Assessment of the Psychometric Properties of the Holland Sleep Disorders Questionnaire in the Iranian Population

Background Assessing sleep disorders and understanding their causes are essential for the proper treatment and management of the disorders. The Holland Sleep Disorders Questionnaire (HSDQ) is a self-assessment questionnaire that measures sleep problems and symptoms based on the six categories of sleep disorders described in the International Classification of Sleep Disorders-2 (ICSD-2). The aim of this study was at validating and assessing the psychometric properties of the HSDQ in Iranian adults. Method The study was carried out as a methodological and validation work. The guidelines for translation and cultural adaptation of patient-reported outcome measures were followed for the translation and the cultural validation of the tool. To examine construct validity, exploratory factor analysis (EFA) with 216 participants and confirmatory factor analysis (CFA) with 355 participants were used. As to the reliability, the test-retest method and, as to internal consistency, Cronbach's alpha were employed. Data analyses were done in SPSS-25 and LISREL-8. Results The CFA and EFA results confirmed the tool with six factors and 31 items. The R2 index of the model was 0.99, which indicated that 99% of changes in the dependent variable (adults' sleep problem) were attributed to the independent variable (the 31 items). In other words, 0.99 of the changes in the dependent variable were due to the independent variables. The main indices of CFA (χ2/DF = 2.65, CFI = 0.91NNFI/TLI = 0.92GFI = 0.81, REMSEA = 0.043, R2 = 0.99) were acceptable. In addition, a correlation coefficient below 0.05 was considered as significant. Reliability of the tool based on internal correlation (Cronbach's alpha) was in the 0.701–0.924 range for the subscales and equal to 0.789 for the whole tool. Conclusion In general, the results showed that the Farsi version of HSDQ (six factors and 31 items) had acceptable and applicable indices and it can be used as a valid tool in the Iranian society. The tool can be used as a reliable tool in different fields of medical sciences.


Introduction
Sleep is a process that is essential for maintaining brain function, and the lack of it can lead to memory and attention impairments [1]. Sleep problems are hazardous for health and treating them is very expensive [2]. Sleep disorders are comorbid of other diseases such as an increased risk of obesity, diabetes, hypertension, tachycardia, and stroke [3]. Sleep disruptions have short-term health consequences (somatic pain, emotional distress, mood disorders, and problems [6] such as somatic pain, emotional distress, and weight-related issues [4]. The lack of balance in sleep and rest process can create excessive fatigue and nervousness [7]. Studies based on questionnaires have shown the high prevalence of sleep disorders in many countries [8][9][10][11][12]. All around the world, these disorders cause negative effects on one's quality of life [12]. Epidemiological studies have shown that about 36% of the total adult United State (US) citizens sleep less than seven hours overnight [13]. A multinational study reported that the prevalence of sleep problems in the United State of America (USA) was the highest (56%), followed by European countries (23-26%) and Japan (23%) [14].
There is a fast-growing number of studies on sleep epidemiology [8,10,12], which is due to the gradual increase in public awareness about the negative effects of inadequate and irregular sleep on human error and health [15]. Therefore, there is a need for collecting correct and reliable data about the prevalence of sleep disorders and also answering questions about epidemiological data and general health [10].
In general, sleep disorders are assessed using polysomnography, actigraphy, and questionnaires (e.g., Pittsburgh Sleep Quality Index, Sleep Hygiene Index, and Insomnia Severity Index) [16][17][18]. Holland Sleep Disorders Questionnaire (HSDQ) is a credible questionnaire based on the International Classification of Sleep Disorders-2 (ICSD-2) [19]. The questionnaire measures sleep problems of patients in six main categories of sleep disorders. These six categories are insomnia, Circadian rhythm sleep disorders (CRSD), parasomnia, hypersomnia, restless legs syndrome (RLS)/periodic limb movement disorder (PLMD), and sleep-disordered breathing (SDB). Using the scale, a physician can evaluate which one of these sleep disorders has inflicted the patient [10,20].
Given the introduction about the problems, costs, and high prevalence of sleep disorders, one of the ways to collect useful information to attenuate and treat the disorders is epidemiological study. There is no valid and integrated tool for Iranian populations to measure sleep disorders. By searching for specific researches and texts related to sleep, we were convinced that there was no suitable tool to assess sleep problems in different categories as insomnia, CRSD, parasomnia, hypersomnia, RLS/PLMD, and SDB. Although the HSDQ is not based on an updated version of the classification of sleep disorders (ICSD-2), considering that the items match the Iranian culture and the classification of sleep disorders into six categories, it was used in this study, and along with cultural validation, the psychometric properties of the HSDQ in Iranian adult were also evaluated. Taking into account that HSDQ has proper items and categorizations and it is a simple and inexpensive tool to use, it can be a good option for the Iranian population. Therefore, the present study is an attempt to examine psychometric properties of HSDQ in the Iranian population.

