Background
Temporomandibular disorders (TMD) are the umbrella term encompassing mostly chronic pain conditions involving the masticatory muscles, temporomandibular joint (TMJ) an associated structures. More than half of the patients with chronic pain conditions report poor sleep quality [
1]. Studies show that disruption of sleep exacerbates pain and, conversely, pain contributes to sleep disturbance [
1]. Therefore, impaired sleep quality can contribute substantially to the suffering of chronic pain patients because the sleep disorders are associated with significant quality of life impairments. Patients with TMD pain are no exception. They frequently suffer from chronic orofacial pain and also have comorbid sleep disorders [
2]. They commonly report poor SQ (up to 90%) [
2]. Since sleep disturbances have been associated with poor treatment outcomes in the TMD patients [
3], the assessment of sleep quality needs to be a part of the comprehensive evaluation of this patient population.
Subjective sleep assessment is challenging and ideally provides data on at least four characteristics of sleep: sleep initiation, sleep maintenance, sleep adequacy, and daytime somnolence. Whereas polysomnography is an objective measure of biophysiological sleep parameters, sleep quality is usually assessed using self-report. There are a number of patient-reported outcome instruments that measure various aspects of sleep quality. Seven of these instruments assess the four sleep characteristics noted above: Basic Nordic Sleep Questionnaire, Leeds Sleep Evaluation Questionnaire, Medical Outcomes Study Sleep Scale, Pittsburgh Sleep Diary, Sleep Dissatisfaction Questionnaire, Self-Rated Sleep Questionnaire, and the Pittsburgh Sleep Quality Index (PSQI) [
4].
Since the introduction of the PSQI in 1989 by Buysse et al. [
5] to measure sleep quality among adult psychiatric patients, this instrument has been employed in numerous other patient populations including cancer, traumatic brain injury and chronic pain patients in over 2,200 published studies. The PSQI is composed of 19 items, which are combined into seven components that are summarized into a global score that represents a unidimensional sleep quality construct [
5]. However, five recent studies [
6‐
10] have questioned the appropriateness of using only the global score. In 2006, Cole et al. examined PSQI structure by exploratory and confirmatory factor analysis in healthy and depressed elderly adults and showed initial evidence that a single global PSQI score did not capture the multidimensional nature of sleep disturbances when examined by PSQI [
6]. Subsequently, four other studies explored the dimensionality of the PSQI and reported that a two- and three- factor scoring model for the PSQI were better to assess sleep quality compared to the originally proposed single-factor model (Table
1) [
5].
Table 1
Published studies investigating two- or three- factor scoring models for the Pittsburgh Sleep Quality Index with samples description
| USA community-dwelling depressed and nondepressed adults > 60 years | 417 | Three-factor | Sleep Efficiency (sleep duration, habitual sleep efficiency) |
Perceived Sleep Quality (subjective sleep quality, sleep latency, use of sleep medication) |
Daily Disturbances (sleep disturbances, daytime dysfunction) |
| Nigerian university students | 520 | Three-factor | First factor (subjective sleep quality, sleep latency, habitual sleep efficiency, sleep disturbances, use of sleep medication) |
Second factor (sleep duration and sleep disturbances) |
Third factor (subjective sleep quality, habitual sleep efficiency, use of sleep medication) |
| Australian adults aged 18 to 59 years | 364 | Two- and three-factor | Same factors as Cole et al. for the three- and without Daily Disturbances for the two-factor model |
| Swiss renal transplant recipients | 135 | Three-factor | Same factors as Cole et al. [ 6] |
| Belgian chronic fatigue syndrome patients | 413 | Three-factor | Same factors as Cole et al. [ 6] |
In a sample of Nigerian university students, a three-factor model of the PSQI was identified [
7], but the factors differed from Cole’s et al. study making comparisons difficult. A study assessing sleep quality in Australian adults also suggested two- and three-factor scoring models [
8]. A three-factor model of the PSQI was also found to have a better fit in a sample of renal transplant recipients [
9] as well as in a sample of chronic fatigue syndrome patients [
10]. Cole et al. found that the PSQI factor structure has three separate factors, that is, dimensions of
sleep efficiency,
perceived sleep quality, and
daily disturbances and these are reported as 3 separate scores [
6].
While both two-factor and three-factor models have been reported, studies which assessed self-reported sleep disturbances of the TMD patients used a global PSQI score [
2,
11‐
14]. These findings indicate that the factor structure of this instrument in TMD patients needs to be further investigated. In particular, it is uncertain how many PSQI scores are needed to characterize sleep quality in this patient population. Moreover, the psychometric properties of the PSQI used in TMD studies are also unknown. This is important because the use of an instrument in a specific patient population is justified only if these properties are known.
The aim of this study was to assess the dimensionality and the psychometric properties of reliability and validity for the PSQI in cases with pain-related TMD and in cases with pain-free TMD.
Discussion
The current study demonstrated that sleep quality in TMD patients, as assessed by the PSQI, is a unidimensional scoring structure. When the PSQI data for TMD cases with or without painful diagnose were analyzed separately with exploratory factor analysis and the inspection of the scree plots, the results showed clearly a single common latent factor that explained item responses. The only exception was a low loading for the sixth component of the PSQI questionnaire called use of sleep medication for the cases with pain-free TMD. These individuals don’t use a lot of sleep medications, and therefore, sleep disturbances due to sleep medications are challenging to identify. Internal consistency of the PSQI was also lower in these latter cases with Cronbach’s alpha value of 0.66, which was below the threshold for the internal consistency. When the use of sleep medication data was omitted from the analysis, the internal consistency coefficient increased to 0.73. Sleep medication use varies across populations and may not have a strong relationship with the other variables. However, greater use of sleep medications by cases with pain-related TMD may mitigate differences in sleep quality, if the medications are effective, compared to pain-free TMD cases thus masking the true difference between these two types of cases. Conversely, cases with pain-related TMD may use more sleep medications because they use more medications in general, including pain medications, and since they see more doctors for their pain, they have more opportunity to get sleep medications.
