnia at baseline. 2. For people with insomnia, higher degrees of the cognitive processes at baseline will increase the likelihood of reporting persistent insomnia rather than remission from insomnia later on. 3. For people with insomnia, a diminishing of the cognitive processes over time will increase the likelihood of reporting remission from insomnia rather than persistent insomnia later on.

Method Overview of the study This research is part of the Prospective Investigation on Psychological Processes for Insomnia (PIPPI) study and was approved by the Regional Ethics Board in Uppsala, Sweden. A survey regarding demographic factors, sleep, daytime impairment, and health was mailed to a random sample from the general public on three occasions, in September 2008, March 2009, and April 2010, spanning over 18 months in total. Data from all three times were used in this longitudinal study. All communication with participants was through post and all data was gathered from questionnaires. No response on the survey or the response promoting efforts within two weeks after was considered a declination to participate. Respondents at time 1 were sent a questionnaire at time 2, and at time 3. In order to increase the response rates at the three assessment points, a number of steps were taken in line with a recent Cochrane review (Edwards et al., 2007).

Sleep measures Night time symptoms. To assess sleep problems the participants were asked to complete the following categorical questions based on the previous month: 1. How many minutes, on average, do you lie awake in bed before you fall asleep (after the lights are out)? (Sleep onset latency: SOL; <15 min, 16e30 min, 31e60 min, >60 min). 2. How many minutes, on average, are you awake during the night? (Wake time after sleep onset: WASO; <15 min, 16e 30 min, 31e60 min, >60 min). 3. How many minutes, on average, do you wake up too soon in the morning (earlier than you would wish for)? (Early morning awakening: EMA; <15 min, 16e30 min, 31e60 min, >60 min). Table 1 Demographic variables at baseline T1.

Age (M & SD) Women Born in Sweden Marital status Single, divorced, or widowed Married, cohabitant, or in a relationship College/university education Occupational status Working/studying Unemployed, on sick leave, retired or other

Normal sleep n ¼ 1706

Insomnia n ¼ 322

c2/t

47.54 (14.62) 708 (51.2%) 1282 (93.6%)

47.18 (14.13) 198 (61.5%) 282 (89.8%)

0.40 10.79** 5.15* 22.23**

239 (17.5%) 1126 (82.5%)

94 (29.4%) 226 (70.6. %)

385 (28.2%)

107 (33.3%)

1027 (76.9%) 308 (23.1%)

216 (69.5%) 95 (30.5%)

Note. *Significant at the .05 level. **Significant at the .01 level.

3.04 7.23**

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

4. How would you describe your sleep quality? (Sleep quality: very good (1), quite good (2), neither good nor poor (3), quite poor (4), very poor (5).) Daytime symptoms and distress. The participants were instructed to report on the degree of sleep-related impairment during the previous month. The first question was “When your sleep has been disturbed: How have you been affected during the day”? The following potential areas of impairment were listed: fatigue/malaise, impairment in attention, concentration, or memory, social dysfunction, vocational dysfunction, mood disturbance, irritability, daytime sleepiness, reduction in motivation, energy, or initiative, proneness for errors or accidents at work or while driving, tension headaches, gastrointestinal symptoms, and concerns or worries about sleep. The response alternatives for these indications of daytime impairment were: not at all (1), somewhat (2), quite much (3), and a lot (4). The second question was “When your sleep has been disturbed: How have you been affected during the day in the following domains”? The following impairment domains were listed: work/studies, leisure time, and social situations. The response alternatives for the three functional domains were: not affected (1), some difficulties but not less ability to function (2), less ability to function (3), much less ability to function (4) could barely function (5). All the impairment items were summed and a composite score for daytime impairment was computed. Sleep disorders. The SLEEP-50 was used to assess five DSM-IVTR sleep disorders: apnoea, narcolepsy, restless legs/periodic limb movement disorder, circadian rhythm disorder, and sleepwalking (Spoormaker, Verbeek, van den Bout, & Klip, 2005). The instrument has high internal consistency, with a testeretest reliability ranging between .71 and .88 and a factor structure that matches the DSMIV-TR sleep disorders (American Psychiatric Association, 2000). The sensitivity and specificity scores have been found to be reasonable for the sleep disorders in a sample of college students, clinical patients and healthy volunteers (the sensitivity for the different sleep disorders ranging between .67 and 1.00 and specificity ranging between .69 and 1.00). The agreement between clinical diagnoses and classification derived from the SLEEP-50 is acceptable (kappa ¼ .77). The response alternatives ranged from 1 to 4 (1 ¼ not at all, 4 ¼ very) regarding whether the items had been applicable during the past month. Five additional subgroups in the SLEEP-50 (insomnia, affective disorder, hypersomnia, nightmares and sleep state misperception) were not used in the current study. The insomnia categorisation in the SLEEP-50 was not employed since it is not based on established research criteria. The internal consistency was .92 for the current sample. Sleep status group classification. The participants were classified into two different sleep status groups, depending on their sleep patterns, indication of other sleep disorders than insomnia, and daytime impairment. 309 participants who reported other sleep disorders were omitted from the analyses. Insomnia. A proxy for DSM-IV-TR definition of insomnia disorder was used (the version of DSM current at the time of data collection), based on the Research Diagnostic Criteria for Insomnia Disorder (Edinger et al., 2004). Participants had to confirm a sleep disturbance on the question “During the last month: Would you say that you have had any sleep problems?”, and report initial, middle, or late insomnia (>30 min awake involuntarily at any stage during an estimated average night) (Lichstein, Durrence, Taylor, Bush, & Riedel, 2003). They also had to report some daytime impairment (score 3, 4 or 5 on at least one daytime symptoms or impaired function distress item). They must not meet criteria for apnoea, narcolepsy, restless legs syndrome/periodic limb movement disorder, circadian rhythm disorder or, sleepwalking as assessed with the SLEEP-50. Using data from the current study, the concordance

