Conceptual overlap of negative thought processes in insomnia: A focus on catastrophizing, worry, and rumination in a student sample

ABSTRACT

Objectives

The association and overlap between different forms of negative thought processes in insomnia is largely unknown. The purpose of the current investigation was to examine conceptual overlap between three insomnia-specific negative thought processes; catastrophizing, worry, and rumination, identify the underlying factors, and explore their associations with insomnia symptoms.

Methods

A total of 360 students completed three insomnia-related negative thought process scales (Catastrophic Thoughts about Insomnia Scale, Anxiety and Preoccupation about Sleep Questionnaire, Daytime Insomnia Symptom Response Scale) and two insomnia symptoms measures (the Insomnia Severity Index and Sleep Condition Indicator).

Results

The three scales and their subscales displayed acceptable reliabilities. Further, confirmatory factor analysis was supportive of the notion of catastrophizing, worry, and rumination measures as distinct. The catastrophizing and worry constructs were significantly associated with insomnia symptoms, but the rumination factor was not.

Conclusions

The findings indicate that catastrophizing, worry, and rumination might be viewed as distinct constructs. Although more research is warranted on the topic of conceptual overlap, the current results might have implications for the development of models of insomnia, clinical research, and practice.

Introduction

Insomnia disorder is characterized by sleep initiation difficulties at bedtime, frequent or extended awakenings, or early-morning awakenings with an inability to return to sleep (American Psychiatric Association, 2013). To meet the full criteria for insomnia disorder, sleep difficulties must occur despite adequate opportunity for sleep and be present for three nights or more per week and lasts for more than 3 months, and be associated with significant distress or impairment in functioning, such as within the individual’s working or personal life. It is estimated that approximately 10% fulfill the criteria for insomnia disorder at a given time-point (C. M. Morin et al., 2015). Insomnia disorder is frequently associated with medical as well as mental disorders and also a risk factor for several other disorders, such as depression. Insomnia disorder is also associated with significant direct and indirect costs (C. M. Morin et al., 2015).

Extant psychological models of insomnia often propose that negative thought processes are involved in the maintenance of insomnia (Borkovec, 1982; Espie, 2002; Harvey, 2002; Lundh & Broman, 2000; C. M. Morin, 1993). A common notion in these conceptualizations is that intrusive thinking prior to sleep and during subsequent awakenings might trigger and intensify the sleep difficulties associated with insomnia. Some of these models also underscore that negative thought processes might maintain daytime problems, such as fatigue (Harvey, 2002; Lundh & Broman, 2000). There are at least three negative thought processes that have been underscored in the scientific literature, namely: catastrophizing, worry, and rumination (Hiller et al., 2015). Catastrophizing has been suggested to involve “dwelling on the worst possible outcomes of any situation in which there is a possibility for an unpleasant outcome. The person overemphasizes the probability of this catastrophic outcome and usually exaggerates the possible consequences of its occurrence” (Beck et al., 2005). Several researchers have highlighted that patients with insomnia often catastrophize about the consequences of sleep loss and its negative impact on health and functioning, which could result in disrupting sleep homeostasis, circadian timing, and sleep perception (Espie, 2002; C. M. Morin, 1993; Perlis et al., 2000). Worry has been defined as “a chain of thoughts and images, negatively affect-laden and relatively uncontrollable” (Borkovec et al., 1983). Several studies have demonstrated that patients with insomnia complain that they cannot get to sleep due to excessive and uncontrollable worry during the pre-sleep period (Borkovec, 1982; Espie et al., 1989; Harvey, 2000; Lichstein & Rosenthal, 1980). Rumination has been used to refer to the repetitive focusing on the “causes, meanings and consequences” of one’s feelings and symptoms (Nolen-Hoeksema, 1991). Research on rumination has shown that poor sleepers are likely to ruminate about symptom-focused content (e.g., fatigue and low mood), and among patients with insomnia that high rumination is associated with longer wakefulness after sleep onset and poorer sleep quality (Carney et al., 2006, 2010). Outside of the insomnia field, there are indications that catastrophizing, worry, and rumination are independent constructs (Pike et al., 2021).

Although three different negative thought processes have been proposed to be involved in insomnia, conceptual questions remain. First, it is unclear whether these three forms of negative thought processes are conceptually different as proposed by Hiller et al. (2015), i.e., they differ by possessing differing thought characteristics. Based on their definitions, which could be argued to be similar, another perspective would be to consider them as being conceptually similar, e.g., in terms of the appraisals and strategies (Harvey & Greenall, 2003; Watkins et al., 2005). A common notion outside the insomnia field is also that worry and rumination are critical thought processes, particularly in relation to generalized anxiety and depression, and that catastrophizing can be conceived as an intense form of worry (Fresco et al., 2002; Nolen-Hoeksema et al., 2008; Yang et al., 2014). The insomnia research literature is scarce on this conceptual topic. One exception is a clinically oriented study by Carney and colleagues (Carney et al., 2010), in which 210 individuals with insomnia completed two generic worry and rumination questionnaires [i.e., the Penn State Worry Questionnaire (PSWQ) and the Symptom-focused Rumination Scale (SYM)]. With correlations and an exploratory factor analysis of the PSWQ and SYM items, findings indicated a strong correlation between the two scales (r = .56) and a three-factor solution (i.e., PSWQ high worry items, SYM items, and reverse-scored PSWQ items). Second, it is also uncertain how the three negative thought processes relate to insomnia symptoms. Yet again the study by Carney and colleagues (Carney et al., 2010) is an exception, since at least two forms of negative thought processes were assessed in the same investigation in combination with measuring insomnia symptoms (in this case using sleep diaries). Relative to those with low rumination, the participants with high rumination displayed lower sleep efficiency, worse sleep quality, and longer nightly awakenings. However, rumination was not significantly associated with sleep initiation difficulties, and worry was not significantly related to any of the sleep diary measures.

