Preteen children’s health related quality of life in Sweden: changes over time and disparities between different sociodemographic groups

Study population

A repeated cross-sectional study was conducted during the school years starting August 2005 and 2009 (onwards named 2005 and 2009). The main objective of this school-based study, the ETICS study (Exploring The Iceberg of Celiac in Sweden), was the screening of celiac disease. The study comprised two cohorts of children in 6th grade at five study sites in Sweden: Lund/Malmö, Växjö, Norrköping, Norrtälje and Umeå [24].

Out of the 18,325 invited children, 13,279 agreed to participate in the study. As part of the study, children answered a questionnaire at school together with the screening while parents answered a questionnaire at home. In total, 12,658 of the children (95%) and 11,239 of the parents (85%) filled in the questionnaire. In this study, we included the 11,009 children (60% of the invited) with responses from both themselves and their parents; 6449 from the first cohort and 4560 from the second cohort. There were 1419 children’s questionnaires without parents’ responses where only information about child-reported HRQoL dimensions are available. Children were asked about their HRQoL, while parents responded to questions about their own and their child’s country of birth, family structure, their employment status, occupation, and education.

Measures

HRQoL in the general population can be measured by a variety of instruments. Of these, the EuroQol 5 Dimensions (EQ-5D) captures the HRQoL of adults and EQ-5D-Y (Youth) is a child-friendly version of the adult instrument. The EQ-5D-Y has been tested in Sweden and international studies present evidence of its feasibility, reliability and validity. These studies support the discriminative validity of the tool i.e. its ability to differentiate between children belonging to different sociodemographic groups [25–27].

Five dimensions of HRQoL were measured by the EQ-5D-Y [28]. These dimensions correspond to mobility in terms of ability to walk; looking after one-self (self-care) in terms of independence in daily personal care, especially regarding washing and dressing; performing usual activities such as going to school, family and spare-time activities, hobbies, sports and playing; having pain or discomfort (pain), and finally, feeling worried, sad or unhappy (mood problems). Response alternatives correspond to no, some or severe current problems. In the analysis, ‘some’ and ‘severe’ problems were combined as one category.

The inclusion of potential explanatory factors of HRQoL disparities was guided and ordered by the PROGRESS framework [9].

Place of residence, the 23 included municipalities were categorized according to population size. As suggested by the Swedish Central Bureau of Statistics [29], large municipalities was defined by areas with more than 50,000 inhabitants, middle size municipalities by areas with 10,000 to 50,000 inhabitants, and small municipalities by areas with less than 10,000 inhabitants.

Migration status was categorized into four groups based on child and parents’ country of birth: Swedish origin, i.e. the child and both parents were born in Sweden. One foreign parent, i.e. the child and one parent were born in Sweden. Foreign parents, i.e. the child was born in Sweden but both parents were born outside Sweden. Foreign origin, i.e. the child and both parents were born outside Sweden. This approach is in accordance with Hjern et al. [21].

Parents’ occupation was measured by the socioeconomic index (SEI) which is based on the Swedish Socioeconomic Classification (SEI-1982) [30]. This index captures the highest occupation and employment status in the household. Household SEI was classified into skilled non-manual workers including executives and self-employed professionals (SEI-codes: 54–57 and 60–87), unskilled non-manual worker including assistants and intermediate non-manual workers (SEI-codes: 33–46), skilled manual worker (SEI-codes: 21–22), unskilled manual worker (SEI-codes: 11–12), and students (SEI-codes 01–03). The SEI of the currently employed parent was used if the other parent was unemployed. The household occupation and employment status was defined as student if both parents were students or one parent was a student and the other unemployed for more than six months.

Parents’ education level was categorized into low, medium and high education levels based on the International Standard Classification of Education (ISCED-1997) [31]. A high education level was defined as holding a college or university degree (ISCED levels 5 and 6), a medium level as 12–13 years of education (ISCED levels 3 and 4), and a low level as nine years of education or less (ISCED levels 0, 1 and 2). Parents’ education level was defined by the highest level of the two parents if the child lived with both parents or in joint custody, and by the education level of the parent that the child lived with when it was only one parent.