41

between this insomnia definition and two validated Insomnia Severity Index (ISI) cut-offs were investigated (Morin, Belleville, Bélanger, & Ivers, 2011). ISI measures subjective insomnia severity and has widely been used as a measure of change after treatment. A score below 8 is a cut-off for sub threshold insomnia in both community and clinical samples and the cut-off at 10, marking insomnia, has been validated in community samples of insomniacs. The two ISI cut-offs that have been validated have both shown nearly 98% correct classification rates differentiating clinical insomnia patients and normal sleepers. The concordance between ISI cut-offs and the insomnia classification in this study was high. The ISI cut-off at 8 points correctly classified 99.4% of people with insomnia and the ISI cut-off at 10 points correctly classified 89.4% (Jansson-Fröjmark et al., 2012). The high concordance with an established insomnia measure indicates that our insomnia definition captures the insomnia construct to an acceptable degree. Normal sleep. Participants had to deny a sleep disturbance on the item “During the last month: Would you say that you have had any sleep problems?” and not fulfil criteria for any sleep disorder as assessed with the SLEEP-50. Process measures Anxiety and preoccupation about sleep. The Anxiety and Preoccupation about Sleep Questionnaire (APSQ) was used to assess sleep-related worry (Tang & Harvey, 2004). The response alternatives for each of the 10 items were changed from the original 1e10 (1 ¼ strongly disagree, 10 ¼ strongly agree) to 1e5 (1 ¼ strongly disagree, 5 ¼ strongly agree) in the current study to make the alternatives similar to the other psychological process measures used in the battery of questionnaires. This version has been evaluated and found to be psychometrically sound (Jansson-Fröjmark, Harvey, Lundh, Norell-Clarke, & Linton, 2011). The score range was 10e50. For the current sample, the internal consistency was .93. Dysfunctional beliefs and attitudes about sleep. The Dysfunctional Beliefs and Attitudes about Sleep Scale (DBAS) was used to measure dysfunctional beliefs (Morin et al., 1993). A short version, the DBAS-10 was used to identify sleep-related dysfunctional beliefs i.e. beliefs at a schema level, and contains 10 items (Espie, Inglis, Harvey, & Tessier, 2000). The alternatives were changed slightly to make the response alternatives similar to other psychological process measures in the questionnaire (1e5; 1 ¼ strongly disagree, 5 ¼ strongly agree) and the score range was 10e50. Based on the current sample, the internal consistency was .85. Pre-sleep arousal. The Pre-Sleep Arousal Scale (PSAS) consists of two subscales which measures pre-sleep somatic arousal and cognitive arousal (Nicassio, Mendlowitz, Fussel, & Petras, 1985). The cognitive subscale (PSAS-C) was not used due to overlapping constructs between APSQ and PSAS-C and psychometric issues regarding PSAS-(Jansson-Fröjmark & Norell-Clarke, 2012). All eight items from the somatic subscale (PSAS-S) were used. The response alternatives ranged from 1 (1 ¼ not at all) to 5 (5 ¼ extremely) and the score range was 8e40. Based on the current sample, the internal consistency for the PSAS-S was .72. Sleep associated monitoring. The Sleep Associated Monitoring Index (SAMI) was used to measure monitoring and attentional bias regarding sleep. In the current study, the original 30 items had been reduced to 8 combined items, representing a subscale each, consisting of several examples of monitoring within similar areas of focus. As the eight subscales are highly correlated with the subscale totals (Semler & Harvey, 2004), this seemed reasonable. The response alternatives ranged from 1 (not at all) to 5 (all the time). Based on the current sample, the internal consistency was .80. Sleep-related behaviours. Sleep-Related Behaviors Questionnaire (SRBQ) was used to measure sleep-related safety behaviours (Ree & Harvey, 2004). The scale was reduced from 32 to 16

42

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

questions and the remaining items were chosen for their .40 correlation or higher with Insomnia Severity Index (correlations from Ree & Harvey, 2004) in a sample of insomnia patients, and healthy subjects from the general population and university students. The response alternatives ranged from 0 (almost never) to 4 (almost always), making the score range 0e64. The internal consistency was .88 for the 16 item scale, based on the current sample. Health measures Anxiety and depression. The Hospital Anxiety and Depression Scale (HADS) was used to measure anxiety and depression (Herrmann, 1997; Zigmond & Snaith, 1983). HADS is a self-rating scale with fourteen questions in which the severity of anxiety and depression is rated on 4-point scales (score range 0e21). To detect a possible case of anxiety or depression, an often-used procedure in research has been to dichotomise the two subscales: a score of 7 or less indicates a non-case and a score of 8 or more a definite case (Herrmann, 1997). This cut-off results in sensitivities and specificities for both subscales of approximately .80 when compared with structured or semi-structured diagnostic interview (Bjelland, Dahl, Tangen Haug, & Neckelmann, 2002). For the current sample, the internal consistency was .85 for the anxiety subscale and .84 for the depression subscale. Medical disorders. The instruction was as follows: “Have you had, during the past month, one or more of the following medical disorders? Please check the appropriate box for all that applies.” The listed medical disorders were: heart disease (e.g., angina, cardiac arrhythmias or heart attack), metabolic disorders (e.g., diabetes), pulmonary disease (COPD, asthma or emphysema), kidney disease (e.g., kidney failure), gastrointestinal disorders [e.g., stomach or duodenal ulcers or reflux disease (GERD)], autoimmune disorders (e.g., lupus), neurologic disorders (e.g., Huntington’s disease, epilepsy, Parkinson’s disease or Alzheimer’s disease), cancer, head trauma, pain in neck, back or shoulders, headaches, and endocrine disorders (e.g., thyroid disease). Almost identical instructions and items have been used in previous research (e.g. Zhang et al., 2012). Statistical analyses Binary logistic regression was used to explore all three hypotheses. Data from all three measure times were chosen for analyses regarding the following time spans: T1 e T2, T2 e T3, and T1 e T3. Preliminary analyses were carried out to test for multicollinearity between the five predictors worry, dysfunctional beliefs about sleep, somatic arousal, selective attention and monitoring, and safety behaviours, using Pearson’s r and collinearity statistics available through SPSS. The variance inflation factor (VIF) was below 10 in all instances, and tolerance statistics were above .20, which indicated that there was no multicollinearity, and none of the correlations were above the critical value of .80 (Field, 2009), although some correlations were in the vicinity. For the first hypothesis, correlations around .70 were found between worry (APSQ) and safety behaviours (SRBQ) as well as between worry (APSQ) and dysfunctional beliefs (DBAS). Further, worry and dysfunctional beliefs in both the second and third aim showed correlations near and above .70. The results of these correlations from the analytical samples are reported in Table 2 for their respective hypothesis. The issue of potential multicollinearity will be elaborated upon in the results. For the first hypothesis the predictors consisted of the five cognitive processes at baseline, with persistent sleep status as the outcome. Persistent sleep status was defined as either being classified as having normal sleep at the two points included in the analysis (e.g. T1 and T2, T2 and T3, or T1 and T3), or insomnia at both

Table 2 Range of correlations between predictor variables for T1eT2, T2eT3 and T1eT3.

Hypothesis 1 1 APSQ 2 DBAS 3 PSAS-S 4 SAMI 5 SRBQ Hypothesis 2 1 APSQ 2 DBAS 3 PSAS-S 4 SAMI 5 SRBQ Hypothesis 3 1 APSQ 2 DBAS 3 PSAS-S 4 SAMI 5 SRBQ

1

2

3

4

.64e.68** .47e.50** .57e.61** .66e.70**

.40e.41** .55e.56** 53e.56**

.57e.61** .54e.56**

.62e.64**

.73e.77** .23e.30** .42e.52** .57e.64**

.23e.25** .34e.50** .47e.55**

.34e.47** .35e.43**

.47e.52**

.63e.74** .24e.30*/** .25e.34*/** .38e.52**

.28e.29** .30e.34** .35e.52**

.26e.45** .13nse.31**

.44e.47**

Note. *Significant at the .05 level. **Significant at the .01 level. ns ¼ not significant.

