Introduction
Healthy sleep requires adequate duration, good quality, appropriate timing and regularity, and the absence of sleep disturbances or disorders. Individual variability in sleep need is influenced by genetic, behavioural, medical, and environmental factors. Sleep quality is considered to be of utmost importance as it contributes to enhanced functioning, psychological well-being and improved quality of life. It plays a very important role in ensuring the wellness of the human body both physically as well as mentally. Current evidence supports the general recommendation for obtaining 7 or more hours of sleep per night to promote optimal health among adults aged 18 to 60 years (Watson et al. 2015). Sleeping less than 7 hours per night is also associated with impaired immune function, increased pain, impaired performance, increased errors, and greater risk of accidents (Laugsand et al. 2014; Leger et al. 2014).
Poor sleep quality is associated with a variety of negative consequences, including health-related problems such as obesity, diabetes, hypertension, and stroke (G.M. Barthlen & C. Stacy 1994; Buysse 2005b; Buysse et al. 2008; Soldatos & Paparrigopoulos 2005a). Besides that, physical (Guallar-Castillón et al. 2014) and mental aspects (Palmer & Alfano 2017) also associated with sleep quality such as muscle pain and depression (G. M. Barthlen & C. Stacy 1994; Buysse 2005a; Buysse et al. 2008; Soldatos & Paparrigopoulos 2005b). Poor sleep has been shown to increase the negative emotions, and modify the ways in which individuals understand and express (Palmer & Alfano 2017). There were few findings from different countries of South Asia showed that individual with mental disorder experienced poor sleep quality and excessive daytime sleepiness (Haregu et al. 2015; Kang et al. 2013; Theorell-Haglow et al. 2006). On the other hand, a significant body of research has investigated the effects of physical activity on sleep (Kredlow et al. 2015). Reid et al. (2010) reported that physical activity, aerobic exercise successfully improved sleep quality, reduce depression symptoms and also daytime sleepiness.
There are few objective and subjective assessments for measuring sleep quality being used in clinical and non-clinical purpose. Polysomnography (PSG) and actigraphy are objective assessments widely used for assessing sleep quality. PSG is a ‘gold standard’ method to accurately assess the quality and quantity of sleep (Appelhans et al. 2013; Banhiran et al. 2013) as well as actual physical and physiological processes during sleep activity periods (O'Donoghue et al. 2009). Actigraphy is a useful and reliable instrument used to estimate sleep-wake schedules by measurement of activity in a variety of populations, particularly in healthy young people (Sato et al. 2014). Subjective assessments of sleep quality such as Consensus Sleep Diary (CSD) (Carney et al. 2012), Subjective Sleep Quality (SSQ) Zverev and Misiri (2009), Pittsburgh Sleep Quality Index (PSQI) (Tzeng et al. 2012) are essential to assess sleep quality and the presence of sleep disturbance that can impact one’s quality of life. Subjective assessments are increasingly being used although subjective measures are difficult to aggregate and interpret, however, it is informative and most of the assessments have benchmark a standard or point of reference in order to compare the result effectively.
Pittsburgh Sleep Quality Index (PSQI) has been found to be a reliable instrument that can easily and qualitatively measure sleep disorders (Tzeng et al. 2012). PSQI has been widely used in Southeast Asia (Cheng et al. 2012) among different population which include college students (Aloba et al. 2007; Dietch et al. 2016; Guo et al. 2016; Manzar et al. 2015), older adults (Beaudreau et al. 2012; Spira et al. 2012), and individual with different medical condition (Burkhalter et al. 2011; Rener-Sitar et al. 2014; Salahuddin et al. 2017). PSQI has been translated into more than 56 different languages, including Spanish (Hita-Contreras et al. 2014), Thailand (Sitasuwan et al. 2014), Hungarian (Takacs et al. 2016), European Portuguese (Fontes et al. 2017), Italian (Curcio et al. 2013), Sinhala (Anandakumar et al. 2016) and others. There is a standardize study has been done in Malay in rural older Malaysian (Yunus et al. 2017). The aim of our study is to examine the validity and reliability of the PSQI following translation into Malay in working adult. The English version of PSQI was first translated into the Malay language before the internal consistency reliability, test-retest reliability and convergent validity were determined. MATERIALS AND METHODS
Participants
A convenience sample of healthy working staff were recruited from the Kuala Lumpur Campus, Universiti Kebangsaan Malaysia (UKM KKL). The inclusion criteria are the adult population who aged from 18 years old to 64 years old (Groeger et al. 2004). On the other hand, individual who work as a shift worker and having insomnia, cancer, depression, heart disease and pregnancy (trimester) were excluded. There were 167 staffs agreed to participate in the study. 14 of them were dropped out which is about 8% of the total subjects. An insomnia checklist is used to determine if the respondents have symptoms of insomnia. 47 of them were excluded because of meeting exclusion criteria. The final sample were 106 staffs, all of the respondents were asked to complete the questionnaires again within 2 to 4 weeks.