Beyond the study by Carney et al. (2010), there is also a large community study, in which measures of insomnia-specific negative thought processes (i.e., catastrophizing and worry) and insomnia symptoms were used (Jansson-Fröjmark et al., 2020). Catastrophizing was assessed with the Insomnia Catastrophizing Scale (Jansson-Fröjmark et al., 2020), worry with the Anxiety and Preoccupation about Sleep Questionnaire (Jansson-Fröjmark et al., 2011), and insomnia symptoms with one-item questions focusing on typical sleep diary variables, such as sleep onset latency. In the study, catastrophizing and worry were highly correlated (ρ = 0.70–0.73). Both catastrophizing and worry were significantly associated at a similar level with several nighttime symptoms (η = 0.20–0.38) and with daytime impairment (ρ = 0.53–0.60). Taken together, the study’s findings imply that insomnia-related catastrophizing and worry are closely related and that both are associated with insomnia symptoms.

Both cited studies have limitations in terms of providing valid answers on how the three negative thought processes are inter-related and whether they are associated with insomnia symptoms. First, in the study by Carney and colleagues (Carney et al., 2010), only generic measures, not specific to insomnia, was used to assess worry and rumination. Generic and insomnia-specific negative thought processes are likely to differ in terms of specificity of content (e.g., worrying in general or worrying about poor sleep) and therefore also impact the association with specific symptoms. For example, Lancee et al. (2017) showed that nighttime sleep-related worry is involved in the maintenance of insomnia, whereas effects of trait repetitive thinking is more benign. Second, both studies rely only on exploratory factor analysis in the examination of latent negative thought processes. Confirmatory factor analysis is viewed as an important addition when exploring latent factors, primarily because it builds on hypothesis testing based on theory and previous research (Flora & Flake, 2017; Widaman, 2012). Third, no extant investigation has explored the inter-relationships between all three of the negative thought processes, thus hampering progress in how to view the negative thought processes as a whole. Thus, several questions remain unanswered concerning the inter-relationships between the three negative thought processes and their association with insomnia symptoms.

To validate existing constructs, such as negative thought processes, research must show that the concepts are conceptually and empirically distinct from related constructs (Shaffer et al., 2016). The route to exploring conceptual distinctness is through a literature review, in which theoretically related constructs are identified and then examined empirically. The most direct way to establish the empirical distinctness of a construct is to assess its discriminant validity, which refers to the extent to which measures of theoretically distinct constructs are unrelated empirically to one another (Campbell & Fiske, 1959). Demonstrating the discriminant validity of measures of newly developed constructs is a vital step in the process of construct validation (Harter & Schmidt, 2008), since this step can expose whether a new construct is empirically redundant with existing constructs. One of the most well-known and validated methods for demonstrating discriminant validity is confirmatory factor analysis (CFA) (Bagozzi et al., 1991).

As stated above, there may be substantial conceptual overlap between insomnia-related catastrophizing, worry, and rumination. As a result, the distinct value of the three negative thought processes as predictors of insomnia and associated outcomes is unclear. This lack of clarity hinders progress in three domains. First, without empirical research on the distinctness of the three negative thought processes, the continued refinement and development of cognitive models of insomnia becomes difficult. Second, research progress is also hampered since knowledge is lacking on how to select appropriate measures for the latent constructs at hand. Relevant research could, for example, clarify the influences of negative thought processes on insomnia symptoms and their potential role as mediators and moderators of cognitive behavioral therapy for insomnia (van Straten et al., 2018). Third, greater clarity about the distinctness of the three negative thought constructs has the potential to provide a basis for the design of more effective interventions for example, by focusing on the process with the greatest impact on insomnia disorder. Research focusing on the overlap and distinctness of cognitive processes in insomnia was recently called for in a review, since 20 cognitive factors were identified in cognitive models and very limited research on their inter-relationships (Tang et al., 2023).

The aim of the current study was therefore to examine the degree of conceptual overlap between the three insomnia-specific negative thought processes (i.e., catastrophizing, worry, and rumination), identify the underlying factors, and explore their validity (in relation to degree of insomnia symptoms). In line with Hiller et al. (2015), the hypothesis was that catastrophizing, worry, and rumination measures would be significantly related but also distinct. A hypothesis concerning the three negative thought processes’ association with insomnia symptoms was not stated a priori.

Methods

Participants

The participants were recruited during 2019 through purposive sampling among registered students at the Department of Psychology, Stockholm University, Sweden. University students were chosen since the rate of insomnia is relatively high at 18.5% (Jiang et al., 2015), thus selecting a group that is likely to display substantial variations in sleep and insomnia symptoms. Recruiting a heterogeneous sample is recommended when exploring the factor structure of a measure, since such a group is likely to display a range of scores within the characteristics of interest (Gaskin et al., 2017). The following student groups were asked to participate through invitations via group-based, anonymized e-mail adresses: those from the Master programme to become clinical psychologists during the first three years (i.e., terms 1–6) and those from undergraduate courses (Psychology I and II). Experiences of study participation was at the time of the data collection a mandatory part of the courses and programs at the department. In this study, each student received one hour course credit for their participation. The study has been ethically reviewed and approved by the Regional Ethical Board in Stockholm, Sweden. Informed consent was obtained from all participants included in the investigation.

