It is crucial to consider both specific and general contextual effects simultaneously when assessing area-level influences on diet and physical activity behaviours in childhood.
Diet and physical activity in childhood are heavily influenced by the living environment, the family socioeconomic status (SES) being a strong determinant of children’s present and future health.
While diet quality follows a socioeconomic pattern at the individual and neighbourhood levels, evidence on children’s physical activity in relation to parenteral and neighbourhood SES is limited.
WHAT THIS STUDY ADDSPhysical activity in childhood is associated to both SES of the household and economic level of the neighbourhood.
Research on children’s health-related behaviours benefits greatly from using population-based databases of primary care medical records, as in this study.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIn countries with universal health coverage and welfare state, policies should pay special attention to children from lower-middle-income families, not covered by the economic and in-kind benefits intended exclusively for lower-income families.
IntroductionUnbalanced diet and lack of physical activity (PA) in children, as leading global risks to health, have long been a focus of prevention strategies. The WHO European Childhood Obesity Study showed a synergistic positive effect on health of having high levels on PA and having high fruit and vegetable intakes, combined with limited screen-time use and low consumption of sugar-sweetened beverages (SSBs).1 Improving healthy eating and PA behaviour in childhood do not exclusively depend on lifestyle choices, as these habits are heavily influenced by the living environment.2 Specifically, the family environment—family eating and PA behaviours, parenteral education, economics and employment status—has an impact on children’s habits. Among them, the family socioeconomic status (SES) is inherited by children as a strong determinant of their present and future health.3 4
While diet quality follows a socioeconomic pattern, where healthier diets are associated with higher SES,5 6 limited evidence is available in relation to parenteral SES and PA in children, probably due, in part, to the lack of categorisation of PA—leisure time, occupational, etc.7 However, evidence related to leisure time shows that low SES children tend to have higher prevalence of sedentary behaviour than those with high SES.8–10
Besides the family environment, other environmental components also shape food and PA behaviours: the social environment—social norms regarding PA and diet—the built environment—healthy food availability, transportation and walkability—and the macroeconomic framework—trade agreements, food system.2 3 11 In particular, literature suggests that the built environment is influenced by the neighbourhood SES (NSES), as living in low SES neighbourhoods is linked to a higher risk of unhealthy habits, independent of individual-level SES.12 13
Given that healthy habits (HH) in children influence their present and future health, we aimed at quantifying the influence of socioeconomic factors at the individual and neighbourhood level in the Canary Islands, a Spanish outermost region of the European Union with childhood obesity and overweight rates above the national average.8 14 15
Taking advantage of the universal coverage of the Spanish health system, using electronic health records data from the Canary Health Service (CHS) well-child visits to their primary care health specialists, this study aimed to assess the effect of SES and NSES on two health determinants—diet and leisure time PA—among children in the Canary Islands.
MethodsStudy design and participantsCross-sectional study based on 2018 data from all public primary healthcare centres in the Canary Islands. The CHS well-child visits to their primary care health professionals are all electronically registered. They occur at 6, 9 and 12–14 years of age, when diet and PA are assessed, among others. Therefore, we restricted our dataset to individuals aged 6–14 years and with at least one contact with primary healthcare during 2018 (n=1 76 027). Individuals with missing values for SES (n=17), for the food (n=26 261) or PA surveys (n=59 685), or with a small sample size for the geographical unit definition—basic health zone (BHZ)—(n=111) were excluded; the final sample consisted of 89 953 individuals from 101 BHZs (table 1).
Table 1Characteristics of the study participants versus paediatric CHS population
Variables and data sourcesIndividual data were obtained from the electronic health records, including date and country of birth, sex, health area, BHZ, drug cost sharing and data on the food and the PA surveys. These surveys are part of the universal ‘healthy child’ programme of the CHS and are performed by primary care health professionals to children aged 6, 9 and 12–14 years during their annual follow-up visits to the health centre.16
The food survey was used to assess the level of adherence to the dietary guideline issued by the CHS.16 This short food frequency questionnaire evaluated the intake of seven food groups on a two-point yes/no scale, except for the fruit group (fruit pieces consumption per day). Items were (1) daily intake of vegetables, (2) 0.5 L of dairy products intake per day, (3) daily intake of meat, eggs or fish, (4) frequent intake of SSB, (5) frequent intake of sweets, (6) 1–2 times a week intake of legumes and (7) number of fruit pieces per day. Items on fruit and vegetable consumption were combined into an indicator as follows: daily intake of vegetables (1 point), at least two pieces of fruit per day (1 point). Analogously, items on sweets and SSB were used as an indicator of unhealthy eating: daily intake of SSB (0 points), daily intake of sweets (0 points). As a result of combining these two indicators—fruit and vegetables, SSB and sweets—we obtained a diet scale range from 0 to 4 points, where 0 denotes an unhealthy diet and 4 refers to a healthy one.
The PA survey was used to assess the level of adherence to the guideline issued by the CHS.16 The questionnaire classifies PA into five groups, namely inactive/sedentary, partially active, active, very active and extremely active (online supplemental file 1), which we combined into two categories to assessed leisure time PA (table 2). Thus, the PA indicator ranges from 0 to 1 point, where 1 refers to meeting the recommendation of 60 min of PA per day.
