Validation of the Chinese version of the Pediatric Quality of Life Inventory^TM (PedsQL^TM) Cancer Module

Abstract

Objective The psychometric properties of the Chinese version of the Pediatric Quality of Life Inventory^TM (PedsQL^TM) Cancer Module were investigated. Methods This instrument and the Generic Core Scales were administered to 359 pediatric patients with cancer (5–18 years) and 413 parents of such patients (2–18 years old). Results Seven and eight factors were, respectively, identified for the patient and parent versions. The Cronbach's alpha coefficients were respectively .89 and .92 for the total scale, and respectively .75–.90 and .76–.93 for the subscales of the patient and parent versions. Test–retest reliability coefficients exceeded .60 for most cases. The total/subscale scores of the Cancer Module significantly correlated with those of the Generic Core Scales. Some of the subscales could distinguish between on-treatment and off-treatment patients. Conclusions The psychometric properties of the patient and parent versions of the Chinese PedsQL^TM Cancer Module were found acceptable.

Health-related quality of life (HRQOL) is a multidimensional construct, assessing the physical, psychological, and social functioning dimensions of illnesses and/or treatments (Revicki, 1989). HRQOL measurements have been used as outcome measures in clinical trials and evaluation of health-related services. The Pediatric Quality of Life Inventory^TM (PedsQL^TM), which has both generic and disease-specific modules as well as patient and parent versions, is one of the very few instruments that is widely used to assess HRQOL among children and adolescents between the age of 2 and 18 years (Varni, Seid, & Kurtin, 2001). The multi-dimensional PedsQL^TM Generic Core Scales (physical, emotional, social, and school functioning dimensions) assess general HRQOL in various pediatric populations (Varni, Seid, & Rode, 1999). It can be applied to healthy children and pediatric patient populations with acute or chronic health conditions (Varni et al., 2001).

Disease-specific modules of the PedsQL^TM (e.g., cancer, asthma, diabetes, cerebral palsy) have also been developed (Varni et al., 2003, 2006; Varni, Burwinkle, Katz, Meeske, & Dickinson, 2002; Varni, Burwinkle, Rapoff, Kamps, & Olson, 2004). Except for the Cancer Module, none of the other disease-specific modules have been validated in non-English languages. Because the disease-specific modules are sensitive to specific diseases and the Generic Core Scales allow for cross-population comparisons (Varni et al., 2001), researchers are advised to use disease-specific modules and the Generic Core Scales together to provide a comprehensive assessment of HRQOL (Varni, Limbers, & Burwinkle, 2007a).

The PedsQL^TM Cancer Module (Varni et al., 2002) was tailor-made to assess the HRQOL among pediatric patients with cancer, with recommendations that both the patient version and the parent version be used in this population. Versions in English (Varni et al., 2002), German (Felder-Puig et al., 2004) and Brazilian (Scarpelli et al., 2008) have been validated. These validation studies surveyed around 100 pediatric cancer patients (2–18 years old) and parents of these patients, establishing the instrument's internal reliability, test–retest reliability (Scarpelli et al., 2008), concordance between the patient and parent versions and discriminant validity on disease severity (Felder-Puig et al., 2004) as well as discriminating between patients who were on-treatment and off-treatment (Scarpelli et al., 2008). No validation of the Chinese version of the Cancer Module has, however, been reported.

The Health Utilities Index (HUI) (Feeny, Torrance, Furlong, & Boyle, 1996) is one of the few existing Chinese pediatric HRQOL measurement tools. However, the PedsQL^TM is more sensitive than the HUI to assess decreased HRQOL after brain injury (Curran, Miller, McCarter, & Sharples, 2003) and HRQOL changes among pediatric patients with cancer undergoing chemotherapy (Banks, Barrowman, & Klaassen, 2008). Unlike the PedsQL^TM Cancer Module (2–18 years old), the HUI is a generic instrument (6–18 years old) and does not have a cancer module. Another instrument, the Quality of Life for Children with Cancer was developed in Chinese, targeting 7–18 years old cancer patients and their parents (Yeh, Chao, & Hung, 2004). It has not been translated into English and was only used in a few studies published by the developers of the scale. Because of the specificity to pediatric patients with cancer, a wider age range of applied populations, and the potential for cross-cultural comparisons, the PedsQL^TM Cancer Module is regarded as one of the ideal instruments for measuring HRQOL among pediatric patients with cancer. Therefore, it is important to validate its Chinese version.

