1 Introduction

Insomnia is the most commonly diagnosed comorbidity disorder among patients with chronic pain [1]: 50% of individuals with insomnia disorders suffer from chronic pain and 50–80% of chronic pain patients develop insomnia symptoms [1, 2, 3, 4, 5, 6, 7]. In Sweden, the comorbidity between clinically significant insomnia and pain is almost 65% [8]. A longitudinal study on an adult female population in Norway found that insomnia preceded 2/3 of the incident cases of fibromyalgia [9]. As these data indicate, it is important to evaluate insomnia severity among chronic pain disorders.

Indeed, there is several assessment tools for assessing insomnia of which polysomnography (PSG) is considered the gold standard [10]. Regarding sleep questionnaires, the Insomnia Severity Index (ISI) and the Pittsburgh Sleep Quality Index (PSQI), are the two commonly used instruments for measuring insomnia [11,12]. However, considering the poor availability of PSG in daily routine and time consuming accomplishment of PSQI among pain patients, there is a need of short and valid self-instruments. Therefore the ISI is the one of the most frequently used scale to distinguish clinically significant insomnia [10, 11, 12, 13, 14].

The ISI consists of only seven items and estimates patient’s perception of insomnia as well as identifies both daytime symptoms and night-time severity of insomnia [11]. Thus, the ISI’s diagnostic criteria for insomnia are in accordance with the Diagnostic and Statistical Manual of Mental Disorders and the International Classification of Sleep Disorders [15,16]. Additionally, the ISI items are related with polysomnography variables [17]. The ISI has been translated into several languages and its psychometric properties have been examined in a variety of versions [13,14,18, 19, 20]. However, the results are mixed and inconsistent across different cultures and populations. Bastien et al. [11] reported a three-factor solution in a clinical sample of middle-aged insomniacs and Savard et al. [19] proposed a two-factor solution using two samples of cancer patients. Results from studies that examined the Spanish ISI are also inconsistent: one study found a one-factor solution using factor analysis in a group of older adults [21], and another study found a three-factor solution in the general population using confirmatory factor analysis [14]. None of these studies examined the factor structure of the Insomnia Severity Index in a sample of chronic pain patients. As a result, the proposed psychometric properties and factor solutions for ISI have not been identified in this target group. Additionally the Swedish version of the Insomnia Severity Index has not been evaluated regarding the generability of the factorial structure underlying it. Even if some studies have reported the psychometric properties of the instrument [22] or its clinical feasibility [8,23], the verification of the proposed factor model as well as the invariance property (e.g., between males and females) are lacking. As such, this study tests the validation of the Swedish ISI and evaluates its psychometric properties using both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) in a large sample of chronic pain patients.

2 Material and methods

3 Results

3.3 Confirmatory factor analysis

Confirmatory factor analysis (CFA) was applied to Sample 2 to test the fit of the one global factor structure; however, the fit model with the seven items, as proposed by initial authors, was not satisfactory (Table 3). Next, a series of CFA models were performed by omitting items from the initial scale via a sequential procedure to identify a model with adequate fit. The item exclusion criteria were (a) low communalities, (b) low item-total correlations (removing items that reduced the Cronbach’s alpha coefficient), and (c) CFA fit indices of the resulting models. This procedure resulted in a four-item Insomnia Severity Index (ISI-4) (Fig. 2) that provided satisfactory fit indices (Table 3). However, two other proposed alternative factor models of the CFA were also applied to compare the results of Fernandez-Mendoza et al. [14]. Next, we assessed the sex invariance by comparing two nested models: one model that hypothesized that the loadings were equal irrespective of sex (constrained model) and one unconstrained. The difference between the models was not significant (Δχ² = 2.86, df=3, p = 41), so the constrained model was preferred. Therefore, there is assumed measurement invariance in the loadings by participants’ sex. The difference of structural invariance between sexes was not statistically significant in either model (Δχ² = 10.6; df=3; p = .69).

Table 3

Absolute and relative fit indices measurement models of the four-item Insomnia Severity Index (ISI-4; Sample 2, n = 502).

Model	Absolute and relative fit measures
	X 2 [*]	df	X2/df	GFI	RMR	RMSEA (p-c lose[*])	NNFI	CFI	AIC
One factor (Fernandez-Mendoza et al. (2011)	267.64	14	19.12	0.84	0.09	NR	0.69	0.79	295.64
One factor global factor (7 items)	150.12	14	10.8	0.85	0.06	0.139 (<.001)	0.87	0.94	192.12
Four-items one global factor	6.54	2	3.27	0.99	0.02	0.06 (.238)	0.99	0.99	22.54
Two factor (Fernandez-Mendoza et al. (2011)	151.05	13	11.61	0.91	0.06	NR	0.82	0.89	181.05
Two factor (7 items)	96.55	13	7.42	0.94	0.05	0.10 (<.001)	0.89	0.96	138.55
Three factor (Fernandez-Mendoza et al. (2011)	31.65	9	3.51	0.98	0.03	NR	0.95	0.98	69.65
Three factor (7 items)	29.56	9	3.28	0.99	0.02	0.39 (<.001)	0.96	0.99	70.00

1. NR = not reported.

Fig. 1

Scree plot of eigenvalues indicating the one-factor solution for Sample 1.