Family structure was categorized into four groups as suggested by Bergström et al. [19]. These groups captured whether the child during the past year had lived with: two parents; in joint custody, i.e. alternating between separated or divorced parents; mainly with one parent, i.e. lived with one parent for most of the week; or with one parent and another adult.

Statistical methods

Logistic regression was used to study the association between each of the five HRQoL dimensions and various potential explanatory factors (study year, municipality population size, migration status, parents’ occupation, sex, parents’ education level, and family structure). The analyses were also performed based on fathers’ or mothers’ education separately. However, the results are not presented here as the analysis showed similar findings. Odds ratios (OR) with 95% confidence interval (CI) were reported. Model fit was evaluated with Wald’s test [32], and was tested for possible multicollinearity with the variance inflation factor (VIF). Results were assessed through stratifications by all used factors to verify that no obvious pattern, such as effect modification, existed. Stata 13 was used for the statistical analyses (StataCorp LP, College Station, Texas, USA). A statistically significant level of 0.05 was applied.

Comparisons between 2005 and 2009

Comparison between sociodemographic groups

Overall, there were few observed disparities in HRQoL between sociodemographic subgroups (Table 2). Detailed numbers and prevalence of children who reported HRQoL problems according to different sociodemographic groups are shown in Additional file 1. Municipality population size was associated with two of the studied HRQoL dimensions. Children who lived in small compared to large municipalities had statistically significantly higher odds in two of the dimensions; OR’s were 2.46 in problems performing usual activities and 1.29 in pain. Regarding migration status, the only statistically significant difference was among children with one foreign parent who had an OR of 1.3 in mood problems compared with children of Swedish origin. For parents’ occupation, there were no statistically significant differences between groups (Table 2).

Compared to boys, girls had OR’s of 1.2 in pain and 2.57 in mood problems (Table 2). Only in mood problems there were statistically significant differences based on parents’ education levels. Children of parents with low or medium levels of education had OR’s of 0.84 respectively 0.65 in mood problems compared to children of parents with high education levels.

Family structure was the factor associated with the most widespread statistically significant HRQoL disparities (Table 2). Compared to children living with two parents, children living with one parent and one other adult had an OR of 1.66 in problems performing daily activities, and the OR of pain was 1.3 in children living mainly with only one parent. Finally, compared to children living with two parents, children in all other family constellations studied had OR’s of 1.33 to 1.57 in mood problems.

Comparison between boys and girls

After stratifying by sex there were few disparities in HRQoL in girls but even fewer in boys (Table 3). For comparisons between cohorts, pain was only statistically significant among girls with an OR of 1.25 in the second cohort compared with the first cohort. Both boys and girls had a statistically significant increase in mood problems in the second cohort (OR’s of 1.37 in girls and 1.34 in boys) compared to the first cohort. Living in small municipalities showed statistically significant disparities only among girls. Girls living in small municipalities reported more than three times higher odds of problems performing usual activities and had an OR of 1.51 in pain compared to those who lived in large municipalities. While, no statistically significant disparities were seen among boys in any of the measured HRQoL dimensions. Percentages of reported HRQoL problems in different municipality population sizes among boys and girls are shown in Fig. 1.

