Health-related quality of life and its association with socioeconomic status and mental health in 5- to 7-year-old children: a cross-sectional study

The current research is a substudy of the Norwegian Starting Right™ project, which uses online proxy- and child-reported questionnaires to support public health nurses (PHNs) in assessing children’s health and development in routine follow-up [36]. The cross-sectional data were collected from students in Grade 1 in Agder County in southern Norway, who were recruited from both public and private schools. All Grade 1 students accompanied by parents were offered free consultation with the PHN. The mothers of the participating children answered online questionnaires concerning their child’s health prior to the consultation.

Data were collected from 2019 to 2023 using the SDQ for the assessment of mental health [37] and the KIDSCREEN-27 for the assessment of HRQoL [38]. Self-reports are the preferred method for assessing HRQoL because they allow individuals to directly provide information about their own health. However, in younger children, proxy reporting is recommended [8, 9, 39]. Prior to the individual consultation with the PHN at school, the proxy-reported online questionnaires were distributed to all Grade 1 parents as part of a start-of-school follow-up routine. All questionnaires were answered online prior to the consultation. Twentynine questionnaires were distributed before the child started Grade 1, as one of the municipalities altered their routines early in the project. With informed consent, the clinical data were aligned with Norwegian statistics and made available for the present study. Using unique personally identifiable IDs, data from the questionnaires were linked with administrative data from Statistics Norway, including maternal household income, educational level, and child sex. The parental role and child age data were obtained from Starting Right™.

To evaluate HRQoL, we applied the proxy version of the KIDSCREEN-27 questionnaire, a multidimensional measure of generic HRQoL consisting of 27 items grouped into five subscales: (1) physical well-being; (2) psychological well-being; (3) autonomy and parent relations; (4) social support and peers; and (5) school environment [40]. Previous research has provided support for its psychometric performance [41, 42]. The proxy version includes items similar to those in the child version, but it also includes questions about parents’ thoughts about their child’s feelings [40]. Item are assessed using a 5-point Likert scale, measuring either the strength of an attitude or the frequency of a behavior or emotion [40]. In line with the KIDSCREEN handbook [40], the Rasch scores were computed and transformed into T-scores with a general population normative mean of 50 and a standard deviation (SD) of 10. Higher scores indicate better HRQoL. Negatively worded items were reversed according to the manual. The normative values were based on data from an international survey sample including children aged 8–11 years old from 12 European countries [40].

We collected two common indicators for SES in the current study, which were maternal household income and maternal education. Our data included household income from mothers’ residences, also including fathers’ income when they are living together. Household income after tax per consumption unit was calculated using the European Union equivalence scale, as measured in 2020 and 2021. Our data did not include the exact household income values but included the mothers’ relative ranks in the household income distribution relative to the population of Norway. Statistics Norway provides income percentile data for individuals > 16 years living in Norway, here as categorized by sex. Our data show maternal household income percentiles. The maternal household income variable was divided into five even categories of percentiles: 1 = 0th–19th percentile, 2 = 20th–39th percentile, 3 = 40th–59th percentile, 4 = 60th–79th percentile and 5 = 80th–100th percentile. Maternal education was measured as the highest level of education completed by 2021. Low educational levels included elementary school, high school, and education based on senior high school but were not considered university or college education. A high educational level included any degree from a college or a university.

The Strength and Difficulties Questionnaire was used to assess children’s mental health problems, as reported by their mothers. The SDQ has been translated into Norwegian and has been used in general population studies of Norwegian children, demonstrating appropriate psychometric properties [43]. We used the 20 SDQ difficulty items representing four problem scales, namely (1) emotional problems, (2) peer problems, (3) conduct problems, and (4) hyperactivity-inattention, forming a Total problem score [44]. The SDQ subscale for prosocial behavior was not used, as we investigated mental health problems. Higher scores indicate more severe problems. The response scale is “not true”, “somewhat true”, or “certainly true”, rated 0–2 for negatively worded items and rated inversely 2–0 for positively worded items. Scores were calculated for each subscale (range 0–10) and for the Total problem score (0–40). Norwegian norms and cutoffs for the SDQ have not been established [43]; therefore, we calculated the 80th and 90th percentiles for the cutoff values within our study population for the SDQ Total problem score, similar to how developers calculate cutoffs in normative populations [45].