In total, 360 students participated in the study. On average, the participants were 27.5 years old (SD = 7.9), and 72.9% were women. Among the participants, 13.6% reported a diagnosed mental disorder and 9.4% a diagnosed physical illness. The three most prevalent mental disorders that were reported by the students (with number of reports in parenthesis) were anxiety disorders (21), depression (20), and attention deficit hyperactivity disorder or attention deficit disorder (8). The three most common physical illnesses that were stated (with a number of reports in parenthesis) were rheumatism (5), asthma (3), and diabetes (3). Based on the two questions asking participants to report on diagnosed mental disorders and mental illnesses, six students also reported a diagnosed sleep disorder (insomnia disorder: n = 2, narcolepsy: n = 1, hypersomnia: n = 1, sleep apnea: n = 1, and restless legs syndrome: n = 1).

Procedure

Potential participants were sent an e-mail which informed about the possibility to participate in a survey concerning sleep and psychological processes. The participants were also informed that their participation was voluntary, that their survey responses would be used for research purposes, and that they could discontinue their participation at any time. Informed consent was obtained from all individual participants included in the study. A survey on participant characteristics and five self-reported scales (see below) was published online at a unique web platform during four weeks. On the first page, the participants were informed about the purpose of the study and ethical considerations. On the subsequent webpages, the participants completed the measures. The survey took approximately 20 min to complete.

Measures

All the participants were asked to report their age in years and their gender (man or woman). The participants were also requested to state whether or not they had been diagnosed with a mental disorder (yes or no) or a physical illness (yes or no) in the health care system. If the participant confirmed a mental disorder or a physical illness, the individual was asked to report on the type of disorder or illness.

Five self-report scales were used: the Catastrophic Thoughts about Insomnia Scale (CTIS), the Anxiety and Preoccupation about Sleep Questionnaire (APSQ), the Daytime Insomnia Symptom Response Scale (DISRS), the Insomnia Severity Index (ISI), and the Sleep Condition Indicator (SCI). Using the back-translation method, the CITS and DISRS were first translated into Swedish and then back-translated into English by two persons with good skills in both Swedish and English (Streiner et al., 2015). We relied on two independent translators, one translating the items from English to Swedish, and the other translating the items from Swedish to English. The APSQ, ISI, and SCI have previously been translated into Swedish using the back-translation method.

The CTIS (Tan et al., 2017) was employed to assess insomnia-related catastrophizing. The response alternatives for the 18 items are from 0 (strongly disagree) to 6 (strongly agree), with 3 (neutral) in between. The score range is thus 0–108, with a higher score reflecting higher catastrophizing. Tan et al. (2017) report that the CTIS displays a three-factor solution (i.e., rumination, magnification, and helplessness) and an overall catastrophizing factor, with satisfactory internal consistencies for the subscales and the total scale (α = .84–.94). The CTIS has also been shown to display criterion validity with sleep measures and construct validity with other psychological constructs. Examples of CTIS items are “When I go to bed and can’t fall asleep, I think of other times that I couldn’t fall asleep” (Rumination factor), “As I’m trying to fall asleep, I think: “If it takes me too long to fall asleep, it would be terrible” (Magnification factor), and “There is nothing that I can do to improve my sleep” (Helplessness factor).

To assess insomnia-specific worry, the APSQ was used (Jansson-Fröjmark et al., 2011; Tang & Harvey, 2004). The APSQ contains 10 items and determines insomnia-specific worry in two core domains: worries about the consequences of poor sleep (e.g., “I worry about the amount of sleep I am going to get every night”) and worries about the uncontrollability of sleep (e.g., I worry about my loss of control over sleep). The response alternatives for each of the items are from 1 (strongly disagree) to 10 (strongly agree). The score range is thus 10–100, with a higher score reflecting elevated worry. In the original study on the psychometric properties of the APSQ, two subscales were identified with acceptable internal reliabilities (α = .86–.91), and there was support for the scale’s discriminant and convergent validity.

As an index of insomnia-specific rumination, the DISRS was used (Carney et al., 2013). The DISRS contains 20 items, and the response alternatives are from 1 (almost never) to 4 (almost always) (total score range 20–80). A higher score on the DISRS represents elevated rumination. Previous research has shown that the DISRS has high internal consistency (α = .93) and consists of three factors (i.e., Cognitive/Motivation, Negative state, and Tired). Also, the DISRS has been shown to display criterion and discriminant validity with insomnia severity and fatigue measures. Examples of DISRS items are “Think about how hard it is to concentrate” (Cognitive/motivation factor), “Think: I won’t be able to do work because I feel so bad” (Negative state factor), and “Think about how you don’t have the energy to get through the day” (Tired factor).

Two self-report measures were used to assess insomnia symptoms. First, the ISI was employed to assess participants’ perception of insomnia severity (Bastien et al., 2001; C. M. Morin et al., 2011). The seven-item ISI questionnaire is rated on a 5-point scale (0–4) with a total score of 0–28 and assesses both night- and daytime symptoms (e.g., difficulty falling asleep, function interference, and worry/distress). A higher score on the ISI reflects an elevated level of insomnia severity. The ISI has demonstrated adequate internal consistency (α = .91) and is viewed as a one-factor measure. Further, the ISI has been shown to display discriminant and convergent validity. Second, the SCI was used to determine insomnia symptoms (Espie et al., 2014). The SCI consists of eight items, reflecting night- and daytime symptoms (e.g., concerns about getting to sleep, daytime performance, and extent troubled by poor sleep). Each item is scored on a 0–5 scale (total score range: 0–32). A higher score is indicative of less insomnia symptoms. The SCI has displayed satisfactory internal consistency (α= .86) and is used as a one-factor scale. The SCI has also been shown to display criterion validity with other insomnia measures (e.g., the ISI) and convergent validity with other health measures.