The present study aimed at translating and validating the Chinese version of the PedsQL^TM Cancer Module among Hong Kong Chinese pediatric patients with cancer and parents of such patients, using similar methodologies employed by validation studies of other translations of the instrument (Felder-Puig et al., 2004; Scarpelli et al., 2008; Varni et al., 2002). This is also the first study investigating the factor structure of the PedsQL^TM Cancer Module in any language.

Methods

Participants and Settings

Inclusion criteria included pediatric patients who were 2–18 years old, Chinese Hong Kong residents, diagnosed as having cancer, and attending one of the three public oncology outpatient clinics in Hong Kong (there are a total of five such clinics): the Prince of Wales Hospital, the Princess Margaret Hospital, and the Queen Elizabeth Hospital. Patients with some comorbidities (e.g., leucocythemia, asthma) or major developmental disorders (e.g., autism, attention deficit hyperactivity disorder) were excluded from the study.

The main study, which investigated the psychometric properties of the Cancer Module and the Generic Core Scales, included 359 patients and 413 parents. The age group distribution of pediatric patients was: 2–4 years (55 parents), 5–7 years (70 patients and 70 parents), 8–12 years (144 patients and 148 parents), and 13–18 years (145 patients and 140 parents). The number of participants from each of the three hospitals was: the Prince of Wales Hospital (220 patients and 258 parents), the Princess Margaret Hospital (50 patients and 55 parents), and the Queen Elizabeth Hospital (89 patients and 100 parents). These 359 patients and 413 parents came from 420 families, of which respectively 46.2% and 53.8% had a child diagnosed as having leukemia/lymphoma or solid tumors (e.g., central nerve system tumor, peripheral nerve system tumor, kidney tumor, retinoblastoma, bone tumor, soft issue sarcoma, germ cell tumors, and liver tumor). A separate sample of 59 patients (20 from each of the three age groups with one incomplete questionnaire) and 80 parents (20 from each of the four age groups) were recruited to establish test–retest reliability for the PedsQL^TM Cancer Module and the Generic Core Scales. They were assessed twice within a 2-week period by using the main study's questionnaire; their data were not used in the main study. During the study period, prospective participants who were scheduled to visit the outpatient clinic after 2 weeks’ time in the Prince of Wales Hospital were invited to join the test–retest study. The recruitment stopped after the quota of 60 patients and 80 parents was met.

Measures

Patients’ and parents’ background characteristics

Patients’ socio-demographic characteristics (age and gender) and parental information (gender and education level) were recorded. Among the patients from the Prince of Wales Hospital, further information about whether he/she was currently on-treatment or off-treatment was recorded. On-treatment status was defined as currently receiving cancer-related treatments (i.e., chemotherapy, radiotherapy, or surgery); off-treatment status was defined as having completed the entire course of treatment and/or being currently not prescribed any cancer-related treatments.

Cancer-Specific HRQOL

The PedsQL^TM Cancer Module instrument includes 27 items which form eight subscales (Pain and Hurt, Nausea, Procedural Anxiety, Treatment Anxiety, Worry, Cognitive Problems, Perceived Physical Appearance, and Communication). There are parallel and identical versions for pediatric patients (young child: 5–7 years old, child: 8–12 years old and adolescent: 13–18 years old) and parents (toddler: 2–4 years old plus the above three age groups). The participants rated how frequently a particular problem occurred in the past month, using a three-point Likert scale (0 = Never, 2 = Sometimes, 4 = Often) for children 5–7 years old and a five-point Likert scale (0 = Never, 4 = Almost all the time) for children of 8–18 years old and for the parents of patients of all ages. Scores range from 0 to 100 with higher scores indicating better HRQOL. The original study reported that the Cronbach's α values for the subscales of the patient and parent versions have been found to range from .49 to .82 and from .81 to .93, respectively (Varni et al., 2002). Clinical validity has been established by using the Nausea, Treatment Anxiety, and Worry subscales of the pediatric patient version and the Pain and Hurt, Nausea, Procedural Anxiety, and Treatment Anxiety subscales of the parent version to discriminate between patients who were on-treatment and off-treatment. Correlations between the subscales of the Cancer Module and those of the Generic Core Scales have demonstrated concurrent validity (Varni et al., 2002).