Descriptive analyses were conducted using IBM SPSS Statistics (version 28). STATA (Stata-Corp. 2019, Stat Statistical Software: Release 17. College Station, TX, USA) was used to fit the robust regression models. Descriptive statistics were calculated for all variables and are presented as the mean and SD for continuous variables and counts and percentages for categorical variables. Because the assumptions for linear regression were not met and the residuals were skewed, we used robust regression to model possible associations between the dependent variable (the KIDSCREEN-27 subscales) and the selected independent variables. Univariate robust regression was conducted to examine possible associations between SES and HRQoL (KIDSCREEN-27 subscales: physical well-being, psychological well-being, autonomy and parent relations, social support and peers, and school environment), and between SDQ and HRQoL. Second, multiple robust regression was performed separately for SES (maternal household income and education) and the SDQ Total problem score in relation to HRQoL subscales while controlling for age and sex. Robust regressions were fitted separately for each of the five KIDSCREEN-27 subscales. Additionally, multiple robust regression was conducted for each KIDSCREEN-27 subscale, with the four SDQ scale scores as the independent variables, again controlling for age and sex. The results are presented as regression coefficients with 95% confidence intervals (CIs). P values ≤ 0.05 were considered to indicate statistical significance. All analyses were considered exploratory, so no correction for multiple testing was performed.

Sensitivity analyses were performed by excluding participants with maternal household income in the 0th–19th percentile because nontaxable income sometimes does not indicate low income in general. Sensitivity analyses were also performed excluding children who had not yet started school because some children were included in the project prior to the start of school.

In total, mothers of 621 children participated in the study, with a response rate of 78%. Just over half (51.4%) of the included children were girls. All participating children were in Grade 1, with an age range of 5.1–7.3 years. Our study sample represented all groups of SES, however, it had a small overrepresentation of mothers with university or college education (65.4%) compared with Norwegian women aged 30–39 years old (54.8% nationally, 52.7% in Agder County, where the study was conducted), including women without children. One-third (32.3%) of the mothers had completed high school as their highest level of education, while 7.7% had completed elementary school only. Almost 30% of mothers had a household income between the 40th and 59th percentiles, while 12.6% of mothers had a household income above the 80th percentile of the national household income percentiles (Table 1).

The mother-reported mean scores for all five KIDSCREEN-27 dimensions were above or well above the 50th percentile for the European KIDSCREEN-27 proxy report norm for children aged 8–11 years [40]. The distribution of each KIDSCREEN-27 subscale score is shown in Fig. 1. The mother reported mean KIDSCREEN-27 and SDQ scores are shown in Table 2. The internal consistency of the KIDSCREEN-27 questionnaire was confirmed, with a Cronbach’s alpha ranging from 0.70 to 0.83 for all subscales for the current sample. Values > 0.7 indicate acceptable reliability [46]. The Cronbach’s alpha [46] for the SDQ Total problem score in our sample was 0.78. The Cronbach’s alpha for the four problem scales ranged from 0.53–0.80, indicating an acceptable consistency for hyperactivity/inattention but low degree of internal consistency for emotional, peer, and conduct problems [46].

When controlling for sex and age, having a mother with an income in the 80th–100th percentile was associated with 3.4 points greater scores for physical well-being (B = 3.4, 95% CI [0.1; 6.8]) and 3.8 points greater scores for the school environment subscale of KIDSCREEN-27 (B = 3.8; 95% CI [0.4; 7.2]) than having a mother with an income in the 0–19th percentile (Table 3). Having a mother with an income in the 60th–79th percentile was associated with a greater autonomy and parent relations score (B = 2.2, 95% CI [0.1; 4.3]) than having a mother with an income in the 0th–19th percentile. Being a boy was associated with lower subscale scores for social support and peers (− 1.7, 95% CI [− 3.1; − 0.4]) and school environment (− 2.5, 95% CI [− 4.1, − 0.9]). Furthermore, having a mother with a high educational level was associated with a higher subscale score for school environment (B = 2.3, 95% CI [0.5; 4.1]). In univariate analyses SES was positively associated with HRQoL (Table S1, supplementary materials).