Statistical analysis

After data cleaning, operationalization of raw data, and production of descriptive statistics, the students were divided into two groups based on their ISI score, 11 points or more and less than 11 points (C. M. Morin et al., 2011). The cutoff at 10 points on the ISI has been shown to detect insomnia cases at an acceptable level (86.1% sensitivity and 87.7% specificity) in a community sample. The statistical analysis continued thereafter with the following three steps, all executed with (R Core Team, R, 2021). First, the five self-reported scales of interest for the study were assessed for normality; CTIS, APSQ, and DISRS for multivariate normality using Royston’s MVN test in the MVN package in R (Korkmaz et al., 2014), whereas the ISI and SCI scales were assessed for normality using the Shapiro–Wilk normality test.

Second, a Confirmatory Factor Analysis (CFA) was used at the subscale level (n = 8) to examine the factor structure of a three-factor solution [CTIS (three subscales), APSQ (two subscales), and DISRS (three subscales)] that would indicate no overlap, a two-factor solution [CTIS + APSQ (five subscales), and DISRS (three subscales)] that would indicate an overlap between Catastrophizing and Worries compared to the most fundamental CFA model, and a one-factor solution [CTIS + APSQ + DISRS (eight subscales)] that would indicate a total overlapping model involving all three negative thought process scales, as a reference. AllCFA analyses in the study were done with the lavaan package in (Rosseel, 2012). The fit statistics reported follows recommendations and consists of a model test statistics and three approximate fit indexes (Kline, 2015). Maximum likelihood estimation with robust standard errors (MLR) was used as the estimation method, since the study variables did not follow a normal distribution. The CFA models were evaluated with an overall model fit approach using recommendations (Bentler, 2007). Several indicators of overall model fit were examined, including (a) the Root Mean Square Error of Approximation (RMSEA) (Steiger, 1990), a parsimony-adjusted index with values ≤ .08 indicating acceptable model fit and values ≤.05 indicating good fit, 0.05–0.08 an acceptable fit, 0.08–0.10 a marginal fit, and >0.10 a poor fit; (b) the Comparative Fit Index (CFI) (Bentler, 1990), an incremental fit index (i.e., the ratio of the deviation of the user model from the baseline model against deviation of the saturated model from the baseline model) with a value of 0.90 considered as a cutoff for an acceptable fit and a value of 0.95 for a good fit; and (c) the Standardized Root Mean Square Residual (SRMR) (Hu & Bentler, 1999), a badness-of-fit statistic, where a value of zero indicates the best result and a value over .10 indicate a poor fit and that the matrix of correlation residuals should be inspected. The likelihood ratio χ² was also reported for completeness; however, it was not used as the only indicator of model fit since it is highly influenced by sample size and does not demonstrate degree of fit (Bentler, 2007).

Third, the best factor solution from the second step was used as independent variables in a linear regression with the two insomnia symptoms scales (the ISI and SCI) as dependent variables.

Results

Correlations between the three negative thought process scales and insomnia symptoms

In Table 1, the descriptive statistics (means and standard deviations) and alpha statistics for the three negative thought process scales and their subscales are displayed. As can be seen in the table, the inter-correlations were high for the Catastrophic Thoughts about Insomnia Scale and its suggested three subscales (r = .57–.95), the Anxiety and Preoccupation about Sleep Questionnaire and its two proposed subscales (r = .78–.97), and the Daytime Insomnia Symptom Response Scale and its proposed three subscales (r = .73–.93). The internal consistencies for the scales and subscales ranged from α = .75 to α = .94.

Table 1. Descriptive statistics for the APSQ, CTIS, DISRS, ISI, and SCI: Observed correlations between the subscales, means, standard deviations, and Cronbach’s alphas.

		1	2	3	4	5	6	7	8	9	10	11	12	13
^1	APSQ: Total
^2	APSQ: Factor 1	0.97
^3	APSQ: Factor 2	0.92	0.78
^4	CTIS: Total	0.80	0.76	0.76
^5	CTIS: Factor 1	0.77	0.74	0.72	0.95
^6	CTIS: Factor 2	0.72	0.71	0.64	0.91	0.79
^7	CTIS: Factor 3	0.62	0.53	0.68	0.79	0.67	0.57
^8	DISRS: Total	0.68	0.68	0.57	0.64	0.62	0.61	.44
^9	DISRS: Factor 1	0.61	0.62	0.50	0.58	0.55	0.55	0.40	0.92
^10	DISRS: Factor 2	0.60	0.60	0.52	0.59	0.58	0.55	0.42	0.93	0.73
^11	DISRS: Factor 3	0.68	0.69	0.57	0.61	0.60	0.60	0.39	0.88	0.74	0.79
^12	ISI: Total	0.73	0.65	0.75	0.72	0.65	0.61	0.67	0.48	0.44	0.43	0.46
^13	SCI: Total	−0.71	−0.64	−0.73	−0.71	−0.65	−0.58	−0.70	−0.51	−0.48	−0.43	−0.48	−0.90
	M	41.5	27.8	13.7	32.9	13.7	12.5	6.6	42.5	18.2	17.3	9.0	8.7	19.9
	SD	23.1	14.8	9.5	21.9	10.0	8.8	5.7	12.4	5.6	5.7	3.0	5.7	7.8
	α	0.94	0.91	0.89	0.92	0.86	0.81	0.75	0.94	0.89	0.88	0.82	0.89	0.88

All the correlations in the table are significantly correlated with one another at the p < .001 level. APSQ =Anxiety and Preoccupation about Sleep Questionnaire, CTIS =Catastrophic Thoughts About Insomnia Scale, DISRS =Daytime Insomnia Symptom Response Scale, ISI =Insomnia Severity Index, SCI =Sleep Condition Indicator.APSQ factor 1: worries about the consequences of poor sleep, APSQ factor 2: worries about the uncontrollability of sleep. CTIS factor 1: rumination. CTIS factor 2: magnification. CTIS factor 3: helplessness. DISRS factor 1: cognitive/motivation, DISRS factor 2: negative state, DISRS factor 3: tired.