Characteristic	Mobility problems		Problems in performing usual activities		Pain		Mood problems
	Boys	Girls	Boys	Girls	Boys	Girls	Boys	Girls
	Adj. ORa	Adj. OR	Adj. OR	Adj. OR	Adj. OR	Adj. OR	Adj. OR	Adj. OR
Study year
2005 (ref)b	1	1	1	1	1	1	1	1
2009	0.92 (0.6–1.3)	1.24 (0.9–1.8)	1.00 (0.7–1.5)	1.03 (0.7–1.5)	1.15 (1.0–1.3)	1.25 (1.1–1.4)	1.37 (1.1–1.7)	1.34 (1.2–1.6)
Municipality size
> 50,000 inhabitants (ref)	1	1	1	1	1	1	1	1
10,000 to 50,000	1.39 (1.0–2.0)	1.09 (0.7–1.6)	0.66 (0.4–1.1)	1.44 (0.9–2.2)	1.02 (0.9–1.2)	0.90 (0.8–1.1)	0.98 (0.8–1.2)	0.97 (0.8–1.2)
< 10,000 inhabitants	1.53 (0.8–3.0)	1.45 (0.7–2.8)	1.59 (0.8–3.1)	3.77 (2.1–6.8)	1.07 (0.8–1.5)	1.51 (1.1–2.0)	0.99 (0.6–1.6)	1.09 (0.8–1.5)
Migration status
Swedish origin (ref)	1	1	1	1	1	1	1	1
One foreign parent	1.20 (0.7–2.1)	1.16 (0.7–2.1)	0.83 (0.4–1.7)	0.78 (0.4–1.6)	0.97 (0.8–1.2)	1.04 (0.8–1.3)	1.32 (1.0–1.8)	1.21 (0.9–1.5)
Two foreign parents	0.46 (0.2–1.3)	1.27 (0.7–2.4)	0.82 (0.4–1.9)	1.01 (0.4–2.4)	0.78 (0.6–1.1)	1.02 (0.8–1.4)	1.19 (0.8–1.8)	0.84 (0.6–1.2)
Foreign origin	1.75 (0.7–4.1)	1.39 (0.5–3.5)	0.65 (0.2–2.8)	1.38 (0.5–3.9)	0.99 (0.6–1.5)	0.59 (0.4–0.9)	1.27 (0.7–2.3)	0.87 (0.6–1.4)
Parents’ occupation
Skilled non-manual (ref)	1	1	1	1	1	1	1	1
Unskilled non-manual	1.09 (0.7–1.6)	0.91 (0.6–1.4)	1.11 (0.7–1.8)	0.67 (0.4–1.1)	1.10 (0.9–1.3)	1.10 (0.9–1.3)	0.82 (0.6–1.1)	1.00 (0.8–1.2)
Skilled manual	0.83 (0.5–1.5)	1.58 (0.9–2.6)	1.39 (0.8–2.6)	0.94 (0.5–1.8)	1.03 (0.8–1.3)	1.07 (0.8–1.4)	1.15 (0.8–1.6)	0.99 (0.8–1.3)
Unskilled manual	1.18 (0.6–2.4)	1.70 (0.9–3.2)	1.35 (0.6–2.9)	1.22 (0.6–2.6)	1.21 (0.9–1.6)	1.17 (0.9–1.6)	0.96 (0.6–1.5)	1.15 (0.8–1.6)
Student	4.29 (1.5–11.)	n/ac	1.52 (0.2–12.)	n/a	2.38 (1.2–4.6)	1.01 (0.5–2.1)	1.32 (0.5–3.5)	1.99 (1.0–3.9)
Parents’ education level
High (ref)	1	1	1	1	1	1	1	1
Medium	1.28 (0.9–1.9)	1.30 (0.9–1.9)	1.30 (0.8–2.0)	0.64 (0.4–1.0)	0.99 (0.8–1.2)	0.94 (0.8–1.1)	0.77 (0.6–1.0)	0.87 (0.7–1.0)
Low	1.23 (0.5–2.9)	0.80 (0.3–2.1)	0.75 (0.2–2.6)	0.74 (0.3–2.0)	0.78 (0.5–1.2)	1.04 (0.7–1.5)	0.57 (0.3–1.1)	0.68 (0.4–1.0)
Family structure
Both parents (ref)	1	1	1	1	1	1	1	1
Joint custody	1.25 (0.7–2.3)	0.97 (0.5–1.9)	0.68 (0.3–1.6)	1.78 (0.9–3.4)	1.10 (0.8–1.4)	1.19 (0.9–1.5)	1.27 (0.9–1.8)	1.42 (1.1–1.9)
Mainly one parent	0.99 (0.5–1.8)	1.18 (0.7–2.0)	0.83 (0.4–1.7)	1.62 (0.9–3.0)	1.23 (1.0–1.6)	1.40 (1.1–1.7)	1.47 (1.0–2.0)	1.63 (1.3–2.1)
Parent and another adult	0.88 (0.4–1.8)	1.61 (0.9–3.0)	1.17 (0.6–2.5)	2.40 (1.2–4.6)	1.20 (0.9–1.6)	1.19 (0.9–1.6)	1.40 (0.9–2.1)	1.28 (0.9–1.8)