Translation
Translation of the PSQI into Malay by following World Health Organization (WHO) process of translation and adaptation of instruments which included forward translation and expert panel backward translation. Forward translation is the translation of the English version of PSQI into Malay by an expert panel who proficient in both English and Malay language. Two different expert panels who also proficient in both English and Malay language as well as familiar with area of health sciences were invited to identify and resolve the inadequate expressions or concepts of the translation, as well as any discrepancies between the forward translation and the English version of PSQI. Backward translation was then performed by another expert panel who has no knowledge of the questionnaire. Comparison between the original English and the backward translated English version were carried out by translation team. Discrepancies were discussed between translation team and adjustments were made. The final version of the PSQI-M was then tested in a pilot study on the target population. 10 staff who were working in UKM KKL were recruited in the pilot study. They were asked about any word they did not understand as well as any word or expression that they found unacceptable or offensive after they complete the questionnaires. The final version was produced after discrepancies were also discussed between translation team and adjustments were made.
Main Instrument
Pittsburgh Sleep Quality Index Malay Version (PSQI-M)
PSQI is a self-rated questionnaire published by Daniel J. Buysse (Buysse et al. 1989a) that can provides a brief, clinically useful assessment of multiple sleep disturbances.. It is composed of 19 questions that evaluate retrospective sleep quality and disturbances within the past month. The 19 questions will then categorize into seven sleep components related to sleep habits including sleep quality (1 item), sleep latency (2 items), sleep duration (1 items), sleep efficiency (3 items), sleep disturbance (9 items), sleep medication (1 item), and daily dysfunction (2 items). All components were then summed to yield a global PSQI scores ranging from 0 to 21. Respondents with a PSQI global score of greater than 5 were classified as poor sleepers. Those with a score of 5 or less were classified as good sleepers (Buysse et al. 1989a).
Secondary Instruments
Epworth Sleepiness Scale Malay Version (ESS-M)
ESS is a self-rated questionnaire published by Dr Murray Johns (Johns 1991). The ESS is an eight-item self-report measure of excessive day-time sleepiness. Respondents indicate, on a four-point Likert-type scale (0 = never, 3 = high chance), the likelihood that they will ‘doze off or fall asleep’ in eight different conditions such as riding as a passenger in a car. The responses are summed to yield a total score from 0 to 24, with higher scores indicating greater sleepiness during common daily activities. The cut-off score presented by an recent study showed that score above 10 indicate high sleepiness (Buysse et al. 2008). Validity and reliability was tested in pilot study. Face validity is performed by expert panel to make sure the items are clear and understandable for public. ESS-M has acceptable internal consistency, Cronbach’s alpha α = 0.731 and fair test-retest reliability Intraclass Correlation Coefficient, ICC = 0.596.
Fatigue Symptom Inventory Malay Version (FSI-M)
The Fatigue Symptom Inventory (FSI) first published in 1998 and established in women with a diagnosis of breast cancer and women with no history of cancer (Hann et al. 1998) Since its development, the FSI has been used to assess fatigue in clinical and nonclinical populations and as an outcome measure in clinical studies. This questionnaire is a 14-item self-report measure designed to assess fatigue intensity (4 items), fatigue interference with quality of life (7 items), fatigue frequency (2 items) and the daily pattern of fatigue (1 item). A score of 3 or greater on the average fatigue score is the recommended cut-off for discriminating cases of clinically meaningful fatigue (Donovan et al. 2008). Face validity is performed by expert panel to make sure the items are clear and understandable for public. FSI-M has excellent internal consistency, Cronbach’s alpha α = 0.973 and substantial test-retest reliability Intraclass Correlation Coefficient, ICC = 0.823.