Table 1 also depicts correlations between the three negative thought process scales and their subscales with the two scales assessing insomnia symptoms (Insomnia Severity Index and Sleep Condition Indicator). Overall, the correlations were moderate in strength with the two measures determining insomnia symptoms (r = .43–.75). Further, the correlations appear stronger for the Catastrophic Thoughts about Insomnia Scale and the Anxiety and Preoccupation about Sleep Questionnaire and their respective subscales (r = .58–.72; .64–.75), relative to the Daytime Insomnia Symptom Response Scale and its subscales (r = .43–.51).

Confirmatory factor analyses of the three negative thought process scales

The test statistic for the three scales indicated that none of the three scales follow a multivariate normality distribution [Catastrophic Thoughts about Insomnia Scale (H = 1336.62, p < .001), Anxiety and Preoccupation about Sleep Questionnaire (H = 443.82, p < .001), and Daytime Insomnia Symptom Response Scale (H = 1281.62, p < .001)]. Moreover, the Shapiro–Wilk test indicated that neither the Insomnia Severity Index nor the Sleep Condition Indicator scales follow a normal distribution [ISI (W = 0.97, p < .001) and SCI (W = 0.96, p < .001)].

The Goodness-of-fit indicators from the CFA for a two-factor solution and a three-factor solution, relative to a fundamental one-factor solution for the overlapping model of the three negative thought scales, are displayed in Table 2. As can be seen in Table 2, both the two-factor and the three-factor solutions displayed a superior fit compared to the one-factor solution. This was found for both groups of students, i.e., those with high and those with low ISI scores. However, a significant chi square value means that the user model should be rejected. Nevertheless, it is well documented in CFA and SEM literature that the chi square statistic is often overly sensitive in model testing, especially for large samples (Bentler, 2007). The RMSEA estimates in Table 2 indicate an acceptable fit only for the three-factor model among students with a low IS value, while all the remaining models had a poor fit according to RMSEA. The CFI values for the two-factor and three-factor models for students with high ISI values showed an acceptable fit, while the three-factor model for students with low ISI scores showed a good fit. The SRMR estimates was not too high (>.10) for any of the models that a revision of the models was advisable. No modification of any of the models was therefore performed.

Table 2. Goodness-of-fit indicators of overlapping models for the three scales (catastrophic thoughts about insomnia scale, anxiety and preoccupation about sleep questionnaire, and daytime insomnia symptom response scale)stratified on students with high respectively low ISI scores.

Model	χ^2	df	RMSEA (90% CI)	CFI	SRMR	Raykov’s rho
Students with an ISI score of 11 or more (n = 128)
Single Factor	144.87***	20	0.221 (0.187–0.256)	0.759	0.095
	Factor 1 = Rumination + Magnification + Helplessness + Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep + Cognitive/motivation + Negative state + Tired					0.887
Two Factor	69.81***	19	0.145 (0.109–0.182)	0.902	0.060
	Factor 1 = Rumination + Magnification + Helplessness + Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep					0.869
	Factor 2 = Cognitive/motivation + Negative state + Tired					0.853
Three Factor	46.71***	17	0.117 (0.078–0.157)	0.943	0.049
	Factor 1 = Rumination + Magnification + Helplessness					0.868
	Factor 2 = Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep					0.853
	Factor 3 = Cognitive/motivation + Negative state + Tired					0.791
Students with an ISI score of 10 or less (n = 232)
Single Factor	153.53***	20	0.170 (0.147–0.193)	0.842	0.070
	Factor 1 = Rumination + Magnification + Helplessness + Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep + Cognitive/motivation + Negative state + Tired					0.903
Two Factor	64.62***	19	0.102 (0.077–0.127)	0.946	0.044
	Factor 1 = Rumination + Magnification + Helplessness + Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep					0.879
	Factor 2 = Cognitive/motivation + Negative state + Tired					0.857
Three Factor	36.22***	17	0.070 (0.040–0.107)	0.977	0.032
	Factor 1 = Rumination + Magnification + Helplessness					0.827
	Factor 2 = Worries about the consequences of poor sleep + Worries about the uncontrollability of sleep					0.854
	Factor 3 = Cognitive/motivation + Negative state + Tired					0.870

***p < .001.

The last column in Table 2 displays the construct reliability for all factor of the models, all of which were over 0.7, which is acceptable, and indicate that the overlapping scale is reliable. Based on the results in Table 2, we can conclude that the three-factor model has the best fit for both groups of students (i.e., high ISI values respectively low ISI values).

A final indication of that goodness-of-fit is better for the two-factor model compared to the single factor model and that the three-factor model is better compared to the two-factor model can be made by comparing BIC (Bayesian Information Criteria) values which is appropriate for both nested models and non-nested models. However, there is no explicit value for what a good fit is. The BIC value is only informative when it is compared to a competing model’s BIC value, where a lower value has a better fit than a higher value. Following Jeffrey-Raftery’s guidelines (Raftery, 1995), if the difference in BICs between two models is 0–2, this constitutes “weak” evidence in favor of the model with the smaller BIC; a difference in BICs between 2 and 6 constitutes “positive” evidence; a difference in BICs between 6 and 10 constitutes “strong” evidence; and a difference in BICs greater than 10 constitutes “very strong” evidence in favor of the model with smaller BIC. The BIC values were 6386.627 for the Single Factor model, 6320.782 for the Two Factor model and 6307.989 for the Three Factor model in the group of students with high ISI values, while the corresponding BIC values for students with low ISI values were as follows: 10863.271, 10768.070, and 10,744.113. This means that each increase of one factor more showed a “very strong” evidence in favor of the model with more factors and this was true for both groups of students (i.e., high ISI values respectively low ISI values).