^aOdds ratio adjusted for study year, municipality population size, migration status, parents’ occupation, sex, parents’ education level, and family structure; ^b Ref = Reference group of each characteristic. Statistically significant results in bold. ^c n/a = not applicable due to too few children with problems in the group. Health-related quality of life is measured by EQ-5D-Y (The prevalence of problems related to self-care were very low and therefore not presented here). ETICS study 2005, 2009

Girls with a foreign origin had statistically significant less perceived pain (OR of 0.59) compared to girls with a Swedish origin. Moreover, children of student parents were identified as a disadvantaged group in both boys and girls. Girls of student parents had a statistically significant double odds of mood problems compared to girls of skilled non-manual worker. While boys who had student parents had more than double as high odds in pain problems and four times higher odds in mobility problems compared to boys of skilled non-manual workers. On the other hand, the association between parents’ education and mood problems was only statistically significant among boys. Boys of parents with a medium level of education had an OR of 0.77 in mood problems compared to boys of parents with high education levels. Regarding family structure, girls living in joint custody had an OR of 1.4 in mood problems compared to girls living with two parents, while no statistically significant differences were detected among boys. However, girls who lived with one parent and another adult had more than double as high odds of problems performing usual activities compared to those living with two parents. Compared to those living with two parents, girls living mainly with one parent had OR’s of 1.6 in mood problems and 1.4 in pain while, boys living mainly with one parent had an OR of 1.5 in mood problems compared to those living with two parents. Contrary to non-stratified data, living with one parent and another adult did not show statistically significant results regarding mood problems.

Except for the statistically significant results when stratifying by sex, other stratification did not show noteworthy differences e.g., stratifications by study sites, municipalities, municipality population size, migration status, parents’ occupation, parents’ education level, or family structure.

Methodological considerations

A strength in our study is that our results are based on a large sample size, corresponding to approximately 10% of the twelve-year-olds in Sweden, with a relatively high participation rate (60% of all invited participants were included in this study). Still, non-respondents might have higher levels of HRQoL problems. Furthermore, there were differences in HRQoL dimensions between our sample and children with missing parents’ responses. However, we expect these differences to at most marginally affect our associations. Moreover, the sample has been considered representative of children of the same age in Sweden according to previous studies based on geographical representation, healthcare consumption, unemployment rates and HRQoL [39, 40]. The high response rate and the similar methods in 2005 and 2009 support the generalization of our results to twelve-year-olds in Sweden. However, these results might not apply on other age groups or countries with different living standards than Sweden.

There are both a strength and a limitation in our use of EQ-5D-Y in this study. On the one hand, there is support from Swedish-based studies on the validity and reliability of this tool including test retest reliability and discriminative validity [25, 26]. On the other hand, the ceiling effect might have limited our analysis.

Even though geographical data were collected in the study, we did not consider using a multilevel analysis as initial analyses showed very little contribution to the models.

We apply logistic regression for four of the EQ-5D-Y domains and with stratified analyses by sex and estimates for the exposure variables, we present results from 184 tests. There is therefore a risk of spurious associations, and adjustments for multiple testing might be justifiable. However, we followed recommendations from methodologists [41, 42] and did not adjust for multiple testing as this might adventure the possibility to provide valid information about associations. Instead, we recommend that the associations provided are interpreted with care due to the multiple testing.

One recognized limitation of our study is that it does not include some relevant variables to children’s health such as social capital or neighborhood’s characteristics. Another major limitation in regard to migration analysis is that there is no differentiation between the reasons of migration e.g. labor, study or asylum seekers. This might influence the results by bringing heterogeneous groups in one category. Although the sample size of our study is large, the number of reported problems by some groups was relatively low in some of the EQ-5D dimensions especially after stratification.