time points. The second and third hypotheses regarded only those with insomnia and the development to remission from insomnia or persistence. For the second hypothesis insomnia status (persistence or remission) was the outcome and the five cognitive processes at baseline were predictors. Persistent insomnia was defined as having insomnia at both time points in the analysis, and remission was defined as having insomnia at the first time but normal sleep at the second time point. For the third hypothesis insomnia status (persistence or remission) was the outcome and the standardised residual change scores of the five cognitive processes between two times were predictor variables. The standardised residual is a way of expressing the score at time 2 as larger or smaller than the score predicted linearly by time 1 score (Cronbach & Furby, 1970). Standardised residual change scores are recommended rather than raw change scores as standardised residuals take into account the score at time one, while at the same time controlling for possible random errors of measurement (Steketee & Chambless, 1992). The standardised residuals were computed by converting the raw scores to Z scores and were further computed as follows: Z2  (Z1  r12) in which Z2 is the follow-up score, Z1 the baseline score and r12 the correlation between both ratings. Results As medical and psychiatric comorbidities could impact the results, this was controlled for by testing the models with and without comorbid conditions that had been found to discriminate between people with insomnia and people with normal sleep, in this sample (Jansson-Fröjmark, Norell-Clarke, & Linton, in preparation). Having an anxiety disorder (HADS-A 8: odds ratio ¼ 5.70), a mood disorder (HADS-D 8: odds ratio ¼ 3.49), pain in neck, back or shoulders (odds ratio ¼ 1.60), and headaches (odds ratio ¼ 1.49) were associated with an increased risk for fulfilling criteria for insomnia at baseline, relative to not having insomnia. The following comorbid disorders were not significantly related to insomnia at baseline in this study and were therefore excluded from the subsequent longitudinal analyses: heart disease, metabolic disorders, pulmonary disease, kidney disease, gastrointestinal disorders, autoimmune disorders, neurologic disorders, cancer, head trauma, and endocrine disorders. It should be noted that most of these conditions were rare, yielding less than 10 participants. Thus only anxiety disorder, mood disorder, pain in the neck, back or shoulders, and headaches were added to the hypotheses testing of the model with cognitive processes.

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

Hypothesis 1: The Cognitive Processes at Baseline Can Discriminate Between People With Persistent Insomnia and People With Persistent Normal Sleep. The descriptive statistics for the cognitive processes across the two sleep groups (persistent insomnia and persistent normal sleep) and the results from logistic regressions are displayed in Table 3. The overall model with worry, dysfunctional beliefs about sleep, somatic arousal, selective attention and monitoring, and safety behaviours as predictors, was significant for all three time spans, and explained between .22 (Cox & Snell r2) and .57 (Nagelkerke r2) of the variance in sleep status, which indicates that the model could distinguish between those with persistent insomnia and persistent normal sleep. APSQ, PSAS-S, SAMI and SRBQ made uniquely significant contributions to the model, for all three time spans, with odds ratios ranging between 1.00 and 1.16. More worry about sleep, somatic arousal, selective attention and monitoring, and safety behaviours were thus associated with a higher likelihood of persistent insomnia than persistent normal sleep. However, DBAS was a significant predictor for T2eT3 only, with negative beta value and an odds ratio below 1, indicating that the more dysfunctional beliefs a person had, the more likely a person would be to have normal sleep. This was contradictory to the hypothesis and also puzzling as people with persistent insomnia had reported higher degrees dysfunctional beliefs than people with normal sleep. The high correlation between APSQ and DBAS (see Table 2) prompted further analyses without APSQ and DBAS respectively, as multicollinearity was suspected between the two predictors. When APSQ was removed, DBAS was a unique significant predictor for T1eT2 and T1eT3 and had positive beta values and odds ratios above 1 for

43

all time spans (1.03e1.08), meaning that high degrees of dysfunctional beliefs increased the likelihood of having persistent insomnia rather than persistent normal sleep. This model explained between .20 (Cox & Snell r2) and .54 (Nagelkerke r2) for the three time spans. With DBAS removed, the remaining four predictors were unique contributors to the model, albeit with slightly less variance explained by the overall model (between .21 and .57) than the analyses with all predictors included. This supports that the variance explained by APSQ and DBAS was highly overlapping, thus causing a negative beta in DBAS. When entering the cognitive processes together with anxiety, depression, pain, and headache simultaneously, across the three time spans, only pain was a significant predictor, i.e., discriminating between those with persistent insomnia and those with persistent normal sleep. While SRBQ, APSQ and SAMI remained significant in the adjusted models, DBAS and PSAS-S lost their significance. The increase in explained variance in the adjusted models was very small (Nagelkerke r2 increase with .01e.02). To summarise, in general our results support the hypothesis that the cognitive processes can discriminate between persistent insomnia and persistent normal sleep. Hypothesis 2: For People With Insomnia, Higher Degrees of the Cognitive Processes at Baseline Will Increase the Likelihood of Reporting Persistent Insomnia Rather than Remission From Insomnia Later On. The descriptive statistics for the cognitive processes at baseline across the two insomnia groups (persistent insomnia and remission from insomnia) and results from logistic regressions are displayed at Table 4. The overall model with worry, dysfunctional beliefs

Table 3 Cognitive processes at baseline between those with persistent insomnia and those with persistent normal sleep: descriptive statistics and logistic regression analyses. Predictor

M (SD)

B (SE)

Insomnia at T1 and T2 (n ¼ 145) e Normal Sleep at T1 and T2 (n ¼ 999). The full model was significant (p < .001) and predicted 91.3% of cases. APSQ (T1) NS: 11.81 (3.86) 0.14 (0.03)** I: 23.42 (9.02) DBAS (T1) NS: 18.37 (5.77) 0.03 (0.03) I: 26.70 (7.11) PSAS (T1) NS: 10.23 (2.75) 0.07 (0.03)* I: 14.57 (4.71) SAMI (T1) NS: 13.98 (4.38) 0.08 (0.03)** I: 21.39 (5.11) SRBQ (T1) NS: 22.04 (5.95) 0.09 (0.02)** I: 35.24 (9.01) Insomnia at T2 and T3 (n ¼ 102) e Normal Sleep at T2 and T3 (n ¼ 911) The full model was significant (p < .001) and predicted 92.3% of cases. APSQ (T2) NS: 11.93 (4.15) 0.13 (0.03)** I: 22.84 (9.89) DBAS (T2) NS: 18.57 (5.91) 0.06 (0.03)* I: 25.76 (7.09) PSAS (T2) NS: 10.25 (2,75) 0.08 (0.04)* I: 14.57 (4.59) SAMI (T2) NS: 14.08 (4.34) 0.08 (0.03)* I: 20.88 (5.32) SRBQ (T2) NS: 22.17 (6.27) 0.08 (0.02)** I: 34.86 (9.32) Insomnia at T1 and T3 (n ¼ 124) e Normal Sleep at T1 and T3 (n ¼ 945) The full model was significant (p < .001) and predicted 92.7% of cases. APSQ (T1) NS: 11.64 (3.46) 0.15 (0.03)** I: 24.37 (10.16) DBAS (T1) NS: 18.38 (5.73) 0.02 (0.03) I: 27.35 (7.39) PSAS (T1) NS: 10.07 (2.60) 0.09 (0.04)* I: 14.75 (4.84) SAMI (T1) NS: 13.88 (4.17) 0.09 (0.03)** I: 21.40 (5.52) SRBQ (T1) NS: 21.79 (5.77) 0.08 (0.02)** I: 35.73 (9.57)

Odds ratio

95% CI

r2

1.15

1.09e1.21

.29e.54

0.97

0.92e1.02

1.07

1.00e1.14

1.09

1.03e1.12

1.10

1.05e1.14

1.14

1.08e1.20

0.94

0.89e1.00

1.09

1.01e1.17

1.08

1.02e1.15

1.08

1.04e1.12

1.16

1.10e1.23

0.98

0.93e1.04

1.09

1.02e1.18

1.09

1.03e1.17

1.09

1.04e1.13

.22e.45

.29e.57

Note. The final solution is presented. CI ¼ confidence interval, I ¼ insomnia, M ¼ mean, NS ¼ normal sleep, SD ¼ standard deviation. *Significant at the .05 level. **Significant at the .01 level. r2 refers to Cox & Snell and Nagelkerke R Square respectively.