General HRQOL

The PedsQL^TM Generic Core Scales assesses HRQOL in healthy and patient pediatric populations (Varni et al., 2001). Its 23-items yield four subscales: the Physical, Emotional, Social, and School Functioning subscales. The questionnaire format, instructions of administration, response categories and scoring methods are identical to those of the Cancer Module. In US pediatric population, high Cronbach's α values (>.80) have been reported for these subscales, which were associated with morbidity and illness burden (e.g., care needed, absence from school, impact on the parent's concentration on work) (Varni et al., 2001). The PedsQL™ Generic Core Scales have also been able to discriminate among healthy children and different types of pediatric patients (Varni et al., 2001). In the present study, the Cronbach's α coefficients for the subscales of the PedsQL^TM Generic Core Scales ranged from .59 to .83 and from .70 to .88 for the patient and parent versions, respectively. The test–retest reliability coefficients ranged from .72 to .92 and .78 to .88 respectively for the patient and parent versions.

Procedures

Linguistic validation was conducted for both the PedsQL^TM 3.0 Cancer Module and the 4.0 Generic Core Scales, following the original developers’ guideline very closely (Varni, 2002). The original English scales were translated into Chinese by two bilingual medical experts, to produce a version which is conceptually equivalent to the original English version. Back-translation was undertaken by two other independent bilingual medical experts. A pretest was conducted among five children and their parents for each of the four age groups (2–4, 5–7, 8–12, and 13–18 years), to ensure that the questionnaires were understood accurately by Chinese respondents. The final version was proof-read by a pediatric oncologist. The entire process was monitored by James W. Varni.

The eligible participants were recruited while they sought medical consultation from the oncology outpatient clinics. An experienced research nurse explained the study to prospective participants and invited them to participate. Written informed consent from the parents and verbal agreement from the pediatric patients were obtained. The questionnaires were self-administered by one of the patient's parents (75.6% were mothers) and by the patients who were 13–18 years old. Patients aged 5–12 years old were face-to-face interviewed by the research nurse. Patients and parents completed the questionnaires (both the Cancer Module and the Generic Core Scales) separately in two private rooms. The Cancer Module and the Generic Core Scales were retested within 2 weeks. Data confidentiality was strictly guaranteed. Ethics approval was obtained from the Chinese University of Hong Kong. Data were collected during March 2004 through March 2006. Information for non-respondents was not recorded; therefore, the response rate is unknown.

Data Analyses

Following the developer's guideline, subscale scores of the Cancer Module were not calculated for those participants who had missing values for over 50% of the subscale's items. The percentage of pediatric patients and parents with a missing subscale score ranged from 0% to 1% and 0% to 2% respectively. The same method was used to define missing subscale scores for the Generic Core Scales.

Construct validity was established by exploratory factor analysis (EFA), using principal component analysis for extraction and the oblique (Promax) method for rotations. The analyses were performed separately for the patient and parent versions and the number of factors was determined by using the criteria of eigenvalue ≥1 and the scree-plot method.

The mean and standard deviation (SD) of the subscale scores and the total scores were computed. Skewness of the score distributions was assessed by using Skewness Statistics (Snedecor & Cochran, 1989) and by reporting the percentage of respondents reporting the extreme scores, i.e., the floor effect (percentage of the respondents reporting the minimum item score) and the ceiling effect (percentage of the respondents reporting the maximum item score) (Andresen, Rothenberg, Panzer, Katz, & McDermott, 1998). Internal consistency was assessed using Cronbach's α coefficients (≥.70 being acceptable); test–retest reliability and consistency between the parent and patient versions were assessed with intra-class correlation (ICC) (small ≤.40; medium .41–.59; large ≥.60; Bartko, 1966). Subscales-total correlation coefficients were derived. To assess concurrent validity, Pearson's correlation coefficients between the total/subscale scores of the Cancer Module and the Generic Core Scales were calculated (small effects .10–.29; medium effects .30–.49; large effects ≥.50; Varni et al., 2002). To assess clinical validity, independent samples t-tests were used to compare the total and subscale scores between participants who were on-treatment (141 patients and 168 parents) and off-treatment (78 patients and 90 parents). SPSS 14.0 for Windows was used for data analyses and p ≤ .05 was considered as statistically significant.

Results

Characteristics of the Participants

The average age of the patients was 10.3 years (SD = 4.5 years) and 59% of the patients were male. Among the father and mother respondents, 41.0% and 41.9% did not attend senior high schools, 32.9% and 41.9% attended senior high schools, and 22.1% and 14.0% attended universities/colleges, respectively (4.0% and 2.2% with missing values).

