Participants
Data used in this study were from 6544 child participants of the Longitudinal Study of Australian Children (LSAC)(24). The LSAC is an ongoing, large population representative survey of children and their families, that collects data on child development and wellbeing. The LSAC recruited two cohorts of children in 2004 using clustered sampling methods: 5107 children in the Birth (B) cohort and 4983 children in the Kindergarten (K) cohort (25). Children and their caregivers were interviewed every two years, with the most recent wave of data collection in 2020. In the present study, we used data from both the B and K cohorts in which both PedsQL and CHU9D were included. This encompassed waves 6,7 and 8 of the B cohort, in which children were aged 10/11, 12/13 and 14/15 years, and waves 6 and 7 of the K cohort in which children were aged 14/15 and 16/17 years.
Demographic Characteristics
Age (in years), sex (male or female), socioeconomic position (SEP) (High or Low), culturally and linguistically diverse (CALD) status (CALD/not CALD) and Indigenous status (Aboriginal or Torres Strait Islander) were included as controls in the analyses. Individual-level socioeconomic position was measured at each wave using a variable developed by the LSAC study investigators which combined the education level, occupation type and income of the child’s caregivers into a z-score (31). For simplicity, we categorised this variable into high (SEP z-score > = 0) and low SEP (SEP z-score < 0). A language other than English regularly spoken to the child, collected at age 2–3 years for the B cohort and age 4–5 years for the K cohort, was used as a proxy for CALD status.
Psychometric properties /statistical analyses
The analyses of psychometric properties were conducted in accordance with practice guidelines and criteria for psychometric assessment (22, 32, 33). For all analyses, except those assessing acceptability through missing data, we used observations that were complete for BMI, PedsQL, and CHU9D.
Reliability The only aspect of reliability we were able to assess with our existing dataset was internal consistency reliability, which is the degree of interrelatedness among items from the same scale (32). Cronbach’s alpha and item-total correlations were used to assess the interrelation of the relevant individual items of PedsQL with the four summary scores and with the total score, and for the individual items of CHU9D with the total utility score scale, among children with overweight and obesity. A Cronbach’s alpha value ≥ 0.7 and item-total correlations ≥ 0.2 are considered acceptable thresholds for internal reliability consistency (33) (34) .
Acceptability measures the quality of the data and is assessed by the completeness of the data and score distributions, including floor and ceiling effects. Acceptability may also include the practicality and feasibility of using a particular instrument among children with overweight and obesity, and may include measures of comprehension or burden of completion. Without access to respondents, we investigated acceptability through the assessment of missing data and the proportion of ceiling and floor values for the PedsQL total scores and CHU9D utility scores (35) across age and weight status. A low and acceptable level of missing data < 5% was used as a benchmark (36), and the threshold for the acceptable floor and ceiling values was < 10% (35).
Validity was addressed through known groups validity and convergent validity. Known groups validity is the extent to which a HRQoL measure can distinguish groups of children with and without a health condition, or between children with different severity of a condition. We hypothesised that children with higher weight status would have lower HRQoL and investigated known groups validity using general estimating equations (GEE) to account for the repeated measures of weight status and HRQoL among the same children, with adjustment for socio-demographic characteristics known to impact on HRQoL (34). The GEE models included binomial family, log-link function and robust variance estimation. PedsQL Total Scale Scores (transformed to 0 to 1 scale) and CHU9D utility scores were the response variables; explanatory variables were weight status (healthy, overweight, obesity) and demographic variables described above. Interaction terms of weight status and significant demographic variables were included to identify whether these parameters modified the association of HRQol and weight status. Models were fitted separately for girls and boys and significance levels were set at p < 0.05 for main effects and p < 0.01 for interaction terms. The margins command in STATA was used to predict marginal effects of weight status on reduced HRQoL, and to predict HRQoL by age and weight status, using final models including interaction terms, where significant.
Convergent validity measures the level of agreement between instruments that purport to measure the same construct, and usually uses an existing health measure as a comparator. As PedsQL and CHU9D are well established and accepted measures of the same general construct i.e. HRQoL, we assessed convergent validity by calculating Spearman’s correlations between the CHU9D utility scores and the PedsQL Total Scale Score among children in each weight status group. Correlation coefficients > 0.8 are regarded as strong, between 0.61 to 0.8 as good, between 0.41 to 0.6 as moderate and < 0.4 as weak convergent validity (23). We hypothesised that there would be moderate correlation between the two instruments, as they are both measures of HRQoL, but one is child report and the other is parent proxy.
Responsiveness is the ability of a measure to detect change over time when there are known changes in health status. (37). This was examined by whether changes in the PedsQL total score and CHU9D utility score were responsive to changes in weight status between subsequent waves in the LSAC. Children were classified as to whether their weight status stayed the same, improved or deteriorated between consecutive waves of LSAC, according to the three weight status groups: healthy, overweight or obese. Both the B and K cohorts were used in the analysis, providing data on the change in HRQoL scores for individual children over two-year intervals from mean ages 11 to 13, 13 to 15 and 15 to 17 years. We hypothesized that deterioration in weight status (healthy to overweight; healthy to obese; overweight to obese) would result in a negative HRQoL score change, whilst improvement in weight status (overweight to healthy; obese to overweight; obese to healthy) would result in a positive change in HRQoL scores, and no change in weight status would result in a HRQoL score change close to zero. Standardised response means (SRM) and effect sizes (ES), which take into account the change in HRQoL score in relation to the SD of baseline score, were calculated according to the method outlined in (38).