44

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

Table 4 Cognitive processes at baseline between those with persistent insomnia and those with remission from insomnia: descriptive statistics and logistic regression analyses. Predictor

M (SD)

B (SE)

Insomnia at T1 and T2 (n ¼ 145) e Remission: I at T1 and NS at T2 (n ¼ 60) The full model was significant (p < .001) and predicted 73.2% of cases. APSQ (T1) REM: 18.58 (8.20) 0.10 (0.03)** I: 23.42 (9.02) DBAS (T1) REM: 26.03 (7.26) 0.12 (0.04)** I: 26.70 (7.11) PSAS (T1) REM: 13.48 (4.41) 0.00 (0.04) I: 14.57 (4.71) SAMI (T1) REM: 19.38 (5.50) 0.04 (0.04) I: 21.39 (5.11) SRBQ (T1) REM: 30.45 (10.09) 0.04 (0.02)# I: 35.24 (9.01) Insomnia at T2 and T3 (n ¼ 102) e Remission: I at T2 and NS at T3 (n ¼ 39) The full model was significant (p ¼ .002) and predicted 71.6% of cases. APSQ (T2) REM: 16.56 (6.56) 0.08 (0.03)* I: 22.82 (9.89) DBAS (T2) REM: 22.92 (7.87) .005 (0.04) I: 25.76 (7.09) PSAS (T2) REM: 12.33 (4.13) 0.06 (0.06) I: 14.57 (4.59) SAMI (T2) REM: 18.49 (5.57) 0.01 (0.05) I: 20.88 (5.33) SRBQ (T2) REM: 28.67 (7.42) 0.05 (0.03) I: 34.86 (9.32) Insomnia at T1 and T3 (n ¼ 124) e Remission: I at T1 and NS at T3 (n ¼ 58) The full model was significant (p ¼ .003) and predicted 69.8% of cases. APSQ (T1) REM: 20.09 (8.65) 0.07 (0.03)* I: 24.37 (10.16) DBAS (T1) REM: 26.53 (8.27) 0.09 (0.04)** I: 27.35 (7.39) PSAS (T1) REM: 13.50 (4.28) 0.01 (0.04) I: 14.75 (4.84) SAMI (T1) REM: 19.40 (5.91) 0.04 (0.04) I: 21.40 (5.51) SRBQ (T1) REM: 31.12 (9.49) 0.04 (0.02)# I: 35.73 (9.57)

Odds ratio

95% CI

r2

1.11

1.04e1.18

.12e.18

0.89

0.83e0.95

1.00

0.92e1.08

1.04

0.96e1.12

1.04

1.00e1.09#

1.08

1.00e1.16

0.96

0.88e1.03

1.06

0.95e1.18

0.99

0.90e1.09

1.05

0.98e1.12

1.07

1.01e1.13

0.91

0.85e0.98

1.01

0.93e1.10

1.04

0.96e1.12

1.04

1.00e1.09

.13e.19

.10e.13

Note. The final solution is presented. CI ¼ confidence interval, I ¼ insomnia, M ¼ mean, REM ¼ remission, SD ¼ standard deviation. *Significant at the .05 level. **Significant at the .01 level. #p ¼ .06e.10. r2 refers to Cox & Snell and Nagelkerke R Square respectively.

about sleep, somatic arousal, selective attention and monitoring, and safety behaviours as predictors, was significant for all three time spans, which indicates that the model could distinguish between those with persistent insomnia and remission from insomnia. The model explained between .10 (Cox & Snell r2) and .19 (Nagelkerke r2) of the total variance in sleep status. APSQ made a significant contribution to the model in all three analyses, with odds ratios ranging between 1.07 and 1.11, meaning that a higher degree of sleep related worry at baseline was associated with persistence rather that remission from insomnia. DBAS was also a unique significant predictor, for the time spans T1eT2 and T1eT3, with odds ratios of .89 and .91 respectively. As for hypothesis 1, this prompted further analyses without APSQ and DBAS respectively, as multicollinearity was suspected between the two predictors. When testing a model without APSQ, SRBQ became a unique significant predictor for all time spans (odds ratios: 1.07, 1.07 & 1.06), rather than having a tendency to significance as for T1eT2 and T1eT3 with the original model. DBAS was no longer a significant predictor and the beta values remained negative, with odds ratios below 1. The model without APSQ explained between .07 (Cox & Snell r2) and .14 (Nagelkerke r2) of the total variance. With DBAS removed, APSQ had a tendency towards significance for T1eT2 (p ¼ .08; odds ratio: 1.05) and T2eT3 (p ¼ .10; odds ratio: 1.06) but no predictor was significant. The model without DBAS explained between .06 (Cox & Snell r2) and .17 (Nagelkerke r2) of the total variance. The theorised multicollinearity between APSQ and DBAS, that would be the cause of negative beta in logistic analyses, could not be confirmed for hypothesis 2. It should be noted that people with remission from insomnia scored only slightly less on dysfunctional beliefs

compared to people with persistent insomnia (see Table 3) thus the clinical implications of this particular result may be negligible. The original overall model, including all five predictors, explained relatively more variance compared to models without APSQ and DBAS respectively. When entering the five cognitive processes and anxiety, depression, pain, and headache simultaneously, across the three time spans, none of the psychiatric or medical conditions were significant predictors, i.e., discriminating between those with persistent insomnia and those with remission from insomnia. The cognitive processes that were significant in the unadjusted models above remained significant across the three time spans. The findings means partial support for our hypothesis, as higher baseline values of only two cognitive processes: worry and safety behaviours increased the likelihood of persistent insomnia rather than remission later on. Hypothesis 3: For People With Insomnia, a Diminishing of the Cognitive Processes Over Time Will Increase the Likelihood of Reporting Remission From Insomnia Rather Than Persistent Insomnia Later On. The descriptive statistics for the raw change scores of the five cognitive processes for the two insomnia groups (remission from insomnia and persistent insomnia) are displayed at Table 5. The overall model, with worry, dysfunctional beliefs about sleep, somatic arousal, selective attention and monitoring, and safety behaviours as predictors, was significant for all three time spans, which indicates that the model could distinguish between those with persistent insomnia and insomnia remission. The overall model explained between .16 (Cox & Snell r2) and .30 (Nagelkerke r2)

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

45

Table 5 Changes in cognitive processes over time between those with persistent insomnia and those with remission from insomnia: descriptive statistics and logistic regression analyses. Predictor

M (SD)

B (SE)