Insomnia Checklist
Insomnia Checklist was used to identifying two outcomes of interest, poor sleep and probable insomnia. The insomnia checklist contains of 4 questions which included difficulty falling asleep, staying asleep, unable to fall asleep again after woke up in the midnight, and the need to take sleep medication. It is four-point Likert-type scale (0 = never, 3 = always), and choices for sleep medication are yes or no. The one who score 2 items and above for 2 scores and having sleep medication will be excluded in this study. This Insomnia checklist has excellent internal consistency, Cronbach’s alpha α = 0.842.
Statistical Analyses
Data were analysed by the Statistical Package for Social Sciences for SPSS version 22.0 for Mac OS X. Cronbach’s alpha values were computed to determine internal consistency of PSQI-M, ESS-M and FSI-M. Test-retest reliability was evaluated by Intraclass Correlation Coefficient (ICC). The convergent validity of PSQI-M was examined with FSI-M and ESS-M to explore the relationship between them. Non parametric data was used to compare the score of three assessments between gender, age and Body Mass Index (BMI) using Mann-Whitney test and Kruskal-Wallis test. The level of statistical significance was set at p < 0.05.
RESULTS
Sociodemographic data of total respondents N=106 was presented in Table 1. Most of the respondents were Malay due to the amount of adult workers in UKM KKL were also Malay. There were 101 Malay, 3 Chinese and 2 Indian. 57 (54%) of them are female and 49 (46%) of them are male. The mean age for 106 respondents are 35.3 ± 7.6 years old, 23 (22%) of them are early adult and 83 (78%) of them are middle aged. The mean weight and height are 71.1kg ± 15.8kg and 1.7m ± 0.1m. The data showed most of the respondents are overweight, 78 (74%) while only 8 (7%) of the respondents are underweight.
Internal Consistency
Table 2 shows the internal consistency, Cronbach’s alpha coefficient if item deleted for pre and post-test. The Cronbach’s alpha, ∝ = 0.33 (pre) and 0.41 (post) for PSQI-M indicating low internal consistency reliability. There was no improvement in internal consistency for PSQI-M when items were deleted.
Test Retest Reliability
Test-retest results for PSQI-M was displayed in table 3. Intraclass Correlation Coefficient (ICC) was used to examine the test-retest reliability between pre and post-test for total 106 respondents. The Intraclass Correlation Coefficient for global PSQI score was 0.50 (95% CI 0.35-0.63) indicating fair test-retest reliability. The highest Intraclass Correlation Coefficient for PSQI Malay subcomponent was Sleep Duration, 0.62 (95% CI 0.50-0.73) indicating moderate test-retest reliability and the lowest Sleep Quality, 0.24 (95% CI 0.05-0.41) indicating slight test-retest reliability.
Convergent Validity
Table 4 shows the convergent validity for PSQI-M with ESS-M and FSI-M. The global PSQI score show significant correlation (p < 0.01) with ESS-M and FSI-M. The Pearson correlation coefficient, r for each significant component for PSQI-M with ESS-M and FSI-M points to an adequate convergent validity. Fatigue Interference of FSI-M and Daily dysfunction of PSQI-M had the highest r value, r = 0.50 (p < 0.10) while Sleep Quality of PSQI-M and ESS-M had the lowest r value, r = 0.19 (p < 0.05).
Comparison of Mean
The result (table 5) showed significant correlation between age, sleep duration, sleep disturbance, ESS overall score, fatigue intensity and FSI overall score. Non parametric test was perform and the result showed that gender and BMI did not show any significant correlation with PSQI-M, ESS-M and FSI-M.
DISCUSSION
PSQI is a widely used assessment in research and clinical practice, providing sleep quality information on respondent. It can discriminate “good” and “poor” sleepers, and in clinical assessment of a variety of sleep disturbances. The present study was performed to assess the internal consistency, test-retest reliability and convergent validity of PSQI Malay version, which was translated into the Malay language using World Health Organisation (WHO) process of translation and adaptation of instruments.