Fig. 2

Factor structure of the one-global factor of the Swedish version of the four-item Insomnia. Severity Index (ISI-4) for Sample 2 Notes. ISI2 = difficulty staying asleep, ISI4 = satisfaction, ISI5 = interference, ISI7 = distress F1 = one-global factor.

4 Discussion

In this study, we examined the factor structure and the psychometric properties of the Insomnia Severity Index (ISI) in a large sample of chronic pain patients. The exploratory factor analysis (EFA) of the Swedish version found one global factor solution, replicating to some extent results found by Sierra et al. [21]. However, the confirmatory factor analysis (CFA) model showed that the fit of the one-factor model with the seven items was not adequate, a finding that was also reported in a previous study [14]. However, both analytic methods did not replicate the three-factor solution described by Fernandez-Mendoza et al. the only previous CFA results reported on the ISI [14]. Therefore, we used an alternative approach assessing the individual item’s contributions to the instrument, which is known as a “refining” procedure [29,30]. This procedure, using a shorter version consisting of only four items, resulted in satisfactory fit indices and supported the proposed structure with respect to exploratory factor analysis. Further evidence was supported in terms of item response pattern, internal consistency, and convergent validity. Given these results, this is the first study to report the factor structure of the ISI in Swedish patients with chronic pain.

The findings of the present study provide additional information about the factor structure of the ISI in chronic pain samples. In particular, we found a better fit to our data after excluding three items of the initial version. The following three items had the lowest communalities or reduced the Cronbach’s alpha coefficient: Item 1 - “Difficulty falling asleep”; Item 3 - “Problem waking up too early”; and Item 6 - “How noticeable to others do you think your sleep problem is in terms of impairing your quality of your life?” These results mean that the dimensions of severity of sleep onset, early morning awakening, and noticeability may not add significantly to the insomnia severity in the chronic pain patients in our study. Morin et al. [13] found higher endorsement of responses for insomnia maintenance (Item 2), for satisfaction (Item 4), for interference (Item 5), and for distress (Item 7). One explanation for this might be the different cultural and clinical context of the study [13]. It has been suggested that many health-related terms tend to have different meanings for individuals in different contexts [11,13,14,18,19,20]. Additionally, many investigators have reported different factor structure of the ISI regarding different cultures and samples [11,18,19,20]. Furthermore, for patients with high levels of pain, the primary impairment is in the perception of sleep quality rather than sleep quality per se [2,3,4,5,6,7,13]. Therefore, Morin et al. [13] concludes that epidemiological studies can use an instrument with only two or three key items to identify insomniacs.

Structural and measurement invariance between men and women groups were supported for the four-item Insomnia Severity Index (ISI-4) scores. Even though our sample included more women than men, the difference on ISI-4 scores between the two sexes was not statistically significant and the dimension of sex only slightly affected the score of the four-item Insomnia Severity Index. Age did not affect the scores with respect to regression analysis. The Cronbach’s alpha for ISI-4 score was very good, almost excellent, with respect to reliability. Furthermore, the correlations between the ISI-4 and the HADS scales were positive, replicating the results that suggest depression and anxiety are related to insomnia severity in chronic pain patients [8]. As expected, we found negative correlations with mental dimension of quality of life and positive correlations with the one item of MPI concerning quality of sleep.

One caveat of our study is that our team did not perform the initial backward translation procedure. To our knowledge, this is the most frequently used version adapted to Swedish language and culture [8,22,23]. All the cultural adaptation procedures have performed by well established guidelines i.e., forward translation, expert panel back-translation, pre-testing and cognitive interviewing and final version [22]. Although we may have not examined the stability of the ISI-4 and the criterion validity of the new version of the instrument, our study’s large sample gives us statistical power to derive solid inferences. Future research efforts should include these issues as well as the screening capacity of the Swedish four-item Insomnia Severity Index. It should be noted that the convergent validity of the ISI-4 scale was measured using only one single-item. Upcoming studies could improve on the present design by using other validated insomnia instruments, which would help to further launch the construct validity of the ISI-4. Additional research is also needed to further refine this version among different chronic pain conditions, duration of pain, and chronic pain intensity.

One may argue that we have not taken into account the two of the most prevalent clinical indicators of sleep problems, by not including item 1 and item 3 in this shorter version. However our results do not exclude that “Difficulty falling asleep” and “problem waking up too early” can be important problems for patients with chronic pain. Hence, herein we only underline that these two aspects had the lowest variation in our sample and therefore they were not considered important in both EFA and CFA. Nevertheless, we have not been able to control for the pharmacological treatment in this sample data and it could partly explain the observed variation among the items of ISI.

Despite these limitations, we conclude that this shorter version of the initial instrument is adequate as it provides satisfactory factor structure, internal reliability, and concurrent validity. Therefore, we believe the four-item Swedish Insomnia Severity Index (ISI-4) accurately identifies insomnia in a chronic pain population (see the Appendix).

5 Conclusions

This is the first study of the psychometric properties and factor structure of the Swedish version of the Insomnia Severity Index in chronic population. The results suggest that a shorter version, consisted of only four items of the initial instrument, could effectively evaluate insomnia in chronic pain patients.

Highlights

• The feasibility of the Insomnia Severity Index (ISI) was examined.

• 836 patients from the Swedish registry for pain rehabilitation were recruited.

• We report a shorter version for evaluation of insomnia in chronic pain patients.