The factor loadings for the non-overlapping, three-factor model are given in Table 3, stratified on the two groups of ISI values (students with an ISI score of 11 or more and students with an ISI score of 10 or less). Note that all factor loadings for all subscales from the three negative thought process measures are acceptable (>.30).

Table 3. Standardized loadings for the three-factor confirmatory model of catastrophizing, worry, and rumination stratified on students with high respectively low ISI scores.

	CTIS		APSQ		DISRS
Subscales	ISI score ≥ 11	ISI score < 11	ISI score ≥ 11	ISI score < 11	ISI score ≥ 11	ISI score < 11
Rumination	.80	.83
Magnification	.87	.83
Helplessness	.78	.82
Worries about the consequences of poor sleep			.92	.88
Worries about the uncontrollability of sleep			.81	.74
Cognitive/Motivation					.84	.92
Negative state					.79	.82
Tired					.47	.59

For students with an ISI score of 11 or more (n = 128), the correlation between the latent variables: CTIS and APSQ = 0.70, CTIS and DISRS = 0.62, and APSQ and DISRS = 0.79, while for students with an ISI score of less than 11 (n = 232), the corresponding figures were 0.77, 0.71, and 0.83.APSQ = Anxiety and Preoccupation about Sleep Questionnaire, CTIS = Catastrophic Thoughts About Insomnia Scale, DISRS = Daytime Insomnia Symptom Response Scale.

Association between negative thought processes and insomnia symptoms

In the final step of the analysis, the association between insomnia symptoms and negative thought process measured by the three factors “Catastrophizing”, “Worry”, and “Rumination” from the three-factor non-overlapping model was examined by linear regression model with first ISI as the dependent variable and secondly SCI as the dependent variable. However, we were concerned that Ordinary Least Squares regression (OLS) might not be appropriate since both the dependent variables were not normally distributed. Not normally distributed outcomes do not necessarily have to be problematic per se but indicate that there may be other problems in data that can influence the estimates in a severe way. After a first preliminary OLS, we studied leverage, Cook’s distance, and outliers and found that there were about two to four potential problematic observations in the data. We therefore tried two robust regression models with different weighting functions, Huber weights, and the bi-square weighting function. Basically, a robust regression model tries to weight down observations with high residuals instead of removing such observations if there is no logical reason for them to be removed. Neither of the two weighting functions had any major impact on the estimates from the OLS model, so the results in Table 4 are from the first preliminary OLS. The beta coefficients in Table 4 are unstandardized coefficients. However, it is difficult to decide the relative importance of the three factors by the beta coefficients since the factors are likely to be correlated with each other. A simple way to overcome this problem is to compare standardized coefficients (z values), which are included in Table 4 and describe the expected change in the outcome (ISI or SCI) expressed in standard deviation units for a standard deviation change in a factor, holding the other factors constant.

Table 4. Associations between three scale factors and two insomnia symptoms measures: Results from a multiple linear regression model stratified on students with high, respectively, low ISI scores.

Insomnia symptoms Factor	ISI score ≥ 11 (n = 128)					ISI score < 10 (n = 232)
	β	SE	t	p	z	β	SE	t	p	z
Insomnia Severity Index
Intercept	10.05	1.059	9.49	<0.001		1.88	0.663	2.84	0.003
Catastrophizing (CTIS)	0.04	0.019	2.19	0.031	0.228	0.05	0.015	3.36	<0.001	0.275
Worries (APSQ)	0.06	0.018	3.45	<0.001	0.384	0.05	0.015	3.57	<0.001	0.298
Rumination (DISRS)	−0.02	0.026	−0.65	0.518	−0.063	0.01	0.021	0.55	0.585	0.041
Sleep Condition Indicator
Intercept	18.30	1.536	11.91	<0.001		31.17	1.113	28.00	<0.001
Catastrophizing (CTIS)	−0.05	0.027	−2.01	0.046	−0.217	−0.10	0.026	−3.72	<0.001	−0.296
Worries (APSQ)	−0.08	0.027	−2.96	0.004	−0.342	−0.08	0.026	−3.05	0.003	−0.247
Rumination (DISRS)	0.03	0.038	0.69	0.491	0.069	−0.06	0.036	−1.64	0.102	−0.119

APSQ = Anxiety and Preoccupation about Sleep Questionnaire, CTIS = Catastrophic Thoughts About Insomnia Scale, DISRS = Daytime Insomnia Symptom Response Scale.

The results in Table 4 indicates that the factors Catastrophizing (CTIS) and Worry (APSQ) have an association with both ISI and SCI scores. The factor Rumination (DISRS) displayed a lower association to both ISI and SCI scores and were not statistically significant. Catastrophizing (CTIS) had the highest estimate and relative importance for SCI among students with a low ISI score while the estimates otherwise were very similar between the two groups of students and where Worries (APSQ) was also the factor with the highest relative importance. The estimates for all factors in the four models in Table 4 are conditional on that the two other factors are held constant.

Discussion

The purpose with the current study was to explore conceptual overlap, underlying factors, and validity of three theory-derived, insomnia-related negative thought processes (i.e., catastrophizing, worry, and rumination). To our knowledge, this is the first study to purposefully explore overlap between the three negative thought processes and insomnia symptoms. The findings provide support for that all three negative thought processes can be conceived as distinct. In multivariate analyses, catastrophizing and worry were significantly associated with insomnia symptoms, but rumination was not.