Table 2Level of leisure time physical activity
A HH score, ranging from 0 to 4, was defined by the weighted sum of the previous indicators as follows: (1) diet scale/2 + (2) physical activity×2.
The Spanish drug cost-sharing scheme of the National Health Service determines a level of coinsurance for each patient, based on their annual income and labour force status.17 We used this classification as a proxy for the SES of the family of participants: (1) integration minimum income, (2) annual income below €18 000 and (3) above €18 000.
Neither the residential addresses of the participants nor their postcodes were available, so we could not use census tract level data to assess NSES. As the BHZ was the smallest geographical unit in our dataset and these areas are defined according to various demographic and geographical criteria, but above all with the aim of guaranteeing service proximity to users,18 we decided to include as NSES variables the percentage of children within the BHZ whose families: (1) had an integration minimum income—low NSES and (2) had an annual income below €18 000—medium NSES.
Statistical analysisA stepwise multilevel linear regression analysis was performed to examine the simultaneous influence that SES, NSES and the BHZ have on the HH score, differentiating between specific (% low NSES, % medium NSES) and general contextual observational effects.19–21 Model 1 is a multilevel linear regression with BHZ as a random effect variable and keeping individual covariates as fixed effects. Finally, model 2 included the specific contextual observational effects. The final model can be written as:
where
is the HH score estimate for an individual
i
living in an area
j
;
are the individual explanatory variables in an area
j
(age group, sex, country of birth and SES);
, the fixed-effect regression coefficients;
, are, respectively, the neighbourhood covariates of interest (NSES: % low NSES, % middle NSES);
, the area-level random error;
, the individual-level random residual error.
We also calculated (1) the variance partition coefficient (VPC) to measure the proportion of total variance attributable to differences between BHZ and (2) the proportional change in variance (PCV) to know if the general contextual variables were substantially mediated by the specific contextual variables.
ResultsTable 3 shows a description of the study population and its distribution according to their diet, PA level and HH score. No differences were found by sex, except for frequent SSB consumption and fulfilling the recommendation of 60 min of PA per day. Specifically, a higher proportion of boys had a frequent consumption of SSB and an active PA level (p<0.01). Adherence to HH decreased with age, so that older children had worst diet and PA habits than younger ones (p<0.01). Although only 5.2% of the sample were foreign-born children, this group had a higher proportion of frequent SSB and sweets consumption (p<0.01). However, no differences were found in daily intake of fruits and vegetables, or in the active PA level. Even though adherence to HH decreased significantly in children with lowest SES (p<0.01), this gradient is not clear at the neighbourhood level. Regarding the percentage of families in the BHZ with annual incomes below the €18 000 threshold (NSES medium), the higher this percentage, the lower the adherence to HH among children (p<0.01). By contrast, a higher proportion of families in the BHZ with an integration minimum income (third tertile of NSES low) did not necessarily mean lower levels of adherence to the dietary recommendations. However, children in these areas were 3.4 percentage points less likely to meet the PA recommendation (69.3%) than those in the first tertile of NSES low (72.7%).
In the study population, the mean HH score was 2.8 points (SD 1.1), with a 23.5% of the children fully complying with the dietary and PA recommendations (4 points) and a 9.7% of the children having an unhealthy diet and being physically inactive (0–1 points).
Among the physically active children (71.8%), the mean score was 3.4 (SD 0.5), meaning that 1.4 points out of 2 were attributable to the diet (87.8% daily intake of vegetables and 51.2% daily intake of 2 or more pieces of fruit). However, being physical active was more common in children with an occasional intake of SBB and sweets (61.4%) than in those with a daily consumption of fruits and vegetables (48.7%) (online supplemental table 3).
Table 4 shows individual and contextual variables associated to the HH score across the two multilevel regression models. Girls, foreign-born and older children had lower adherence to the diet and PA recommendations. In contrast, children from more affluent families were more likely to follow a healthy diet and being physically active (0.436) than those from low SES families, with individual SES having the strongest association with the HH score.
Table 4Individual and contextual factors associated with the healthy habits score
Regarding the specific contextual effects, the addition of NSES variables did not improve much the model fit to the data, with only a 0.002 (95% CI 0.002 to 0.003) reduction in the BHZ estimated variance. While NSES low was not associated with the HH score, the adherence to the dietary and PA recommendations was 0.007 points lower in the BHZ with a higher percentage of annual incomes below the €18 000 threshold (p=0.03).
Differences in the HH score between BHZ were of minor relevance, as these variations accounted for only 1.8% of the total individual variance in both models. Even though middle NSES was significantly associated with the HH score, the small PCV found suggests that the general contextual effects were not substantially mediated by NSES.