Setting.
The study was carried out as a methodological and validation work.

Participants.
The study population consisted of adults in age range of 18-85 living in Kermanshah city. The characteristics of the participants are presented in Table 1.
To perform cluster sampling, six urban districts were selected among 10 urban districts of the city. Then, three districts were randomly selected and six clinics (private or public) were selected randomly from these three districts. The participants were selected based on a set of inclusion criteria among the persons who had a sleep disorder file in the clinics (n = 230). Inclusion criteria were residence in Kermanshah city, reading and writing literacy, interested in participation, no dependence on psychedelic and narcotic drugs, and no physical and psychological problems in the last six months, except for sleep disorders. In addition, 300 individuals without sleep problem who visited the clinics for skin and hair problems were selected. The sample size for face validity included 10 eligible adult persons (with a sleep disorder file) in Kermanshah city, 12 university professors for the construct validity phase, 216 individuals for exploratory factor analysis (EFA), and 355 individuals for confirmatory factor analysis (CFA) phases Content validity ratio: the content validity ratio (Lawshe) is one of the earliest and most widely used methods for quantifying content validity. b Content validity index is the most widely used index in quantitative evaluation. c Skewness is a measure of symmetry or, more precisely, the lack of symmetry. d Kurtosis is a measure of whether the data are heavy-tailed or light-tailed relative to a normal distribution.

Sleep Disorders
(a large number of the questionnaires were not returned or returned not completely filled out).
Given the limitations of the COVID-19 pandemic, the questionnaires were filled out through a blended method. So that hard copies were provided to the individual who had an easy access to the clinics (for EFA, 118 electronic and 112 hard copies were sent, and for CFA, 204 electronic and 151 hard copies were sent) and for others, an electronic version of the tool was sent to them via email or using WhatsApp (an electronic questionnaire link was sent). In this study, an electronic questionnaire was sent to 540 people (86 people via email and 454 people via WhatsApp), out of which only 248 returned and only 204 were usable. 2.4. Cultural Validity. The guidelines for translation and cultural adaptation of patient-reported outcome measures were followed for the translation and the cultural validation of the tool [21]. The tool was translated independently by two native speakers from English into Farsi. The results were examined by a panel of experts in the presence of the research team members, and one version was developed out of the two translations. Afterwards, the Farsi version of the tool was backward translated into English by two other translators. After revising the translation works, a cultural comparison process was performed. The two translations were compared with the original version to make sure of conceptual equivalence of the two translation works and the original version. Eventually, the final version of the translated tool was sent to the developer of the tool for confirmation. To examine cognitive equivalence, the final translated copy was provided to 10 adult persons with sleep disorder in the sleep disorder clinics in Kermanshah city to examine their ability to comprehend, interpret, and understand the items. The tool was revised based on the cognitive findings to make sure of cultural comparability. Eventually, the final translation was reviewed to spot and remove any grammar error or typos.

Face Validity.
To check face validity, the scale was provided to another 10 adult persons, and in face-to-face interviews, they were asked to highlight any vague item and word or ambiguity or wrong perception in the text.
2.5.3. Content Validity. As to content validity, the scale was provided to 12 researchers, members of faculty boards, and experts in pertinent fields for revising and giving opinions. Through this, content validity was conducted qualitatively [22]. To determine quantitative content validity, the content validity index (CVI) was used based on the Walts and Bassel index [23] on all indices ( Table 2).