PSQI-M was easy to be complete in around 10 minutes when respondents answer the questionnaire, as supported by few studies which suggested that PSQI can be successfully self-administered in around 10 minutes for clinical and non-clinical population (Beck et al. 2004; Buysse et al. 2006; Hita-Contreras et al. 2014; Tzeng et al. 2012). However, previous study (Farrahi Moghaddam et al. 2012) suggested to give some explanation before the administration. Similar situation happened where the respondents of pilot study complained that there was confusion between question 1 and question 4 which is “pada pukul berapa biasanya anda masuk tidur?” and “berapa jumlah jam tidur sebenar yang anda biasa perolehi pada waktu malam? (Ini mungkin berbeza dengan jumlah jam yang anda luangkan di atas katil.)”. Therefore, some of the respondents tend to answer the questionnaire more than 10 minutes. We agree that it is not necessary for trained staff to be present to assist in the completion of the questionnaire as reported by Kotronoulas et al. (2011) but a staff who has knowledge about this questionnaire has to be there to give explanation to clarify both of these questions and give example to make them understand about the questions. Other than that, question 1 to question 4 are not Likert scale like question 5 to question 9. Some of the respondents answered these questions with time range such as 15 to 30 minutes instead of providing an exact time. In such cases, the average of the time range was taken as the final responds but it might cause instability and poor consistency of the questionnaire.
For the sociodemographic data, there was no significant correlation of gender and BMI between the three assessments. Similar result were showed by (Mollayeva et al. 2016), a meta-analysis of PSQI original version that lack of consistency in gender differences between the studies. Therefore, gender and BMI will not affect the result of PSQI-M. Mean age in this study was 35.28 ± 7.59 years old and 78.3% of them are middle aged adult. The youngest subject is 20 years old and the oldest is 57 years old. This data showed that the staffs who working in UKMKKL mostly are middle aged adult which is 30 to 64 years old. There was significant correlation between between age and the three assessments used in this study which were PSQI-M, ESS-M and FSI-M, indicating middle aged adult has shorter sleep duration, more sleep disturbance, greater daytime sleepiness, and more intensive fatigue. Result supported by (Van Cauter et al. 2000), reported that healthy young adult subjects having better subjective and also objective sleep quality on average than healthy elderly subjects. Study also showed that sleep parameters significantly change with age where percentage of stage 1 and 2 sleep increases with age and percentage of rapid eye-movement (REM) sleep decreases with age in adults (Ohayon et al. 2004). In sum, age may affect the result of sleep quality, further studies may need to have same amount of subjects for different age groups in order to maintain the stability and reliability of the questionnaire regarding sleep quality.
The results of the present study show acceptable but low internal consistency, with a Cronbach’s alpha for PSQI-M ∝ = 0.33 – 0.40. Similar result found in PSQI Japanese version, ∝ = 0.43 for control group (Doi et al. 2000). Studies that included control group and clinical group in their study will have different value between their internal consistency reliability where the Cronbach’s Alpha of control group will be lower than the clinical group such as PSQI Thailand version, 0.50 for control group, 0.77 for clinical group (Sitasuwan et al. 2014), PSQI Hungarian version, 0.78 for clinical group (Takacs et al. 2016) and Hebrew version 0.52 for control group, 0.70 for clinical group (Shochat et al. 2007). The similarities between previous studies and this study was those who were taking sleep medication were excluded, so the sleep medication subcomponents score had zero variance. Therefore, sleep medication component was excluded when calculating Cronbach’s Alpha. Also, they have large variances in the component means which might be an effect of the population variance on reliability, as reported by Sitasuwan et al. (2014) and Doi et al. (2000). The low value of the Cronbach’s alpha of PSQI-M may also due to confusion of the time frame used. In this study, three instruments were used and three of them were measuring different time frame. Furthermore, sleep quality might be different for each month or each week. Therefore, time frame is important for an individual to determine his or her sleep quality for the particular time frame. The same time frame had to be measured in order to maintain the high stability and reliability of the questionnaire. Also, the population included in this study were only good sleepers who has good sleep quality. Original version of PSQI was designed to assess the clinical population (Buysse et al. 1989b). It has overall high internal consistent in other studies that included clinical sample, as reported by a systematic review study (Mollayeva et al. 2016). PSQI provides a standardized questionnaire of measuring sleep quality that quickly identified good and poor sleepers where global score > 5 provided a sensitive and specific measure of poor sleep quality. PSQI might not sensitive enough to compare the sleep quality between good sleepers.