Insomnia at T1 and T2 (n ¼ 140) e Remission: Insomnia at T1 and NS at T2 (n ¼ 59) The full model was significant (p < .001) and predicted 74.4% of cases. APSQ (T1eT2) REM: 1.25 (9.00) 0.77 (0.36)* I: 1.72 (8.48) DBAS (T1eT2) REM: 2.44 (7.62) 0.22 (0.29) I: .50 (6.08) PSAS (T1eT2) REM: .54 (3.87) 0.05 (0.25) I: .23 (4.30) SAMI (T1eT2) REM: .63 (4.43) 0.65 (0.28)* I: .08 (5.17) SRBQ (T1eT2) REM: 2.69 (10.66) 0.64 (0.29)* I: .58 (8.31) Insomnia at T2 and T3 (n ¼ 80) e Remission: Insomnia at T2 and NS at T3 (n ¼ 16) The full model was significant (p ¼ .005) and predicted 85.4% of cases. APSQ (T2eT3) REM: 2.11 (5.20) 0.99 (0.65) I: .23 (9.26) DBAS (T2eT3) REM: .82 (7.54) 0.53 (0.61) I: .33 (6.63) PSAS (T2eT3) REM: .71 (3.29) 0.55 (0.47) I: .61 (4.29) SAMI (T2eT 3) REM: 1.45 (4.90) 1.49 (0.54)** I: 1.08 (5.42) SRBQ (T2eT3) REM: .21 (8.83) 0.05 (0.53) I: .76 (8.48) Insomnia at T1 and T3 (n ¼ 96) e Remission: Insomnia at T1 and NS at T3 (n ¼ 40) The full model was significant (p < .001) and predicted 75.0% of cases. APSQ (T1eT3) REM: 4.02 (7.52) 0.35 (0.43) I: .46 (8.90) DBAS (T1eT3) REM: 3.75 (6.69) 0.03 (0.44) I: .51 (6.20) PSAS (T1eT3) REM: .82 (3.62) 0.05 (0.33) I: .69 (4.30) SAMI (T1eT3) REM: 1.98 (4.60) 0.69 (0.32)* I: .52 (5.46) SRBQ (T1eT3) REM: 5.85 (9.43) 0.94 (0.37)** I: .28 (8.53)

Odds ratio

95% CI

r2

2.16

1.06e4.39

.16e.23

0.81

0.46e1.42

1.05

0.64e1.72

1.91

1.11e3.30

1.90

1.08e3.36

2.69

0.75e9.64

0.59

0.18e1.94

0.58

0.23e1.46

4.43

1.53e12.75

1.05

0.37e2.98

1.41

0.61e3.28

0.97

0.41e2.23

1.05

0.56e1.99

1.99

1.06e3.72

2.57

1.25e5.27

.16e.27

.21e.30

Note. The final solution is presented. CI ¼ confidence interval, I ¼ insomnia, M ¼ mean, REM ¼ remission, SD ¼ standard deviation. *Significant at the .05 level. **Significant at the .01 level. r2 refers to Cox & Snell and Nagelkerke R Square respectively.

of the variance. APSQ, SAMI and SRBQ made significant contributions to the model in T1eT2 with odds ratios 2.16, 1.91 and 1.90 respectively, meaning that a lowering over time of worry, selective attention and monitoring, and safety behaviours increased the likelihood of remission from insomnia. For T2eT3, only SAMI was a unique significant contributor to the model (odds ratio 4.43). SAMI and SRBQ made significant contributions to the model for T1eT3, with odds ratios 1.99 and 2.57 respectively. When entering the five cognitive processes (standardised residual change scores) and anxiety, depression, pain, and headache simultaneously, across the three time spans, none of the psychiatric or medical conditions were significant predictors, i.e., discriminating between those with persistent insomnia and those with remission from insomnia. The cognitive processes that were significant in the unadjusted models above remained significant across the three time spans. The findings provide partial support for the third hypothesis, as diminishing change scores of three out of five cognitive processes were significant predictors of sleep status later on, and only selective attention and monitoring was a consistent finding over all time spans. Discussion Models of insomnia have emphasised maintaining cognitive processes (Espie, 2002; Harvey, 2005; Lundh & Broman, 2000; Morin, 1993; Perlis et al., 1997). The overall aim was to investigate

if worry, dysfunctional beliefs, somatic arousal, selective attention and monitoring, and safety behaviours were associated with the persistence and remission of insomnia, by using a longitudinal design. This study is the first to investigate the five cognitive processes, simultaneously and prospectively, in the general population. The results partially support the overall hypothesis that cognitive processes are involved in persistence and remittance of insomnia. The associations between insomnia and the five cognitive processes were examined according to three specific hypotheses. The first hypothesis was that cognitive processes would discriminate between persistent insomnia and persistent normal sleep. The overall model was significant and four out of five processes, worry, somatic arousal, selective attention and monitoring, and safety behaviours, were unique significant predictors of sleep status. As the cognitive processes are theorised to thrive on concern about insomnia and the feared consequences of poor sleep, it was expected that people with insomnia would experience the cognitive processes to a larger degree than people without insomnia. The first hypothesis was thus supported and this is in line with theories of insomnia persistence that include one or more of the concepts worry, dysfunctional beliefs, somatic arousal, selective attention and monitoring, and safety behaviours (Espie, 2002; Harvey, 2005; Lundh & Broman, 2000; Morin, 1993; Perlis et al., 1997). The results are also consistent with previous studies where people with insomnia were found to worry more during the pre-sleep period (Harvey, 2000), and reported more somatic arousal and dysfunctional beliefs (Carney et al., 2010; Jansson & Linton, 2007) compared with people without

46

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

insomnia. Selective attention and monitoring and safety behaviours are more recent additions to theories of insomnia persistence (Harvey, 2002) and the current study is the first in the general population to show that selective attention and monitoring, and safety behaviours, in addition to the more established processes, were also unique predictors for persistent insomnia. Multicollinearity between APSQ and DBAS explained a contradictory finding regarding the direction of the relationship between dysfunctional beliefs and outcome in sleep status, and dysfunctional beliefs was a significant predictor if worry was removed from the model. As a consequence of this strong relationship between the concepts, we cannot separate the unique predictive contributions of the two variables in this study. Theoretically the great overlap in explained variance may be explained by the cognitive model of insomnia, where a relationship between dysfunctional beliefs and worry is proposed: dysfunctional beliefs exacerbate worry, which in turn lead to safety behaviours, which prevent disconfirmation of dysfunctional beliefs (Harvey, 2002). Looking at the face value of the questionnaires’ items, one might speculate if APSQ and DBAS-10 might tap into the very same constructs, even though they represent distinct concepts. For example, both questionnaires have items beginning with the statement “I worry.” which in itself might tap into repetitive negative thinking, and both questionnaires give examples of common fears of people with insomnia such as consequences for one’s health, which might tap into specific schemata. Obviously a statistical solution for a future study could be to remove any items that correlate too highly, in order to investigate the processes’ unique predictive power. A question on a more conceptual level is whether it is feasible to separate the two processes or not? Can there be sleep-related worry without something to fear for, in other words, dysfunctional beliefs about sleep? Can a person with insomnia have dysfunctional beliefs about sleep without also worrying about them? An area for future research would be to investigate the association between worry and dysfunctional beliefs, theoretically and psychometrically. The second hypothesis proposed that higher degrees of the cognitive processes would increase the likelihood of reporting persistent insomnia instead or remission from insomnia later on. The second hypothesis was partially supported as the overall model was significant, with worry and dysfunctional beliefs as unique significant predictors of sleep status. Again dysfunctional beliefs presented with contradictory direction of relationship to the outcome but this time multicollinearity with worry could not be confirmed by complimentary analyses. While investigating the suspected multicollinearity, it was found that in models without worry, safety behaviours at baseline became a unique predictor of persistent insomnia later. Conceptually, both worry and safety behaviours could be defined as avoidance behaviours, undertaken in efforts to regulate emotions (Campbell-Sills & Barlow, 2007). In this line of theory, safety behaviours could be divided into several subcategories of overt forms of avoidance, as in situational avoidance (e. g. not engage in activities that are believed to disturb sleep), and the use of safety signals (e. g. using alcohol to fall asleep) but also as covert avoidance, in the form of cognitive avoidance (e. g. distraction or thought suppression) (Helbig-Lang & Petermann, 2010). It may seem paradoxical that worry would be a form of (covert) avoidance when people who worry actively think about their concerns, but it has been found that worry has an immediate effect on somatic anxiety, which in turn may reinforce worry as a strategy (Sibrava & Borkovec, 2006). The abstract nature of worry prevents stronger emotional experiences associated with mental imagery, thus exposure to the fear is less likely to happen. The association between higher degrees of worry and safety behaviours with persistent insomnia could thus be understood as a higher risk