The central finding in this study was that catastrophizing, worry, and rumination appears to be distinct constructs from one another, a finding supportive of the proposal by Hiller et al. (2015). The question is then how this should be interpreted. One interpretation is that the current findings reflect how these insomnia-specific processes operate, namely that catastrophizing, worry, and rumination represent three related but distinct forms of negative thought processes. Of note is that this interpretation is not at odds with current cognitive models of insomnia (Espie, 2002; Harvey, 2002; Lundh & Broman, 2000). For example, in the cognitive model, the key process in the maintenance of insomnia is “excessive negatively toned cognitive activity” (Harvey, 2002; Harvey et al., 2005), and it is conceivable that such activity might vary in its content and focus. It is not surprising that the worry and rumination measures were related and distinct based on a large literature outside of the insomnia field (Ehring & Watkins, 2008). Worry and rumination appear similar in that both processes are self-directed, passive, and associated with symptoms of psychopathology (Nolen-Hoeksema et al., 2008; Segerstrom et al., 2000; Siegle et al., 2004). However, the two processes also consist of differing thought content (possible threat themes in worry versus a focus on self-worth and loss in rumination) and time orientation (mainly future-oriented in worry versus primarily past-oriented in rumination) (Ehring & Watkins, 2008; McLaughlin et al., 2007; Nolen-Hoeksema et al., 2008). These differences might also explain why worry (and catastrophizing) were significantly related to insomnia symptoms in this study and rumination was not; worry is more likely to trigger activation of the central nervous system and arousal, which has detrimental effects for sleep (Harvey, 2002). Another possible reason why rumination was not significantly associated with insomnia symptoms is that both the ISI and SCI are primarily measures of nocturnal symptoms, and, according to Carney et al. (2006), rumination is believed to be triggered by the daytime symptoms of insomnia (e.g., fatigue and inertia) and not the nocturnal symptoms (Tutek et al., 2021).

It is more surprising that the catastrophizing and worry measures displayed distinctness since catastrophizing have been conceived as merely an intense form of worry, at least in the anxiety and pain literature (Davey & Levy, 1998; Flink et al., 2013). However, one could argue that even though the form of thinking is similar (worry to catastrophic worry on a continuum), the function of the two processes might be slightly different. For example, it is conceivable that catastrophizing, relative to worry, results in a stronger activation of arousal and negative emotions and subsequent more pronounced insomnia symptoms. Based on research in the anxiety area, one could also argue that catastrophizing is distinct from worry since it is a process that exacerbates the adverse effects of worry, meaning that excessive worry occurs through catastrophizing (Kendall & Ingram, 1987). Research on the three negative thought processes simultaneously have been very scarce. To our knowledge, there is only one previous study that have used measures of catastrophizing, worry, and rumination in this case, generic scales completed by several convenience samples (Pike et al., 2021). The main findings from the study were that catastrophizing emerged as a unitary construct following factor analysis, and catastrophizing was independent of both worry and rumination. Although more factor analytical research is needed to further explore the conceptual overlap between catastrophizing and worry, there is at least some indications that the two constructs might be distinct.

The three-factor solution demonstrated to have the best fit in the current study could, however, be questioned based on some key findings. Although the three-factor solution displayed better fit than the one- and two-factor solutions, it did not produce good fit on all the indicators: (1) the chi square values were significant for all tested models, although the chi square statistic has been criticized for being overly sensitive in model testing (Bentler, 2007), (b) the RMSEA values indicated poor to acceptable fit in the high and low insomnia symptom groups, and (c) the CFI value suggested only an acceptable fit in the high insomnia symptom group (11 points or more on the Insomnia Severity Index). These fit indicators thus display varying model fit, and further research is therefore warranted to explore which factors that impact on the model fit indicators, such as participant characteristics, sample size, and choice of measures to assess negative thought processes.

A second interpretation is that these findings are context-dependent, e.g., due to the choice of measures indexing the three negative thought processes and sample characteristics. When reading the instructions and items of the three scales, our interpretation is that all three measures aim to assess trait-like negative thinking, e.g., all scales instruct respondents to complete the items based on how their thinking has been in general or during the past month. Also, all three measures consist of items tapping how people think in specific terms, e.g., distinct worries or unique symptoms of distress. One difference between the three measures is, however, that the response alternatives differ slightly from “not true” – “very true”, “strongly disagree” – “strongly agree”, and “almost never” – “almost always”. It is thus possible that the slightly differing response alternatives enhances the observed distinctness between the three measures in this study. Future research could explore this interpretation by using similar or identical response alternatives or by varying self-report scales, e.g., the Insomnia Catastrophizing Scale (Jansson-Fröjmark et al., 2020) instead of the Catastrophic Thoughts about Insomnia Scale or the Daytime Insomnia Worry Scale instead of the Anxiety and Preoccupation about Sleep Questionnaire (Kallestad et al., 2010). Another feature of the three measures that might explain some of the findings is that there is an element of construct contamination. More specifically, five items in the Catastrophic Thoughts about Insomnia Scale uses the terms “worry” or “worrying”, of which two items belong to the magnification subscale and three items to the rumination subscale. Another concern is that one subscale in the Catastrophic Thoughts about Insomnia Scale is labeled “Rumination” and could therefore invoke construct contamination versus the Daytime Insomnia Symptom Response Scale. These two instances of possible construct contamination could potentially increase the overlap and decrease the distinctness of the analysis involving the Catastrophic Thoughts about Insomnia Scale, thereby underestimating the distinctness of the three measures. Contamination-free measures would be preferable in future research.

