This study aimed to translate and adapt the Pittsburgh Sleep Quality Index (PSQI) for Hong Kong Chinese cancer survivors and examine its psychometric properties and factor structure.
A cross-sectional study was conducted to validate the Chinese version of the PSQI. The study was carried out in an outpatient clinic of a public acute care hospital in Hong Kong.
Hong Kong Chinese childhood cancer survivors who had outpatient medical follow-ups were invited. Eligible survivors were aged 6–18 years, and able to communicate in Cantonese and read Chinese. We excluded survivors whose medical records identified cognitive or behavioral problems. To ensure that all participants could understand the questionnaire, we included only those older than 6 years.
To assess the known-group validity, 50 pediatric cancer patients and 50 healthy counterparts were recruited using the same inclusion and exclusion criteria.
There is no universal guideline on the minimal sample size required for confirmatory factor analysis. Nunnally  suggested that at least 10 subjects are required for each item. As the PSQI contains 19 items, we planned to recruit at least 380 childhood cancer survivors.
(1) Demographic and clinical characteristics
A structured questionnaire was used to record participant demographic and clinical characteristics.
(2) The Chinese version of the Pittsburgh Sleep Quality Index (PSQI)
The 19-item PSQI assesses subjective sleep quality in the last month. Items are categorized into seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, sleep medication, and daytime dysfunction. The possible score range for each component is 0 (no difficulty) to 3 (severe difficulty). The seven component scores are summed to produce a global score; higher scores represent poorer subjective sleep quality. Previous psychometric studies of the English version of the PSQI have confirmed that it is a reliable and valid instrument for assessing subjective sleep quality among adult cancer survivors . The factor structure of the PSQI has also been widely examined [15,25,26]. Results vary according to sample characteristics [15,25,26]. Although Buysse et al. suggested a single-factor model underlying the PSQI in depressive patients, Cole et al.  proposed a three-factor model in older adults and Magee et al.  identified a two-factor model in Australian adults.
(3) The Chinese version of the Center for Epidemiologic Studies Depression Scale for Children (CES-DC)
The CES-DC assesses depressive symptoms in people aged 6–17 years. It comprises 20 items measured on a 4-point scale. Higher scores represent more depressive symptoms. This is a reliable and valid instrument for children .
(4) The Chinese version of the Fatigue Scale-Child (FS-C)/Fatigue Scale-Adolescent (FS-A)
These two scales measure cancer-related fatigue levels in pediatric oncology patients. The FS-C comprises 13 items for patients aged 7–12 years; the FS-A contains 12 items for patients aged 13–18 years. Higher scores represent higher levels of cancer-related fatigue. They are reliable and valid measures of cancer-related fatigue among Hong Kong Chinese childhood cancer survivors [28,29].
(5) The Chinese version of the Pediatric Quality of Life Inventory 4.0 Generic Core Scale (PedsQL 4.0)
The PedsQL assesses health-related quality of life of people aged 2–18 years. This scale comprises 23 items categorized into four different subscales: physical functioning, emotional functioning, social functioning, and school functioning. Subscale scores are obtained by averaging item scores for the corresponding subscales; the total scale score is the average of all item scores. Higher PedsQL scores indicate better quality of life. The psychometric properties of this scale among Chinese children have been well established .
An expert panel was set up for translation. The panel contained an associate professor, two assistant professors, a research assistant professor, a postdoctoral fellow, a pediatric oncologist, and a nurse specialist. The procedures suggested by Bracken and Barona  were followed. The PSQI was translated from English into Chinese by two independent bilingual translators. The two translations were compared and reconciled after panel member discussions. Another two independent bilingual translators blinded to the original English version of the scale were asked to translate the Chinese version back into English. The panel members compared the back translations against the original English version and decided whether the meaning of each item in the back translations had been retained. Disagreement was resolved during regular meetings.
After the informed consent process, parents were invited to complete a simple questionnaire that documented their child’s demographic and clinical characteristics. Children were asked to complete the PSQI, CES-DC, FS-A/FS-C, and PedsQL. To assess test–retest reliability, a sample of 40 children were invited to respond by phone to the PSQI 2 weeks later.
For sematic equivalence, the expert panel were asked to compare the original and translated versions of the PSQI, and rate the equivalence of each translated item using a 4-point scale (from 1 = not equivalent to 4 = most equivalent). The equivalence rate was calculated as the percentage of items rated as either 3 or 4. Any item rated 1 or 2 by 20% of the experts was deemed not equivalent and was amended.
For content equivalence, the expert panel rated the relevancy of each PSQI item to the concept (i.e., subjective sleep quality) using a 4-point scale (from 1 = not relevant to 4 = very relevant). The content validity index (CVI) of an item (I-CVI) was calculated as the percentage of experts who rated the item as either 3 or 4. A CVI of the translated PSQI (S-CVI) was the average of the I-CVIs for all items. An I-CVI ≥ 0.78 or higher and S-CVI ≥ 0.9 were considered acceptable .
The internal consistency of the Chinese version of the PSQI was evaluated using Cronbach’s alpha, and the test–retest reliability using the intraclass correlation coefficient (ICC).
Convergent validity was established by identifying correlations between PSQI and CES-DC scores, between PSQI and FS-A/FS-C scores, and between PSQI and PedsQL scores. Following previous studies [10,12,13], we hypothesized positive correlations between PSQI and CES-DC scores and between PSQI and FS-A/FS-C scores, but a negative correlation between PSQI and PedsQL scores.
Confirmatory factor analysis was performed. The overall fits of the single-factor, two-factor, and three-factor models were examined and compared because previous studies report varying PSQI factor structures [15,25,26]. The overall model fit was determined using following fit indices: the χ2/degrees of freedom (df) ratio, root mean square error of approximation (RMSEA), and comparative fix index (CFI).
A cutoff score for the translated PSQI used with the survivors was determined using receiver operating characteristic (ROC) analysis. Participants were regarded as having insomnia if they met all diagnostic criteria for insomnia in Diagnostic and Statistic Manual of Mental Disorders Fifth Edition (DSM-5). The area under the curve (AUC) was calculated. Sensitivity and specificity were used to determine the best cutoff score.
Ethical approval was obtained from the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (UW 18-366). A research assistant approached all parents in the outpatient clinic, and introduced the study. After checking eligibility, the research assistant obtained written consent from parents who allowed their child to participate. These parents were ensured that withdrawal would not result in any prejudice to the care received. The children were also invited to write their names on a child assent form to indicate their willingness to participate.