of persistence for people with higher degrees of overt and covert avoidance strategies. This emphasises the clinical importance of thorough investigations of which strategies people with insomnia use to cope with their sleep problems and sleep-related negative emotions. It also points towards the importance of taking sleeprelated worry seriously. According to the results in this study, people who are worried about their sleep are at greater risk of developing persistent insomnia. The third hypothesis proposed that insomnia remission would be associated with a larger decrease of the cognitive processes over time compared to persistent insomnia. This hypothesis was partially supported as the model was significant and a lowering of three cognitive processes: worry, selective attention and monitoring, and safety behaviours predicted sleep status. Selective attention and monitoring was a consistent finding for all three analyses, whilst safety behaviours were significant in two and worry in only one analysis. It should be noted that the group of people with remission from insomnia was very small (n ¼ 16) in the analysis where only selective attention and monitoring was a significant predictor, and this may have caused issues with low power. The finding that a decrease of selective attention and monitoring was associated with remission supports the Attention-IntentionEffort theory of insomnia (Espie et al., 2006). The theory proposes that attentional bias sets the stage for a vicious cycle of deliberate intention to fall asleep and efforts to control sleep, which disturbs the automatic and otherwise effortless transition into sleep. Safety behaviours as predictor fits in the same model, as safety behaviours could be conceptualised as efforts to control sleep. Looking beyond the insomnia literature, the cognitive theory of social anxiety emphasises the need to re-focus selective attention and stop using safety behaviours (Clark & Wells, 1995) and mediation analyses showed that changes in self-focused attention and safety behaviours mediated treatment outcomes (Hedman et al., 2013). It was surprising that neither dysfunctional beliefs nor somatic arousal were predictive of insomnia remission, as successful CBT-I has consistently been associated with a reduction of dysfunctional beliefs about sleep and cognitive and somatic arousal (for a review see Schwartz & Carney, 2012). This may reflect a difference between clinical samples and the general population and warrants further investigation in future studies. Also it should be noted that although we cannot know for sure if people whose insomnia remitted sought treatment and if so, what kind of treatment they received, it is unlikely that many of them would have received CBT-I as availability is low in Sweden. Thus an area for future research would be to investigate if and which processes act as mediators/ moderators in insomnia treatment studies, when all five processes are investigated simultaneously. It could be that how people think (worry); their schemas regarding sleep (dysfunctional beliefs) and somatic arousal are of less importance for remission than the use of dysfunctional adjustments, such as selective attention and safety behaviours. High degrees of worry, somatic arousal and dysfunctional beliefs may not be perpetuating persistent insomnia unless sleep-disrupting behaviour in terms of attentional focus and unhelpful strategies to promote sleep coexist. One potential future advantage of focussing on cognitive processes of insomnia is that it is possible to change them through targeted treatment. A lowering of dysfunctional beliefs and arousal is associated with symptom improvements after CBT-I (Schwartz & Carney, 2012) and pilot studies have shown using a cognitive treatment for insomnia has been associated with a reduction of worry, dysfunctional beliefs, selective attention and monitoring, and safety behaviours after treatment (Harvey, Sharpley, Ree, Stinson, & Clark, 2007; Norell-Clarke, Nyander, & Jansson-Fröjmark, 2011). Future studies should investigate these and other treatments further in randomised controlled studies.

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

There are several strengths with the current study. The concepts of worry, dysfunctional beliefs, somatic arousal, selective attention and monitoring, and safety behaviours are apparent in several theories of insomnia but had never been studied together, prospectively, in the general population before. The three measuring points: baseline, six month follow-up and 18 month follow-up, allowed for the hypotheses to be tested on three different longitudinal time spans, which strengthens the reliability of the results. Other strengths with the study are that other sleep disorders were assessed, and medical health and anxiety and depression were controlled for. Although the study has several strengths, the study is not without limitations that should be kept in mind when interpreting the results. The moderate response rate (47.1% at T1) is problematic, as the attrition analysis showed that responders were older than non-responders. Given that our aim was to investigate associations between variables rather than estimates of a population a low response rate may not have a large impact on the results (Curtin, Presser, & Singer, 2000; Keeter, Kennedy, Dimock, Best, & Craighill, 2006). It should be noted that our sample was similar to the Swedish population on demographic variables, including insomnia prevalence. Another issue is that a proxy for insomnia diagnosis was used, thus we cannot know for sure if the people classified as insomnia sufferers would be diagnosed as such by health professionals, although the high concordance between our definition and an established insomnia measure (ISI) supports the validity of our definition (Jansson-Fröjmark et al., 2012). Also, it cannot be determined if persistent insomnia really captured one long insomnia episode or several episodes during the 18 months. The equivalent concern regards the participants classified as having persistent normal sleep. Further, it is also unknown if the people classified as having insomnia are similar to help seeking individuals in clinical settings and if the results can be generalised to clinical reality. Previous population studies have shown that many people with reported sleep problems never seek help (Ancoli-Israel & Roth, 1999) and the nature of their problems is therefore largely unknown. Another possible concern with the study is that shorter versions were used for the questionnaires DBAS, SAMI and SRBQ. DBAS had been validated previously (Espie et al., 2000) but the others were new as of the PIPPI project. However, the validity and reliability of the shorter versions were investigated and it was concluded that they measure the underlying constructs to a satisfactory degree (Jansson-Fröjmark et al., 2012). The scale range of APSQ had been slightly altered (from 1 to 10 in the original version to 1e5) which could mean that it would be difficult to know whether the results from this study are comparable with studies using the original version. It should however be noted that all the items remained the same and that the current study’s version has been evaluated and found to have sound psychometric properties regarding discriminant validity, convergent validity and high correlations with insomnia daytime and night-time symptoms (Jansson-Fröjmark et al., 2011). A limitation with the prospective method is that although a timeline can be established, and it can be confirmed that a lowering of worry, safety behaviours, selective attention and monitoring, and dysfunctional beliefs was observed at the same time point as remission of insomnia, a causal link cannot be made. Possible third factors could have effects on both insomnia severity and the processes, or maybe the processes are products of insomnia and would reduce as insomnia remitted. An area for future research could be to investigate how the cognitive processes relate to insomnia through experimental studies and treatment studies. Concerning the individual suffering and the high costs for society it is vital to understand the processes that drive persistent insomnia. The results suggest that high degrees of cognitive processes may be maintaining persistent insomnia, especially sleep

