In this study, reference data of the Dutch PedsQL Generic Core Scale were collected and psychometric properties were assessed in a sample of 496 children aged 5 to 18 years and their parents. The psychometric properties of the PedsQL in the Netherlands range from sufficient to good.
The PedsQL reference scores obtained in this study are an adequate representation of the general Dutch population. According to Statistics Netherlands [15
] in 2008 80.4% (of the total population) were from Dutch origin and 77.9% (of population aged 45-55 years) were employed. These figures are comparable to the socio-demographic data from our study. Furthermore, it is not uncommon [16
] to find a large percentage of highly educated parents (48.4%) in the current study, compared to the Dutch working population (33.5%). It is likely that highly educated parents are better aware of the necessity of this type of research, and thus be more willing to participate. During data collection we experienced that willingness to take part in the study seemed lower on schools with lower educational levels and higher percentages of migrant children. Possibly, parents not born in the Netherlands experienced language problems and for this reason did not fully understand the information letter to participate in the study. These findings are supported by the fact that there were fewer parents with a high education and Dutch ethnicity in the non-participants group, compared to the group that did participate.
The Dutch sample shows the same trend in reliability across subscales as the US [12
], however with slightly lower alphas. In both samples, the total score appears to be most reliable and subscale school functioning least. A possible clarification for the lower Dutch reliability figures could be the difference in sample size. The US sample ranges from 1159 to 2674 children per age group; the Dutch from 92 to 219. More over, the US sample included children from the State Children's Health Insurance Program (SCHIP) that provides health insurance coverage to uninsured children from low-income families [17
]. In the current study a stratified sampling technique was applied in order to create a sample as diverse as possible, which can be generalized to the overall Dutch population. Bastiaansen et al [13
] have investigated the psychometric properties of the PedsQL in the Netherlands as well. In their study a sample (n = 74) from the general Dutch population (6-18 years) was compared with 310 children referred for psychiatric problems: alpha's were in concordance with Varni et al [12
], however, their sample size was very small.
Regarding the socio-demographic within-group differences, this study demonstrates a gender difference in group 8-12 on the emotional subscale with girls obtaining lower scores than boys. Similar results in adolescent girls have been reported by Reinfjell et al [18
Salient finding is that country of birth of the parent had an effect on PedsQL scores in group 8-12. Children of parents born in the Netherlands scored significantly lower on all subscales except physical functioning. Previous studies only correspond partly with this result. For instance in the US, White together with Asian children scored higher than Hispanic and Black children [12
]. Important to note is that we only collected data regarding country of birth of the parent, and not country of birth of the child. Moreover, this difference was not found in groups 5-7 and 13-18.
Results of our study indicate a relationship between educational level and PedsQL scores in group 13-18: children of parents with a low education perceived a significantly better HRQOL. This phenomenon is difficult to explain, since previous research mainly pointed out that high quality of life scores were related to high parental education, or that education had no effect at all [19
Findings on parental employment are also notable: having a job had no influence on the child's HRQOL in our sample. Previous research has shown that children with low socioeconomic status (SES) functioned worse than children from middle SES backgrounds [21
]. Employment and SES are not exchangeable, though. Having a job does not necessarily implicate a middle or high SES.
With respect to construct validity, this study demonstrates that the Dutch PedsQL version differentiates between children with and without a chronic condition in group 5-7 and 13-18. Differences in group 8-12 were not significant - however, except for the emotional functioning subscale, healthy children obtained higher scores than their chronically ill peers. Although several studies have shown that the PedsQL differentiates between children with and without a chronic health condition [9
] it is not exceptional to find adequate functioning for chronically ill children [22
]. Another explanation could be that severely ill children did not take part in our study, because they were not present at the time of administration due to illness or that parents did not want to burden them with participation. A further possible reason might be the fact that the presence of a chronic health condition in our sample was determined by the parent and not diagnosed by a physician. Physician-diagnosed chronic health conditions are often stricter than those reported by the parent. Therefore, the 8-12 year old chronic health condition sample in our study might not be completely representative of children with more serious chronic diseases.
Limitations of the study need to be taken into account. First, a considerable number of children were approached, which eventually resulted in nearly 500 participants. However, considering the different age groups and the socio-demographic within-group differences, sample sizes per group were relatively small. Furthermore, it is possible that presenting PedsQL items one at a time on a computer - with missing values not being allowed - could have some psychometric implications. Digital administration of the PedsQL has demonstrated equivalent measurement properties to the paper version [23
], yet in the study of Varni et al (2008) each PedsQL scale was depicted on a separate screen and participants had the option to skip items. The fact that this possibility was lacking in the current study could have forced participants to choose an answer that did not really apply to them. Nonetheless, this probably concerns a minimum of items since Varni et al (2008) also demonstrated similar (low) percentages of missing-item responses of digital and paper version PedsQL. Additionally, it would have been interesting to have examined all age versions of the PedsQL involving all regions of the Netherlands, but this was unfeasible for the purpose of our study. Therefore, we recommend that future research with respect to the PedsQL in the Netherlands should include more regions of the country and incorporate the remaining PedsQL versions.