Construct Validity

The EFA identified seven factors for the pediatric patient version of the PedsQL™ Cancer Module (Table I). Six of these factors (Pain and Hurt, Nausea, Procedural Anxiety, Treatment Anxiety, Cognitive Problems, and Communication) were identical to the theoretical dimensions proposed in the original (US English) scale (Varni et al., 2002). The items of two original factors (Worry and Perceived Physical Appearance) however, loaded on a single factor (named Worry about Disease and Appearance). The variance accounted for by this seven-factor structure was 68.3%. The correlation coefficients ranged from .19 to .44 (p < .001) among the seven subscale scores, and from .54 to .80 (p < .001) among the subscale scores and total scores.

The parent version's factor structure replicated that of the original eight-factor structure, accounting for 78.1% of the variance (Table II). These factors, named after those of the original study, were Pain and Hurt, Nausea, Procedural Anxiety, Treatment Anxiety, Worry, Cognitive Problems, Perceived Physical Appearance, and Communication. All the factor loadings were above .40. The correlation coefficients ranged from .14 to .51 (p < .001) among the eight subscale scores, and from .47 to .68 (p < .001) among the subscale scores and total scores.

Descriptive Statistics

The means and SD of the subscale scores and total scores are presented in Table III. There were no floor effects for total scores of the patient and parent versions, and the ceiling effects for the total scores of the two versions were respectively only 4.2% and 2.7%. The frequency of floor effect for the subscales of the patient and parent versions ranged from 0 to 4.5%, whereas the frequency of ceiling effect (Table III) was quite high for the subscales of the patient version (12.9–77.7%) and for the subscales of the parent version (8.3–69.7%). The Skewness Statistics were −1.2 and −0.8 for the patient and parent version total scores, respectively, and ranged from −3.6 to −0.7 and −2.3 to −0.4 for the subscale scores. There was no significant gender difference on the subscale/total scores for either of the two versions.

Internal Consistency and Test–Retest Reliability

For the entire sample (pooling all age groups), the Cronbach's α values of the total scale were .89 and .92 for the patient and parent versions, respectively. All subscales of the two versions had Cronbach's α ≥ .70 (standard). The Cronbach's α values were <.70 in two of the subscales in the 5–7-year age group for the patient version, and in two and one of the subscales in the 2–4-year and the 5–7-year age groups, respectively, for the parent version (Table IV).

For the entire sample, the test–retest ICC values were .79 and .71 for the total scale of the patient and parent versions (Table IV). All but two of the subscales of the patient version and all but one of the subscales of the parent version had ICC values exceeding the .60 standard. The ICC coefficients were lower than .60 in four, three, and one of the subscales in the 5–7-year, the 8–12-year, and the 13–18-year age group for the patient version, respectively. These coefficients were lower than 0.60 in three, one, four and three of the subscales in the 2–4-year, the 5–7-year, the 8–12-year, and the 13–18-year age group for the parent version (Table IV).

Consistency between the Patient and Parent Versions

Table V summarizes the results of a comparison of patient and parent responses, using the original eight-factor structure. For the entire sample (5–18 years old only, no data from 2–4 years old patients), ICC values ranged from .33 (Communication) to .75 (Nausea). The ICC values ranged from .20 (Worry) to .80 (Nausea) for the 5–7-year age group, from .38 (Perceived Physical Appearance) to .77 (Nausea) for the 8–12-year age group, and from .19 (Communication) to .69 (Nausea) for the 13–18-year age group. In general, the ICC values were strongest for the 8–12-year age group, with the 13–18-year age group yielding the lowest ICC among the three age groups for five out of the eight subscales. The ICC value of the total scale for the entire sample was .49 (5–7 years: .65; 8–12 years: .54; 13–18 years: .32).

Correlations between the Cancer Module and the Generic Core Scales

The correlation coefficients between the total score of the Cancer Module and the Generic Core Scales were .72 and .64 for patient and parent versions, respectively (Table VI). All reported correlation coefficients between the subscales of these two instruments were positive and all but one of these coefficients (i.e., between the Nausea Subscale and the Schooling Functioning Subscale of the patient version) was statistically significant. Most of the correlation coefficients among subscales were of medium or large effect size (Table VI).

Clinical Validity

Patients who were off-treatment, as compared to those who were on-treatment, reported significantly higher scores on the Communication Subscale (82.05 vs. 75.18, p = .05) of the patient version. Parents whose children were off-treatment reported higher scores on the Pain and Hurt (90.09 vs. 82.95, p < .01), Nausea (90.56 vs. 85.40, p < .05), Cognitive Problems (72.00 vs. 66.61, p = .05), Communication subscales (77.13 vs. 67.02, p = .01) and the total score (82.01 vs. 77.35, p = .01) than those whose children were on-treatment (Table VII).