47

related worry, and that a lowering of selective attention and monitoring, and safety behaviours are associated with remission, over time. This creates interesting openings for further research into mediating factors of insomnia. References American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: American Psychiatric Association. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing. Ancoli-Israel, S., & Roth, T. (1999). Characteristics of insomnia in the United States: results of the 1991 National Sleep Foundation Survey. SLEEP, 22, 347e353. Bjelland, I., Dahl, A. A., Tangen Haug, T., & Neckelmann, D. (2002). The validity of the Hospital Anxiety and Depression Scale: an updated literature review. Journal of Psychosomatic Research, 52, 69e77. Breslau, N., Roth, T., Rosenthal, L., & Andreski, P. (1996). Sleep disturbance and psychiatric disorders: a longitudinal epidemiological study of young adults. Biological Psychiatry, 39, 411e418. Buysse, D. J., Germain, A., Hall, M. L., Monk, T. H., & Nofzinger, E. A. (2011). A neurobiological model of insomnia. Drug Discovery Today: Disease Models, 8(4), 129e137. Campbell-Sills, L., & Barlow, D. H. (2007). Incorporating emotion regulation into conceptualizations and treatments of anxiety and mood disorders. In J. J. Gross (Ed.), Handbook of emotion regulation. New York: The Guilford Press. Carney, C. E., Edinger, J. D., Morin, C. M., Manber, R., Rybarczyk, B., Stepanski, E. J., et al. (2010). Examining maladaptive beliefs about sleep across insomnia patient groups. Journal of Psychosomatic Research, 68(1), 57e65. Carney, C. E., & Waters, W. F. (2006). Effects of a structured problem-solving procedure on pre-sleep cognitive arousal in college students with insomnia. Behavioral Sleep Medicine, 4(1), 13e28. Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier (Eds.), Social phobia: Diagnosis, assessment, and treatment (pp. 69e93). New York, NY, US: Guilford Press. Cronbach, L. J., & Furby, L. (1970). How we should measure “change”: or should we? Psychological Bulletin, 74, 68e80. Curtin, R., Presser, S., & Singer, E. (2000). The effects on response rate changes on the index of consumer sentiment. Public Opinion Quarterly, 64, 413e428. Daley, M., LeBlanc, M., Grégoire, J. P., & Savard, J. (2009). The economic burden of insomnia: direct and indirect costs for individuals with insomnia syndrome, insomnia symptoms, and good sleepers. SLEEP, 32(1), 55e67. Drummond, S. P. A., Smith, M. T., Orff, H. J., Chengazi, V., & Perlis, M. L. (2004). Functional imaging of the sleeping brain: review of findings and implications for the study of insomnia. Sleep Medicine Reviews, 8(3), 227e242. Edinger, J. D., Bonnet, M. H., Bootzin, R. R., Doghramji, K., Dorsey, C. M., Espie, C. A., et al. (2004). Derivation of research diagnostic criteria for insomnia: report of an American Academy of Sleep Medicine Work Group. Sleep, 27(8), 1567e1596. Edwards, P., Roberts, I., Clarke, M., DiGuiseppi, C., Pratap, S., Wentz, R., et al. (2007). Methods to increase response rates to postal questionnaires. Cochrane Database of Systematic Reviews, 2. Ellis, J. G., Gehrman, P., Espie, C. A., Riemann, D., & Perlis, M. L. (2012). Acute insomnia: current conceptualizations and future directions. Sleep Medicine Reviews, 16(1), 5e14. Espie, C. A. (2002). Insomnia: conceptual issues in the development, persistence, and treatment of sleep disorder in adults. Annual Review of Psychology, 53, 215e 243. Espie, C. A., Broomfield, N. M., MacMahon, K. M. A., Macphee, L. M., & Taylor, L. M. (2006). The attentioneintentioneeffort pathway in the development of psychophysiologic insomnia: a theoretical review. Sleep Medicine Reviews, 10(4), 215e245. Espie, C. A., Inglis, S. J., Harvey, L., & Tessier, S. (2000). Insomniacs’ attributions: psychometric properties of the Dysfunctional Beliefs and Attitudes about Sleep Scale and the Sleep Disturbance Questionnaire. Journal of Psychosomatic Research, 48(2), 141e148. Field, A. (2009). Discovering statistics using SPSS (3rd ed.). London: SAGE Publications. Ford, D. E., & Kamerow, D. B. (1989). Epidemiologic study of sleep disturbances and psychiatric disorders. An opportunity for prevention? Journal of the American Medical Association, 262(11), 1479e1484. Gross, R. T., & Borkovec, T. D. (1982). Effects of a cognitive intrusion manipulation on the sleep-onset latency of good sleepers. Behavior Therapy, 13(1), 112e116. Harvey, A. G. (2000). Pre-sleep cognitive activity: a comparison of sleep-onset insomniacs and good sleepers. British Journal of Clinical Psychology, 39(3), 275e 286. Harvey, A. G. (2002). A cognitive model of insomnia. Behaviour Research and Therapy, 40, 869e893. Harvey, A. G. (2005). A cognitive theory and therapy for chronic insomnia. Journal of Cognitive Psychotherapy, 19(1), 41e59. Harvey, A. G., Sharpley, A. L., Ree, M. J., Stinson, K., & Clark, D. M. (2007). An open trial of cognitive therapy for chronic insomnia. Behaviour Research and Therapy, 45(10), 2491e2501. Hedman, E., Mörtberg, E., Hesser, H., Clark, D. M., Lekander, M., Andersson, E., et al. (2013). Mediators in psychological treatment of social anxiety disorder:

48

A. Norell-Clarke et al. / Behaviour Research and Therapy 54 (2014) 38e48

Individual cognitive therapy compared to cognitive behavioral group therapy. Behaviour Research and Therapy, 51(10), 696e705. Helbig-Lang, S., & Petermann, F. (2010). Tolerate or eliminate? A systematic review on the effects of safety behavior across anxiety disorders. Clinical Psychology: Science and Practice, 17(3), 218e233. Herrmann, C. (1997). International experiences with the Hospital Anxiety and Depression Scale-a review of validation data and clinical results. Journal of Psychosomatic Research, 42(1), 17e41. Jansson, M., & Linton, S. J. (2007). Psychological mechanisms in the maintenance of insomnia: arousal, distress, and sleep-related beliefs. Behaviour Research and Therapy, 45, 511e521. Jansson-Fröjmark, M., Harvey, A. G., Lundh, L.-G., Norell-Clarke, A., & Linton, S. J. (2011). Psychometric properties of an insomnia-specific measure of worry: the Anxiety and Preoccupation about Sleep Questionnaire. Cognitive Behaviour Therapy, 40(1), 65e76. Jansson-Fröjmark, M., Harvey, A. G., Norell-Clarke, A., & Linton, S. J. (2012). Associations between psychological factors and nighttime/daytime symtomatology in insomnia. Cognitive Behaviour Therapy, 41(4), 273e287. Jansson-Fröjmark, M., Lundquist, D., Lundquist, N., & Linton, S. J. (2008). How is persistent insomnia maintained? A prospective study on 50e60 years old adults in the general population. British Journal of Health Psychology, 13, 121e 133. Jansson-Fröjmark, M., & Norell-Clarke, A. (2012). Psychometric properties of the Pre-Sleep Arousal Scale in a large community sample. Journal of Psychosomatic Research, 72(2), 103e110. Jansson-Fröjmark, M., Norell-Clarke, A., & Linton, S. J. (2014). The role of anxiety, depression, pain and headache on the persistence, remission, and incidence of insomnia: A longitudinal, community study (unpublished) (in preparation). Keeter, S., Kennedy, C., Dimock, M., Best, J., & Craighill, P. (2006). Gauging the impact of growing nonresponse on estimates from a national RDD telephone survey. Public Opinion Quarterly, 70, 759e779. LeBlanc, M., Mérette, C., Savard, J., Ivers, H., Baillargeon, L., & Morin, C. M. (2009). Incidence and risk factors of insomnia in a population-based sample. Sleep, 32(8), 1027e1037. Léger, D., Massuel, M.-A., Metlaine, A., &, The SISYPHE Study Group. (2006). Professional correlates of insomnia. SLEEP, 29(1), 171e178. Lichstein, K. L., Durrence, H. H., Taylor, D. J., Bush, A. J., & Riedel, B. W. (2003). Quantitative criteria for insomnia. Behaviour Research and Therapy, 41, 427e445. Linton, S. J., & Bryngelsson, I.-L. (2000). Insomnia and its relationship to work and health in a working-age population. Journal of Occupational Rehabilitation, 10(2), 169e183. Lundh, L.-G., & Broman, J.-E. (2000). Insomnia as an interaction between sleepinterfering and sleep-interpreting processes. Journal of Psychosomatic Research, 49, 299e310. Marchetti, L. M., Biello, S. M., Broomfield, N. M., Macmahon, K. M. A., & Espie, C. A. (2006). Who is pre-occupied with sleep? A comparison of attention bias in people with psychophysiological insomnia, delayed sleep phase syndrome and good sleepers using the induced change blindness paradigm. Journal of Sleep Research, 15(2), 212e221. Morin, C. M. (1993). Insomnia: Psychological assessment and management. New York: The Guilford Press. Morin, C. M., Bélanger, L., LeBlanc, M., Ivers, H., Savard, J., Espie, C. A., et al. (2009). The natural history of insomnia: a population-based 3-year longitudinal study. Archives of Internal Medicine, 169(5), 447e453. Morin, C. M., Belleville, G., Bélanger, L., & Ivers, H. (2011). The insomnia severity index: psychometric indicators to detect insomnia cases and evaluate treatment response. SLEEP, 34(5), 601e608. Morin, C. M., LeBlanc, M., Daley, M., Gregoire, J. P., & Mérette, C. (2006). Epidemiology of insomnia: prevalence, self-help treatments, consultations, and determinants of help-seeking behaviors. Sleep Medicine, 7, 123e130. Morin, C. M., Stone, J., Trinkle, D., Mercer, J., & Remsberg, S. (1993). Dysfunctional beliefs and attitudes about sleep among older adults with and without insomnia complaints. Psychology and Aging, 8, 463e467.

Nicassio, P. M., Mendlowitz, D. R., Fussel, J. J., & Petras, L. (1985). The phenomenology of the pre-sleep state: the development of the pre-sleep arousal scale. Behaviour Research and Therapy, 23(3), 263e271. Norell-Clarke, A., Nyander, E.-L., & Jansson-Fröjmark, M. (2011). Sleepless in Sweden: a single subject study of effects of cognitive therapy for insomnia on three adolescents. Behavioural and Cognitive Psychotherapy, 39, 367e374. Ohayon, M. M. (2002). Epidemiology of insomnia: questions of cause and effect. Sleep Medicine Reviews, 6, 97e111. Ohayon, M. M., & Reynolds, C. F., III (2009). Epidemiological and clinical relevance of insomnia diagnosis algorithms according to the DSM-IV and the International Classification of Sleep Disorders (ICSD). Sleep Medicine, 10(9), 952e960. Perlis, M. L., Giles, D. E., Mendelson, W. B., Bootzin, R. R., & Wyatt, J. K. (1997). Psychophysiological insomnia: the behavioural model and a neurocognitive perspective. Journal of Sleep Research, 6, 179e188. Ree, M. J., & Harvey, A. G. (2004). Investigating safety behaviours in insomnia: the development of the Sleep-related Behaviours Questionnaire (SRBQ). Behaviour Change, 21(1), 26e36. Riemann, D., Spiegelhalder, K., Feige, B., Voderholzer, U., Berger, M., Perlis, M. L., et al. (2010). The hyperarousal model of insomnia: a review of the concept and its evidence. Sleep Medicine Reviews, 14(1), 19e31. Salkovskis, P. M. (1991). The importance of behaviour in the maintenance of anxiety and panic: a cognitive account. Behavioural Psychotherapy, 19(1), 6e19. Schwartz, D. R., & Carney, C. E. (2012). Mediators of cognitive-behavioral therapy for insomnia: a review of randomized controlled trials and secondary analysis studies. Clinical Psychology Review, 32(7), 664e675. Semler, C. N., & Harvey, A. G. (2004). Monitoring for sleep-related threat: a pilot study of the Sleep Associated Monitoring Index (SAMI). Psychosomatic Medicine, 66, 242e250. Sibrava, N. J., & Borkovec, T. D. (2006). The cognitive avoidance theory of worry. In G. C. L. D. A. Wells (Ed.), Worry and its psychological disorders: Theory, assessment and treatment (pp. 239e256). Hoboken, NJ, US: Wiley Publishing. Singareddy, R., Vgontzas, A. N., Fernandez-Mendoza, J., Liao, D., Calhoun, S., Shaffer, M. L., et al. (2012). Risk factors for incident chronic insomnia: a general population prospective study. Sleep Medicine, 13(4), 346e353. Spiegelhalder, K., Espie, C. A., & Riemann, D. (2009). Is sleep-related attentional bias due to sleepiness or sleeplessness? Cognition & Emotion, 23(3), 541e550. Spoormaker, V. I., Verbeek, I., van den Bout, J., & Klip, C. (2005). Initial Validation of the SLEEP-50 Questionnaire. Behavioral Sleep Medicine, 3(4), 227e246. Steketee, G., & Chambless, D. L. (1992). Methodological issues in prediction of treatment outcome. Clinical Psychology Review, 12(4), 387e400. Tang, N. K. Y., & Harvey, A. G. (2004). Correcting distorted perception of sleep in insomnia: a novel behavioural experiment? Behaviour Research and Therapy, 42, 27e39. Tang, N. K. Y., Schmidt, D. A., & Harvey, A. G. (2007). Sleeping with the enemy: clock monitoring in the maintenance of insomnia. Journal of Behavior Therapy and Experimental Psychiatry, 38, 40e55. Vollrath, M., Wicki, W., & Angst, J. (1989). The Zurich study. European Archives of Psychiatry and Neurological Sciences, 239(2), 113e124. Woodley, J., & Smith, S. (2006). Safety behaviors and dysfunctional beliefs about sleep: testing a cognitive model of the maintenance of insomnia. Journal of Psychosomatic Research, 60, 551e557. Woods, H. C., Marchetti, L. M., Biello, S. M., & Espie, C. A. (2009). The clock as a focus of selective attention in those with primary insomnia: an experimental study using a modified Posner paradigm. Behaviour Research and Therapy, 47(3), 231e 236. Woods, H. C., Scheepers, C., Ross, K. A., Espie, C. A., & Biello, S. M. (2013). What are you looking at? Moving toward an attentional timeline in insomnia: a novel semantic eye tracking study. Sleep, 36(10), 1491e1499. Zhang, J., Lam, S. P., Li, S. X., Yu, M. W. M., Li, A. M., Ma, R. C. W., et al. (2012). Longterm outcomes and predictors of chronic insomnia: a prospective study in Hong Kong Chinese adults. Sleep Medicine, 13(5), 455e462. Zigmond, A. S., & Snaith, R. P. (1983). The Hospital Anxiety and Depression Scale. Acta Psychiatrica Scandinavica, 67(6), 361e370.