In multiple robust regression analyses, the possible association between the SDQ Total problem score and HRQoL was investigated after adjusting for sex and age. As shown in Table 4, results indicated a significant association between mental health problems and lower HRQoL. Children scoring above the 90th percentile for the SDQ Total problem score had lower scores for KIDSCREEN-27 subscales physical well-being (B = − 4.3, 95% CI [− 6.8; − 1.8]), social support and peers (B = − 4.5, 95% CI [− 6.7; − 2.2]), psychological well-being (B = − 7.0, 95% CI [− 8.9; − 5.1]), and school environment (B = − 10.7, 95% CI [− 13.2; − 8.2]). Children classified between the 80th and 90th percentiles for the SDQ Total problem score showed 4.4 points lower on the KIDSCREEN-27 school environment subscale (B = − 4.4, 95% CI [− 6.8; − 2.0]) than children scoring below the 80th percentile for the SDQ. Finally, being a boy (B = − 1.9, 95% CI [− 3.4; − 0.4]) was associated with a lower HRQoL score in terms of school environment than being a girl.

Multiple robust regression analyses, including the four SDQ subscales and controlling for sex and age, are shown in Table S3 in the supplementary materials. These analyses showed that having peer problems was associated with a significantly lower HRQoL for all KIDSCREEN-27 subscales. Furthermore, emotional symptoms, peer problems, conduct problems, and hyperactivity/inattention were all associated with lower KIDSCREEN-27 scores in school environment. Univariate analyses showed statistically significant associations between mental health problems and HRQoL (Table S2, supplementary materials).

To check whether there were significant differences in HRQoL measures pre and post the COVID-19 pandemic, we included a binary covariate pre versus post pandemic and reran all the multiple analyses with this new variable, as shown in supplementary materials Tables S4, S5 and S6. The COVID-19-variable was not statistically significant in any of the models.

When we removed mothers with the lowest income, our results remained unchanged (data not shown). When children who had not started school were removed, our results also confirmed the main analyses (data not shown).

The major strengths of the present study are the relatively large sample size, the fact that participants were recruited from a variety of schools and that a high consent rate was achieved. Regarding the limitations of our study, we did not have information about the group that did not consent to participate. Our study included mothers’ responses only, which may differ from fathers’ reporting. Our study sample represented all groups of socioeconomic status; however, it had a small overrepresentation of mothers with higher education. One study recommended that the Norwegian proxy version of the KIDSCREEN-27 should be used with caution in 6-year-olds, particularly regarding psychological well-being and autonomy and parent relations [42]. Another limitation was the low Cronbach’s alpha for three of the SDQ subscales, which may have affected our findings. Furthermore, the present study included cross-sectional data only; hence, results should not be interpreted as causal. We did not intend to investigate the impact of COVID-19 on children’s HRQoL or mental health. However, restrictions during the pandemic have affected children’s health and well-being, especially in the least privileged families [61].Our SES variables are based on objective administrative data, which were the available variables in the project. However, use of subjectively reported scales could have added nuances and strengthened the study [62]. Regarding the association between SES and HRQoL, when controlling for possible confounders, maternal household income had a large CI [0.1; 6.8] [0.4; 7.2] (Table 3), which may indicate a large degree of heterogeneity in the observations obtained. The large CI suggests that the association between SES and HRQoL may vary in our sample. If we had a larger sample size, we could model interactions and possibly identify subgroups. The associations found between SES, mental health problems, and HRQoL may be influenced by other confounders that were not controlled for in the present study. We have only adjusted our multiple models for age and sex. Thus, we cannot rule out that the effect of both SES and SDQ on the outcome could be modified by such confounders. Further and more detailed analyses are just warranted.