Discussion

The patient and parent versions of the Chinese PedsQL^TM Cancer Module were validated for the first time, using a relatively large sample size. Even though the Cancer Module has been translated into different languages and some psychometric properties of these translated versions have been reported (Felder-Puig et al., 2004; Scarpelli et al., 2008; Varni et al., 2002), this is the first study analyzing the factor structure of this scale. The eight-factor structure of the parent version replicated those of the original theoretical dimensions (Varni et al., 2002). With regard to the patient version, those items of the Worry (three items) and the Perceived Physical Appearance (three items) subscales of the original theoretical dimension loaded on the same factor, reducing the number of factors to seven. Parents may be more concerned about disease/treatment outcomes than children's appearance whereas children/adolescents may be equally concerned about their appearance and disease/treatment outcomes. Such construct validity needs to be investigated for other translations of the Cancer Module.

The Chinese version of the Cancer Module was internally consistent. Only two subscales of the patient version (Cognitive Problems and Worry about Disease and Appearance) for the age 5–7-year group had Cronbach's alpha values below the .70 standard (.43 and .51). Similar to these findings, the Cognitive and Perceived Physical Appearance subscales of the English version have shown poor internal consistency, with values equal to .51 and .38, respectively (Varni et al., 2002).

The Brazilian study (Scarpelli et al., 2008) is the only validation study that has reported test–retest reliability coefficients; however, their analysis was not performed for separate age groups. The present study is hence the first to report reliability coefficients for all age groups. Most of our ICC values for 2-week test–retest reliability for the total scale and relevant subscales demonstrated medium to large effect size. However, test–retest reliability coefficients for the patient version for young children (5–7 years) were relatively low (ICC<.60 for the total scale and four subscales). This finding suggests that the patient version of the Cancer Module should be applied with caution for this age group and supports Varni's recommendation that the patient version for children 5–7 years should only be used for descriptive and exploratory purposes.

The ceiling and flooring effects indicate whether the distributions of the responses are skewed to the left or to the right. This study documented a small floor effect but a relatively large ceiling effect (e.g., the Pain and Hurt, Nausea, and Treatment Anxiety subscales). These results suggest the need for caution when applying this instrument to pediatric patients with very mild illness severity. Some cultural differences might be involved: Asians are more tolerant of sufferings and seldom complain about pain (Im, Liu, Kim, & Chee, 2008) and Japanese scoliosis patients (10–21 years old) reported higher HRQOL as compared to US patients (Watanabe et al., 2007).

The Brazilian validation study of the Cancer Module is also the only other study that computed ICC coefficients to assess the consistency between the patient and parent versions (Scarpelli et al., 2008). Our study further strengthens evidence of the validity of the Cancer Module, showing ICC coefficients of medium to large effect sizes, comparable to those reported by Scarpelli et al. (2008). Strongest agreement between patients and parents was observed for the Nausea Subscale while other subscales such as the Worry (using the original eight-factor structure) and the Communication subscales showed lower ICC values (<.40). This finding may reflect that nausea is an external and often observable problem while worry and communication problems are internal and less observable. A systematic review of the other pediatric HRQOL studies pointed out that agreement between patient and parent versions tend to be stronger for external behaviors (e.g., walking, hyperactivity) and weaker for internal problems (e.g., worry, depression, pain) (Eiser & Morse, 2001). This hypothesis should be further tested by comparing patient and parent Cancer Module data obtained in other countries to determine cross-cultural similarities. It is also interesting to note that adolescents tend to have weaker response agreement with their parents. Results of a meta-analysis study also reported that in general, correlations between child and parent reports were lower in the adolescent group (12–19 years) than in the younger children group (6–11 years) (Achenbach, McConaughy, & Howell, 1987). Such observations may be interpreted from a developmental perspective. Parental influence on formation of perceptions may be weaker for adolescents than for younger children, which is understandable as tensions (e.g., emotional disruption, negative interactions) are common between adolescents and their parents; adolescents were more likely than children to pursue greater autonomy and self-regulation (Collins, Laursen, Mortensen, Luebker, & Ferreira, 1997). Communication between adolescents and their parents may hence be weakened. The age effect on the agreement between the patient and parent versions needs further exploration in future studies.