Another context-dependent feature of importance is the study participants. The current study recruited relatively young students and divided them into two groups based on their levels of insomnia symptoms. It is not certain that identical results would have been obtained with well-screened participants with insomnia disorder, particularly since the level of distress (e.g., depression) and other potent insomnia-related factors may elevate the risk of catastrophizing, worrying, and ruminating. Future research could examine this possibility by alternating type of study participants, e.g., patients with insomnia disorder and normal sleepers as well as older participants in the community.

A third interpretation of the fact that catastrophizing, worry, and rumination appear to be distinct constructs from one another is considering how negative thought processes can be conceptualized in general. It is essential to underscore that confirmatory factor analysis has been criticized for its overly restrictive independent cluster model, which demands that each item (in this study three scales and their subscales) is defined by one, and only one, content domain. As this procedure can result in inflated factor correlations, confirmatory factor analysis is not always the optimal tool for identifying factors (A. J. S. Morin et al., 2017). As a result, we highlight two additional conceptualizations that can be relevant when reflecting upon the current study’s findings and for future research. First, a hierarchical model is important to mention when considering the interpretation of the three distinct constructs (Lahey et al., 2021; Shihata et al., 2017). In particular, “repetitive negative thinking”or “perseverative thinking” could be viewed as higher-order factors in a hierarchical model with catastrophizing, worry, and rumination as specific factors at a lower level. In that line, it is also conceivable that additional specific negative thought processes may be distinct factors at a lower level, such as perfectionistic concerns and intolerance of uncertainty. Second, a bifactor model could also be envisioned in which catastrophizing, worry, and rumination are independent and non-redundant factors in relation to a general factor, such as “repetitive negative thinking” (Alamer, 2022). One study outside of the insomnia field using a bifactor model have provided empirical support that repetitive negative thinking could be conceived as a general, latent factor and worry and rumination as two uncorrelated factors (Topper et al., 2014). Further research is therefore warranted on how to optimally understand the structure of catastrophizing, worry, and rumination with general and additional specific factors.

There are several limitations with the current study that might impact the interpretation. One limitation of the current study is that the sample was homogenous in its demographics (i.e., largely young female adults), which limits conclusions about the general population. A second limitation is that the study is cross-sectional, which precludes examinations of these relationships across time. More research is therefore needed to identify which of the negative thought processes that have the most potent value in predicting insomnia symptoms over time and for whom. Further research might also investigate how to conceptualize negative thought processes in insomnia in the most optimal way. For example, should catastrophizing, worry, and rumination be thought of as first-level factors and that there is a higher-order factor above, i.e., “excessive negatively toned cognitive activity” or “negative thought processes”? If this notion is valid, developing measures with items assessing higher-order factors could be one way forward. Future research might also assess insomnia symptoms in different ways, e.g., self-report vs non-subjective assessment and survey vs sleep diary. Although CFA is commonly recommended as a tool to explore discriminant validity, it is important to underscore that CFA is only a test of acceptance of the fit of a model to the data; it is plausible that other factor models fit equally well or better. It is therefore recommended that future factor analytical studies explore different factor solutions and statistical approaches [e.g., bifactor and exploratory structural equation modeling; (A. J. S. Morin et al., 2020)]. A final limitation is that the insomnia categorization into “high” and “low” insomnia symptoms was based only on a cutoff on the Insomnia Severity Index. Despite that the cutoff has been shown to be relatively precise in distinguishing between insomnia and non-insomnia, our categorization should not be interpreted as indicating that one group meets criteria for insomnia disorder and the other group does not (C. M. Morin et al., 2011). As a result, the study aims need to be explored further in a well-defined insomnia disorder sample against a well-screened good sleeper sample.

There are several implications that follow from the current findings. In theoretical terms, it appears valid to view catastrophizing, worry, and rumination as related but distinct processes. This could potentially lead to a more parsimonious theory on cognitive processes in insomnia, particularly for the plethora of cognitively oriented models of insomnia (Espie, 2002; Harvey, 2002). In particular, the current results suggest that the three negative thought processes are distinct to a certain degree and need to be separated in cognitive conceptualizations of insomnia (Tang et al., 2023). It is, for example, likely that the thought processes trigger different arousal and emotional responses and varying levels of insomnia symptoms. Since it appears as if catastrophizing, worry, and rumination are distinct processes and could play different roles in insomnia, continued assessment of the thought processes with separate self-report scales seems warranted in clinical and research settings. The current findings might also have implications for future clinical research that aim to modify negative thought processes or include measures of negative thought processes as mediators or outcomes. One clinical implication is to examine which psychosocial treatments have the strongest effect on catastrophizing, worry, and rumination and how these interventions also impact on insomnia symptoms. In a systematic review exploring outcomes for mostly patients from clinical settings or with diagnosed mental disorders, mindfulness-based and cognitive behavioral interventions were found effective to reduce both worry and rumination (Querstret & Cropley, 2013). In the insomnia field, cognitive behavioral therapy for insomnia produces moderate-to-large effects on worry but small and non-reliable changes on rumination (Ballesio et al., 2021). Yet again, the smaller effect on rumination, relative to worry, could be interpreted as if the two constructs are distinct.

The findings in the current study suggest that there is considerable conceptual overlap between insomnia-specific measures of catastrophizing and worry, but that rumination emerged as a distinct construct. In the long run, this might have implications for theory formation and clinical research and practice. The concept of negative thought processes is multidimensional, and more work is therefore warranted to examine what thought concepts are key to the development and maintenance of insomnia.

Acknowledgments

We would like to acknowledge the students who participated in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Funding

The authors received no financial support for the research, authorship, or publication of this article.