DiscussionWe observed a strong association between SES and the HH score. The geographical component was small since accounting for the BHZ gave a VPC of just 1.8%. This does not mean that there are no geographical inequalities in the adherence to the dietary and PA recommendations, but that there might be other types of contextual factors different than the BHZ conditioning the HH score. As children spent most of their time at school, using the surrounding school areas as geographical unit might have better captured the relevant NSES context that influences children’s habits. Although NSES does not appear to be a relevant factor in the contextual effect of the BHZ, we found a negative association between middle NSES and the HH score, so that the BHZ with lower incomes showed a lower adherence to the dietary and PA recommendations. A difference of 0.18 in the HH score was observed between the BHZ with the highest percentage of families with annual incomes below the €18 000 threshold and the lowest one, which represents around 8.5% of the mean HH score. By contrast, we found no association with being in a BHZ with a higher percentage of population with an integration minimum income. This could be explained due to the high degree of heterogeneity of this group—people with non-contributory pension, with severe disabilities, long-term unemployed, etc17; where people are classified not only according to their income but also according to their current administrative situation. In fact, the low NSES group may have similar income levels than the middle NSES, as the first ones are entitled to income and benefits in kind.
Overall, previous studies have also found an association between healthier diet patterns and higher SES in children.5 6 However, only a few studies showed a positive significant association with PA.7 22 This might be due to the different socioeconomic indicators used, which address different areas of the social construct.23 Studies that use composite scores of SES do not usually find a significant relationship with PA, but when single components of SES are used, such as maternal education level or family income, a positive association is observed in adolescents.22 In Spain, single SES indicators are associated to movement behaviours in children and adolescents,8 24 25 though the effect on PA varies depending on the indicator used. While active play and transport to and from school is more frequent in low SES children, it is easier for children from high SES families to engage in sport, as it usually involves additional costs.7 10
Regarding NSES and its association with dietary and PA habits in children, findings in the literature are inconsistent. While most studies have shown a non-significant association,26 a study on US children found that neighbourhood affluence was inversely associated with unhealthy eating and sedentary behaviours.27 Another study on Scottish pupils also found an inverse relationship between an area-level multiple deprivation index and having a poorer diet.28 A population-based cohort study in Finland showed that individuals living in low NSES areas had a lower consumption of vegetable and fruits from the age of 6 years, and also a lower PA from 12 years old. However, it was the cumulative exposure to low NSES that yielded the greatest differences from childhood to adulthood, increasing the odds of developing cardiometabolic risk factors.29 Thus, it is highly likely that the widespread use of single-point measures of NSES may worse capture the effect of contextual factors that take time to reverse.30 Understanding that interactions with the neighbourhood environment vary with age is crucial to foster meaningful changes in children’s habits. Whereas younger children are more influenced by their caregivers, older ones are more constrained by their peers and directly engaged with the living environment. The effect of these interactions also differs by NSES level, being youth from high NSES less dependent on the environment that surrounds them,31 32 perhaps because they are less exposed to obesogenic environments in more affluent areas than their peers from low NSES.2 4 12 Therefore, interventions to promote HH in children would necessarily need to consider their changing conditions over time and their interactions across the different domains of the living environment to succeed.
There are several strengths and limitations in this study. The universal health coverage allowed us to study all the individuals aged 6–14 years from a Spanish region with a population of more than 2 million people. However, the high proportion of children with missing values, which made multiple-imputed analysis not desirable, might have biased our results but made them representative of paediatric patients who had well-child visits in the study year. While the cross-sectional design does not allow to establish causal associations, in this study, the putative causal direction of the observed relationships can plausibly be hypothesised. Furthermore, the analysis provides valuable descriptive information. Because of the surveys design, we used dichotomised variables instead of quantitative intake to assess adherence to the dietary recommendations, and classified children according to their overall leisure time PA, which may not have captured all aspects of the movement behaviours —PA, sedentary lifestyle and sleep time—. Although the use of a composite dependent variable to assess HH in children did not allow us to make accurate comparisons with other studies, from a public health perspective, our results provide valuable information for the design of health promotion strategies in childhood. Finally, the potential confounding effects of unmeasured variables warrant additional studies without this limitation.
ConclusionsIn this study, HH in children followed a socioeconomic gradient at the individual and the neighbourhood level, so that children from more affluent families and areas showed higher levels of adherence to the dietary and PA recommendations. As the geographical component of the inequalities found was low, further research should focus on other contextual factors conditioning the HH score. Targeting socioeconomic inequalities will be crucial to promote HH in children, with a positive impact in health throughout the life course.
Data availability statementData may be obtained from a third party and are not publicly available. This study is based on electronic health records data from the Canary Health Service well-child visits to their primary care health professionals as part as their care and support. The data used in this study are used under license for the current study and so are not publicly available. Requests to access to electronic health records data are reviewed via the Regional Research Ethics Committee to ensure that the proposed research is of benefit to patients and public health.
Ethics statementsPatient consent for publicationNot applicable.
Ethics approvalThis study was approved by the Regional Research Ethics Committee of the Hospital Universitario de Gran Canaria Dr. Negrín, Canary Islands, Spain (2020-287-1).
AcknowledgmentsThe authors would like to thank individuals from the General Directorate of Healthcare Programmes of the Canary Islands Health Service who provided the dataset used in this study.
留言 (0)