It is widely acknowledged that information obtained from parents is important in assessing pediatric HRQOL (Cremeens, Eiser, & Blades, 2006). Parental information is especially important when the patients are unable (too young, cognitively impaired, or ill) or unwilling to complete the measurement (Varni et al., 2007b), or when the patient version shows poor reliability (e.g., patient version for the 5–7-year age group in the present study). It is recommended that both the patient and parent versions of the Cancer Module should be used since they collect similar but not identical information (Varni et al., 2007b).

In general, this study establishes good concurrent validity for the Chinese versions of the PedsQL™ Cancer Module and Generic Core Scales. While the relevant correlation coefficients obtained from this study were lower than or similar to those reported in the United States (Varni et al., 2002) they were higher than those found in the Brazilian study (Scarpelli et al., 2008). In this study, the correlation between the Communication Subscale of the Cancer Module and the Schooling Functioning Subscale of the Generic Core Scales was not statistically significant, but these subscales measure two rather different constructs. The correlations between the Nausea and Cognitive Problems subscales of the Cancer Module and the subscales of the Generic Core Scales were relatively low (<0.40). However, this may reflect clinical problems that are specific to cancer conditions and not included in the Generic Core Scales. Because the Cancer Module and Generic Core Scales assess different aspects of HRQOL, it has been recommended that they should be used together in a complementary fashion (Varni et al., 2007b).

To demonstrate clinical validity, some but not all subscales of the Cancer Module were able to differentiate between patients who were on or off treatment (the Communication subscale of the patient version and the total scale and four subscales of the parent version). Similar to our study, some but not all subscales of results reported in the US (Varni et al., 2002) and the Brazilian studies (Scarpelli et al., 2008) were able to demonstrate clinical validity by this criteria (English version: three subscales of the patient version and five subscales of the parent version; Brazilian version: two subscales of the patient version and two subscales of the parent version). In our study, the parent version was more likely than the patient version to have good clinical validity in terms of discriminating on- and off-treatment patient groups. Parents might be more aware of the clinical significance of treatments than pediatric patients and therefore more sensitive than their children to the differences between being on or off treatment. Information about treatments might also be filtered by parents (Eiser & Morse, 2001), which is likely to happen in the Chinese populations, reducing doctors ability to communicate treatment-related information directly to pediatric patients (e.g., reasons for treatment, risk, side-effects, and response rates). Pediatric patients with cancer would then find it more difficult to communicate with medical professionals while they are on treatment which may result in lower scores on the Communication subscales than those not currently on treatment. These hypotheses can be tested in future studies to better inform medical professionals about how to communicate with pediatric cancer patients.

Overall, our findings are consistent with those reported in other countries. The moderate to high mean scores obtained from our sample were comparable to similar figures reported in the United States (Varni et al., 2002). The percentage of missing values was lower than 1% among all individual items of the patient version and lower than 3% among all individual items of the parent version, values roughly comparable to those found among American pediatric patients with cancer (Varni et al., 2002), indicating good data quality.

However, this study has several methodological limitations. First, information on the non-responders was not available; the response rate was unknown and relevant biases could not be assessed. Second, the relatively high ceiling effects imply that we should be cautious about applying this tool to the pediatric populations of less severe cancer stages. Third, concurrent validity between the Cancer Module and other cancer-specific HRQOL tools such as the (Chinese) Quality of Life for Children with Cancer was not assessed, due to the limited length of time for the interview. Fourth, the type of treatment was not recorded so that differences among treatment modalities could not be assessed. Fifth, non-random samples were drawn from three out of the five oncology centers in the region. Generalization should be made with caution although it is noted that the other validation studies were also based on non-random samples and had much smaller sample size, suggesting that the sampling method of this study is acceptable. Last, the study was conducted among Hong Kong Chinese. Further validation among mainland Chinese and oversea Chinese are required before the instrument is to be used in other Chinese populations.

In summary, the Chinese version of the PedsQL^TM Cancer Module shows acceptable psychometric properties, and many of the results are quite comparable to those of the original version and translated versions validated in other countries. This study therefore extends and supports cross-cultural applications of the PedsQL^TM Cancer Module, which can be used in pediatric cancer patients of a wide age range (2–18 years). This study also provides new information about the factor structure, test–retest reliability, and concordance between patient–parent versions, providing suggestions and hypotheses for future research on similar validation studies. Overall, our results support the reliability and validity of both the patient and parent versions of the Chinese PedsQL^TM Cancer Module as an instrument to be used for assessing HRQOL among Chinese pediatric patients with cancer.

Funding

This study was partly supported by the Hong Kong Children's; Cancer Foundation Peter Nash Paediatric Oncology Research Grant.

Conflicts of interest: None declared.

References