Assessment of selection bias in a health survey of children and families – the IDEFICS Sweden-study

Abstract

BackgroundA health survey was performed in 2007–2008 in the IDEFICS/Sweden study (Identification and prevention of dietary- and lifestyle-induced health effects in children and infants) in children aged 2–9 years. We hypothesized that families with disadvantageous socioeconomic and -demographic backgrounds and children with overweight and obesity were underrepresented.MethodsIn a cross-sectional study, we compared Swedish IDEFICS participants (N=1,825) with referent children (N=1,825) using data from Statistics Sweden population registers. IDEFICS participants were matched for age and gender with a referent child living in the same municipality. Longitudinal weight and height data from birth to 8 years was collected for both populations (n=3,650) from the children’s local health services. Outcome measures included the family’s socioeconomic and demographic characteristics, maternal body mass index (BMI) and smoking habits before pregnancy, the children’s BMI standard deviation score (SDS) at the age of inclusion in the IDEFICS study (BMISDS-index), and the children’s BMI-categories during the age-span. Comparisons between groups were done and a multiple logistic regression analysis for the study of determinants of participation in the IDEFICS study was performed.ResultsCompared with IDEFICS participants, referent families were more likely to have lower education and income, foreign backgrounds, be single parents, and have mothers who smoked before pregnancy. Maternal BMI before pregnancy and child’s BMISDS-index did not differ between groups. Comparing the longitudinal data-set, the prevalence of obesity was significantly different at age 8 years n= 45 (4.5%) versus n= 31 (2.9%) in the referent and IDEFICS populations, respectively. In the multivariable adjusted model, the strongest significant association with IDEFICS study participation was parental Swedish background (odds ratio (OR) = 1.91, 95% confidence interval (CI) (1.48–2.47) followed by parents having high education OR 1.80, 95% CI (1.02-3.16) and being married or co-habiting OR 1.75 95% CI (1.38-2.23).ConclusionFamilies with single parenthood, foreign background, low education and income were underrepresented in the IDEFICS Sweden study. BMI at inclusion had no selection effect, but developing obesity was significantly greater among referents.

Behavioural effects of a community‐oriented setting‐based intervention for prevention of childhood obesity in eight European countries. Main results from the IDEFICS study

“…The number of 'drop-outs' was n = 5,187 (32% of total study population in T 0 ). Dropout analyses extensively reported before (16)(17)(18) showed that the percentage of overweight children, the mean BMI z-score and percentage body fat were slightly increased in the dropout group. This might be due to drop-outs having more often less educated parents, a migrant background and a lower well-being score.…”

Parental marital status and childhood overweight and obesity in Norway: a nationally representative cross-sectional study

“…Second, an obvious limitation is that our cross-sectional design provided no basis for studying causality; whether the development of overweight and obesity was initiated before the divorce or whether the impact on the children's weight status was primarily attributed to marital conflict or the divorce. Third, one cannot exclude the possibility that a higher proportion of overweight children were absent from school on the day measurements were taken and were therefore over-represented among non-participants, which in turn could imply that children of divorced parents were under-represented in NCG, as previously stated 25. If so, the associations shown in this study could be underestimated, but, given that the children were recruited into the NCG by the school health service, selection bias is most likely not a big issue in our study.…”

Parental education and family income affect birthweight, early longitudinal growth and body mass index development differently

“…The first cohort included 1,825 children recruited from 2007 to 2008 who participated in the multicentre European Identification and Prevention of Dietary‐ and Lifestyle‐induced Health Effects in Children and Infants (IDEFICS) study . The second cohort was a register‐based population of 1,825 children matched to the IDEFICS population by municipality, gender and age , and was created for a study of possible selection bias in the Swedish IDEFICS study , using the unique personal identity number allocated to all Swedish residents . Complete cross‐sectional data from 2007 on parental education, family disposable income and parental national background were available for 3,647 of the 3,650 children.…”

“…Data on maternal BMI and smoking habits, the children's gestational age and diagnoses at birth were retrieved from the Medical Birth Register at the National Board of Health and Welfare . In addition, data on subsequent weight and height on the pooled study cohort were collected from records at child health centres and school health services during 2010 and 2011. Growth data were collected for the following ages: birth, 6 months, 1.0, 1.5, 2.5, 4.0, 5.5 and 8.0 years.…”