Construct Validity.
In this study, EFA and CFA methods were used to confirm the construct validity. In each stage of EFA and CFA, normal distribution of the data was checked using the multivariate test.
2.5.5. Multivariate Normality Data. The skewness value for each statement varied from −1.09 to 1.86, and it was at a (−2, 2) range. This means that the statements are normal in terms of skewness with symmetric distribution [24]. Moreover, kurtosis ranged from −1.7 to 2.7 ( Table 2). Normal distribution of the data in each stage of CFA was checked using skewness and kurtosis. The validity of the model was examined based on the factor load of each item (for t value > 1.96, significant level = 95%; and for t value > 2.576 & 3.29, significant levels were 99% and 999%, respectively).
2.5.6. Internal Consistency. To examine fitness of the model, the maximum likelihood method was adopted. In addition, to check tool reliability, internal consistency was used using Cronbach's alpha for each item and then for the whole tool.

Explorative Factor
Analysis. EFA was conducted with 216 participants. To make sure of the adequacy of the participants to conduct EFA, the Kaiser-Meyer-Olkin (KMO) test was used (0.741). To examine correlation of the items, Bartlett's test was used (chi-square = 4393:152, degrees of freedom = 796, p value = 0.0001). The p value for significance of Bartlett's test was less than 0.05. Given the results and significance level, performing EFA on this questionnaire was acceptable [25].
Having checked the requirement for EFA, principal components and varimax rotation were used to extract factors. To determine the number of factors, the three following rules were followed.
(1) The factors with Kaiser's criterion (or eigenvalue) are higher than 1 in scree plots and Horn's parallel analysis [26].
(2) The factors corresponding to the actual eigenvalue higher than the parallel random eigenvalue were accepted in Horn's parallel analysis. In addition, the factors with actual eigenvalue less than or equal to the mean value of the parallel random eigenvalue were removed as sampling error [27].
(3) The items with factor load > 0:3 and higher were loaded on the items under consideration [26].
According to these three rules, the primary results showed seven factors with an eigenvalue > 1 for the analysis. In addition, one factor was removed based on Horn's parallel analysis and EFA was repeated with six fix factors (Table 3 and Figure 1).
In general, the results showed ( Table 4) that six factors elaborated 62.918% of the variance of 32 items. That is, 14.835% of the variance was attributed to factor 1, 12.509% was attributed to factor 2, 11.466% was attributed to factor 3, 10.8% was attributed to factor 4, 7.624% was attributed to factor 5, and 5.684% was attributed to factor 6. Annexed Table 1 or the rotated factor matrix lists the factor load of each item based on the four factors.
In general, the EFA results confirmed six factor loads [20]. However, the factors' content did not necessary overlap with the constructs of the main article, which was not a necessity condition either. As listed in Annexed Table 1, item no. 14 was loaded on the RLS/PLMD factor, while in Kerkhof et al.'s study [20], it was loaded on insomnia. Eventually, the factors and items were allocated, and to measure internal consistency of the tool, Cronbach's alpha was used (alpha ≥ 0:7 was acceptable, and alpha ≤ 0:5 was unacceptable) [28] (Table 5).   Table 5 and Figure 2 demonstrate CFA in standard condition without coefficient.
As the results showed, none of the items were removed except for item 26 of the CRSD factor as its factor load was less than the critical value (±1.96). In addition, Table 6 lists the goodness of fit indices of CFA. Given the goodness of fit indices listed in Table 6, the model has an acceptable goodness of fit and it fits the collected data.
The reliability of the tool was determined through the test-retest method with participation of 15 participants (not among the main group of participants) who filled out the tool twice with a 10-day interval, and the correlation coefficient was obtained equal to 0.875.

Internal Consistency.
To examine internal consistency (internal reliability) of the tool, Cronbach's alpha was obtained for the whole tool with 31 items equal to 0.789. Cronbach's alpha for the subscales of the tool was in the 0.701 and 0.924 range. Therefore, the subscales had the reliability to measure the variables ( Table 5).
As listed in Table 7, correlation coefficients between HSDQ and its subscales were positive and significant in all cases. Eventually, based on EFA and CFA, the Farsi version of HSDQ was confirmed for Iranian adults' society with 31 items and six subscales.