As for the test–retest reliability assessment, the questionnaire was administered again on 106 staffs with 2 to 4 weeks which had the similar time interval to other studies (Fontes et al. 2017; Kotronoulas et al. 2011; Sitasuwan et al. 2014). The result showed fair test-retest reliability ICC, 0.503 (95% CI 0.345-0.633) in this study. This value is lower to that found by European Portuguese version, ICC = 0.76 (95% CI 0.72-0.89), Greek version, ICC 0.82 (p <0.001) (Kotronoulas et al. 2011) and Thai version, ICC 0.89 (95% CI 0.85-0.92) (Sitasuwan et al. 2014). This study had a fair test-retest reliability might due to lack of sleep quality control of subjects. Sitasuwan et al. (2014) conducted test-retest reliability with the control group who able to have a regular sleep pattern during the research. This study only used an Insomnia checklist to recruited good sleepers in order to retain a good reliability of the study. The one who has difficulty in doze off, difficulty in maintain sleep and get back to sleep after woke up will be excluded in this study. However, this might to not enough to maintain the sleep quality within different weeks. Sleep quality may be affect due to different external or internal factor such as work overload and also illness during different weeks (Knudsen et al. 2007). A good sleep quality had to be maintain during the research by controlling a regular sleep pattern of the respondents, as done by other research (Sitasuwan et al. 2014).
Our analysis demonstrated an adequate convergent validity with the ESS-M and FSI-M. The global PSQI score significantly correlated with ESS-M and FSI-M which indicate an individual who has poor sleep quality will also have intensive sleepiness and also high level of fatigue. Similar observation were made by (Khosravifar et al. 2015; Kotronoulas et al. 2011; Siegel et al. 2006). The study between PSQI Greek version and ESS Greek version showed poor correlation, r = 0.30; p = 0.02 Kotronoulas et al. (2011) and PSQI Persian version and ESS in patients with depression also showed poor correlation, r = 0.264; p = 0.011(Khosravifar et al. 2015). Studies suggested that possibly ESS and PSQI are indicating the different goal ESS is to assess more habitual than actual symptoms of sleepiness Kotronoulas et al. (2011) and PSQI is less sensitive to condition that produce excess daytime sleepiness compared to insomnia (Khosravifar et al. 2015). Previous studies have shown that PSQI is measuring arguably primarily a specific condition, insomnia which is highly correlated with daytime fatigue but not with excessive daytime sleepiness (Riedel & Lichstein 2000). In this study, fatigue and daytime sleepiness were both having poor but significant correlation with PSQI. Siegel et al. (2006) reported significant but poor correlation between PSQI and FSI, r = 0.40. There are limited study discussing about the correlation between PSQI and FSI although studies discuss about the sleep quality and fatigue.
There are potential limitations to this study. First, the sample size was a limitation and future validation studies on the Malay population with larger sample size are needed. Current studies always comparing control group and clinical group. Further studies may also include clinical sample for example sleep disorder population to perform comparison of their sleep quality. Second, the samples included for these studies were from a single institution, and may not be generalizable to staffs at institutions in other regions or with other racial or socioeconomic make-ups. Furthermore, a convenience sampling methodology was used which is inferior to a consecutive sample and random sampling. Lastly, instruments used in this study measured sleep and symptoms are not during the same time frame. For example, the PSQI measured sleep quality during the last month; however, the ESS measured daytime sleepiness generally and FSI measured level of fatigue during last week. This may cause confusion to respondents when they are answering the questionnaires.
CONCLUSION
The PSQI-M has acceptable and fair reliability as well as adequate convergent validity with ESS-M and FSI-M. It is a valid and reliable questionnaire to be use for assessing the sleep quality information on healthy population Malaysian context. Further validation test such as identify optimal cut-off point to identify “good” and “poor” sleeper is essential to strengthen the psychometric properties of PSQI Malay version.
ACKNOWLEDGEMENT
The authors are thankful to MAPI Research Trust for kind permission to translate the PSQI into the Malay language. The authors would like to thank all the expert panel their precious time for translating the questionnaires. Finally, many thanks to staff at UKM KKL who participated in the present study.