The polychoric correlations between the subjective sleep quality and the sleep disturbances, between the subjective sleep quality and the sleep duration, and between the subjective sleep quality and the sleep latency had the largest magnitudes (up to 0.74). These large correlations were similar in cases with pain-related TMD and in cases with pain-free TMD.
The PSQI questionnaire comprises 19 individual items assessing sleep quality, and 15 of them are further combined into seven components. The calculation of these components is performed differently than most multi-item self-report instruments which use simple sum of their items (i.e. Oral Health Impact Profile questionnaire) [
34]. A particular computation is needed for each of the seven PSQI components. This burden may potentially obscure the meaning of each component and also impede a widespread clinical application.
Our results can be generalized to other TMD populations because the RDC/TMD Validation Project data set was derived from a diverse spectrum of TMD clinic and community cases [
15]. All participants were thoroughly evaluated to ensure correct TMD diagnosis necessary for inclusion as study cases [
15]. Even among TMD patient populations with different distributions of TMD diagnostic subtypes, such discrepancies would not limit generalizability, because we have demonstrated that the dimensional structure of the PSQI is not substantially different between the cases with pain-related TMD and cases with pain-free TMD which is a major classification used in TMD clinical practice and research.
We found nine studies using the PSQI instrument in TMD patients [
2,
11,
12,
14,
35‐
39]. All of these studies reported sleep quality as being one dimension represented by one global PSQI score. Some of these studies reported only a global PSQI score and defined poor sleep quality as the global PSQI score being larger than 5 [
35,
38,
39]. However, Yatani et al. [
2] used the median cutoff of a global PSQI score ≥ 10 to divide “good” and “poor sleepers”. The six remaining studies reported results for some [
14,
37] or for all seven PSQI subscores [
11‐
13,
36], but none of these studies reported correlations among the seven components. Thus, this is the first study to report correlations between PSQI subscores in TMD patient population. Besides reporting a global PSQI score for TMD patients, Abrahamsen et al. [
37] listed also the average answers obtained from the individual questions contained in the PSQI questionnaire, e.g. Hours of sleep, Minutes before falling asleep, Number of awakenings, Number of awakenings due to pain, and Episodes of daytime sleep.
We did not find any other multiple-item instrument that has been used to assess sleep quality in the TMD patients. Although some studies have used single–item questionnaires to assess sleep quality, multi-item instruments have advantages in terms of validity and reliability compared to a single question. Therefore, there is a limit to how brief an instrument can be, and in part, depends on whether it is intended for clinical or research purposes. Although the causation is currently unknown [
40], sleep quality is an important issue for successful management of TMD patients. Finally, a briefer version of the PSQI and simplification of its computation would probably popularize its clinical utility significantly.
Our study had some limitations. More than one third of our TMD patient population had one missing value in the PSQI questionnaire. The results were probably not affected substantially, because the unidimensional construct of PSQI is characterized with 15 items. For all the subjects in the analyses we had at least 14 items. It is highly unlikely that this situation may have prevented the detection of a second factor – the major alternative for a unidimensional model. We divided all the TMD diagnoses only in two categories, the pain-related TMD and pain-free TMD. The pain-related TMD diagnoses comprised a myofascial pain as well as TMJ pain, which have different clinical characteristics. When we assessed model fit for our dimensionality results not all findings agreed. All results of the exploratory factor analysis favored unidimensionality, but some indices for the confirmatory factor analysis came only close to guideline recommendations. One reason for this situation might be that our TMD cases only used sleep medication rarely and this item had the lowest correlation with the latent factor and consequently a decreased model fit. Furthermore, a convenience sampling methodology was used. Nevertheless, the psychometric properties of the PSQI were assessed in TMD patients for the first time, and dimensionality of the PSQI was graphically assessed also by the use of scree plots and two factor-analytic methods. While the majority of our methods agreed that sleep disturbances in TMD patients can be characterized with one score and we have explored dimensionality in subgroups, more sophisticated analyses of measurement invariance [
41] across populations are a next step in a rigorous assessment of psychometric properties. We used a convenience sample which is inferior to a consecutive sample, but our sample size was large and our patients covered the entire spectrum of TMD patients.
In conclusion, although the PSQI instrument was initially developed for psychiatric practice and research, our study provides additional evidence that it has good psychometric properties and excellent comparability of score results with other published studies for different patient populations. For the TMD patient population, the results obtained from the PSQI questionnaire can continue to be reported in the form of one global score.
Competing interests
The authors declare that they have no competing interest.
Authors’ contributions
KRS designed the study, extracted data, performed the statistical analysis, interpreted the results, and drafted the manuscript. MTJ contributed to the study design, contributed to the data analysis, interpretation of the results, and to the revision of the manuscript. DB and MJH contributed to the interpretation of the results and revised the manuscript. ELS collected data, contributed to the interpretation of the results, and revised the manuscript. All authors read and approved the final manuscript.