Discussion
The HSDQ was translated into Farsi, and its psychometric properties were examined in the Iranian population. The  [21] were followed for the translation and the cultural validation of the tool. As to construct validity, EFA was carried out on 216 participants, and afterwards, the number of participants was increased to 355 for CFA.
The EFA results showed that 62.98% of the variance of the 32 items was attributed to the eight factors, and actu-ally, 32 items and six factors were confirmed. Kerkhof et al. analyzed the tool using principal component analysis (PCA) and confirmed it with six factors and 32 items as well [20].
The results showed that item no. 14 was loaded on RLS/ PLMD and the factor encompassed six items. However, item 14 in Kerkhof et al.'s questionnaire was placed on the insomnia factor and the RLS/PLMD factor (subscale) had five items [20]. To explain the differences, cultural and traditional differences in the study populations, number of participants, and social condition (e.g., COVID-19 pandemic) that has created severed stresses in Iranian society is notable. At any rate, with a closer look at the item, it is clear that the concept of the item has consistency with the RLS/PLMD factor. Item 14 emphasizes the problems that the person will have in the next day due to ill health and factors such as fatigue, sleepiness, bad mood, poor concentration, memory problems, and lack of energy, in this regard, it can be emphasized that because some of sleep disorders, such as PLMD, occur in deep sleep, the individual is not aware that they have such a disorder. These cases are usually associated with morning symptoms such as drowsiness, fatigue, low mood, and concentration disorders [29]. People with PLMD may also experience sleep deprivation during   [20]. However, in the present study, EFA was used to confirm the validity of the structure, followed by CFA. The CFA is widely used for confirming factors and items [31,32]. To explain the results, it is notable that the responses of the participants to the items were affected by cultural condition, age, sleep hygiene, life style, and mental condition; therefore, these variables affect the response to all of items in scale.
In addition, reliability of the tool was equal to 0.875 with Cronbach's alpha equal to 0.789 (0.701-0.924). These results confirmed reliability and stability of the tool for the target population. In Kerkhof et al.'s study, Cronbach's alpha was equal to 0.9 (0.73-0.81) [20], which is consistent with the present study.
The correlation coefficient of HSDQ and the subscale were positive and significant in all cases ranging from 0.21 for Parasomnia to 0.75 for RLS/PLMD. The Pearson correlation results in Kerkhof et al.'s study ranged from 0.73 (SDB) to 0.81 (CRSD) [20]. As to the differences between the two studies, differences of the study population and the participants are notable. In Kerkhof et al.'s study [20], the sample size was approximately four times the sample size in the present study. In addition, lifestyle, social interactions, sleep hygiene, monthly income, diet, and even lifestyle are completely different in the two communities studied and these can affect the response to items in the two communities.
Two different groups of participants were selected for EFA and CFA is one of the advantages of this study. In addition, the study design, translation, and cultural validation were based on the ten steps of Wild et al. [21]. Confirmation of the content validity of the tool quantitatively and qualitatively is also among the strengths of this study.
It is notable that the questionnaire is based on ISCD-2 and it is older than ISCD-3. The research team is determined to prepare a revised form of this questionnaire for the Iranian adult community in the future.

Limitation
Because of the high prevalence of COVID-19 and limitations of finding participants, individuals with and without confirmed sleep disorders were selected. In addition, the HSDQ was administered through the blended method (hard copy and e-form sent via email and WhatsApp account) again because of the limitations caused by the COVID-19 pandemic. The HSDQ represents a screen questionnaire that does not show sleep disorders very accurately. It is also based on the ICSD-2 and is older than the ICSD-3, although the two versions do not differ much in diagnostic classes. And differences in some subclassifications.

Conclusion
The results indicated that the Farsi version of HSDQ with six factors and 31 items had acceptable indices and applicability for the Iranian society. The tool can be used as a reliable tool in different fields of medical sciences. In general, it can be said that in this study, the HSDQ was validated using appropriate cultural validation methods. Formal validity, content validity quantitatively and qualitatively, structural validity, internal validation, and instrument stability have been performed using standard methods, and finally, this scale has been validated and studied in the Iranian society. Therefore, it can be said that this tool can be used to assess the status of sleep disorders in the Iranian adult community.
Abbreviations HSDQ: Holland Sleep Disorders Questionnaire CVI: Content validity index CVR: Content validity ratio KMO: Kaiser-Meyer-Olkin EFA: Exploratory factor analysis CFA: Confirmatory factor analysis TLI: Tucker-Lewis index NFI: Normed fit index GFI: Goodness of fit index CFI: Comparative fit index R 2 : Root mean square