Opposite associations of household income with adolescent body mass index according to migrant status: Hong Kong’s “Children of 1997” birth cohort

In economically developed settings, household income is usually inversely associated with child and adolescent adiposity, but this association may not extend to migrants. Hong Kong is a unique developed setting to study how household income and adolescent adiposity vary by migrant status given many Hong Kong-born Chinese children were born to parents who migrated from neighboring provinces of Mainland China. We examined differences between the associations of absolute household income vs. relative household income on adolescent body mass index (BMI) z-score or overweight (including obesity) status using a linear or logistic model in a Chinese birth cohort (n = 5613, 68% follow-up). We focused on whether the associations differed by mother’s or father’s migrant status (birthplace). No association was found between absolute household income and BMI z-score among adolescents with either native or migrant mothers. However, the association of relative household income with BMI z-score varied by mother’s migrant status (P-values for interaction <0.0005). In adolescents of native born mothers, greater relative household income deprivation was associated with higher BMI z-score (0.03 z-score per USD 128 difference in Yitzhaki index, 95% confidence interval (CI) 0.01 to 0.05). However, in adolescents of migrant mothers, greater relative household income deprivation was associated with lower BMI z-score (−0.05, 95% CI −0.09 to −0.01). Similar association of relative household income with overweight (including obesity) status was found in adolescents of native born mothers but not in adolescents of migrant mothers. Relative income (mediated by social comparisons with others in society) appears to be relevant to adolescent adiposity, but the association depends on the interplay between individual characteristics (migrant background) and societal context.


54
The nutrition transition is associated with a reversal in the social patterning of child and 55 adolescent adiposity with economic development. 1 In economically less developed countries, 56 lower household income is associated with under-nutrition, while higher incomes are 57 associated with greater prevalence of child and adolescent adiposity. 2, 3 This pattern becomes 58 more mixed with a rise in living standards. 4,5 In economically developed countries, such as 59 the United Kingdom, a mixed social pattern prior to the late 1980s 2 has shifted to a largely 60 inverse association of household income with child and adolescent adiposity since the 1990s. 6 61 However, even within affluent countries, such as the United States, a positive association 62 between income and adiposity can be still observed in minority ethnic groups 7 and among 63 immigrants of lower socioeconomic position. 8

65
Decomposing income into absolute (income per se) or relative (social comparisons of income) 66 at household level may provide insights into how income affects adiposity. 9 Absolute 67 household income confers the ability to purchase and access material resources, such that poor 68 households in any neighbourhood are likely affected. Food expenditure composes a larger 69 income share in poor households 10 and lower-priced calorie-dense, nutrient-low food that is 70 more filling encourages unhealthy diet in low-income children. 11 They also participate less 71 organized physical activity 12 and underutilize preventive health service. 13 Relative household 72 income refers to a lack of resources compared with others. 14 A relatively poor household may 73 not be deprived in any absolute sense (e.g. lacking the ability to afford food, shelter, clothing), 74 but they may be deprived in the sense of not being able to participate fully in the norms and 75 lifestyles set by the middle class. For example, an adolescent growing up in a relatively poor 76 household may own a pair of shoes, but they might not be the "right" type to participate in 77 certain sports activities, such as basketball. Relative deprivation can also increase stress and adiposity is a concern, or higher weight if undernutrition is a concern. Given adiposity is 90 shaped by socio-cultural norms, children from migrant families might be more or less prone to 91 adiposity than those from native families even within countries and income levels depending  To date, the relative contribution of absolute and relative income to adolescent adiposity has 96 not been systematically studied, but is important to distinguish for the design of 97 interventions. 18 If only absolute income matters, increasing income by the same amount for 98 every household (e.g. providing universal basic income) would be relevant. If only relative 99 income matters, reducing income inequality or promoting intragenerational relative income 100 mobility would be more relevant. If both absolute and relative income matter, then increasing 5 relative income to adolescent adiposity. The Hong Kong population was mainly formed in the 106 mid-20th century by an influx of young workers from neighbouring provinces such as 107 Guangdong after the World War II. 19  We considered age-and sex-specific BMI z-score (standard deviation) at age 13 years (using 168 the closest measurement at 12-<15 years) relative to the 2007 World Health Organization 169 growth references for 5-19 years. 25 We also defined adolescent overweight (including obesity) 170 as a BMI for age and sex corresponding to an adult BMI of ≥25 kg/m 2 using the International 171 Obesity Task Force cut-offs. 26 Due to the relatively small number of obese adolescents (n = 172 215; 3.8%), overweight adolescents here included those who were obese.  Complete information on BMI at ~13 years and household income was available for 68%. We 203 used multiple imputation with inverse probability weighting (IPW/MI) to recover the entire 204 sample. 28 We first imputed any missing potential confounders (i.e. parent's age, migrant status 205 and highest parental education), but not exposures (absolute or relative income) or outcomes 206 (BMI z-score or overweight (including obesity) status), 20 times for respondents with income 207 and BMI based on a flexible additive regression model with predictive mean matching 29 208 incorporating data on the outcomes, exposures, and covariates (sex, birth order, mode of 209 delivery, second-hand smoke exposure, infant residency, parents' age, parents' migrant status, 210 highest parental education, highest parental occupation, housing and interaction terms between 211 income measures and parents' migrant status as well as those between absolute and relative 212 income measures). 30 We predicted the probability of exclusion due to missing income or BMI 213 based on these covariates to generate the IPW. We combined the results from 20 imputed 214 dataset adjusted for the IPW into single estimates with confidence intervals to allow variability 215 between imputations. 28 We also performed an available case analysis for comparison.

226
Of the original 8,327 recruited, as of January 2016, 29 had permanently withdrawn. Of the 227 remaining 8,298 adolescents, 2,685 were excluded because of missing income or BMI.

231
Parents with lower absolute household income were younger, less educated, and more often 232 migrants from Mainland China or elsewhere (Table 1).

234
The associations of absolute or relative household income with BMI z-score varied by 235 mother's migrant status (P-values for interaction <0.0005), but not by father's migrant status 236 (P >0.46). The associations of relative household income with BMI z-score did not vary by 237 absolute household income (P >0.18).  Table 2 shows that when each income measure was considered separately, greater relative 240 household income deprivation (Yitzhaki index) or lower relative household income rank was 241 each associated with higher BMI z-score in adolescents of native born mothers, but with lower 242 BMI z-score in adolescents with migrant mothers (Model 1). Absolute household income was 243 not associated with adolescent BMI z-score. The associations remained unchanged when all 244 income measures were considered together (Models 2 and 3). The estimates of relative 245 household income for adolescents of native born mothers were attenuated, but those for 246 adolescents of migrant mothers remained similar after adjustment for mother's and father's  The available case analysis (with listwise deletion of missing data) produced similar results 260 (Appendix Table 1). In sensitivity analyses, the opposite associations of relative household 261 income deprivation or rank defined by residential neighbourhood or parents' migrant status 262 with adolescent BMI z-score according to mother's migrant status were similarly found 263 (Appendix Table 2). In a recent economically developed Chinese setting, relative household income (deprivation or 267 rank) was associated with higher BMI z-score and overweight (including obesity) in 268 adolescents of native born mothers, but with lower BMI z-score in adolescents with migrant 269 mothers from Mainland China or elsewhere. Absolute household income was unrelated to 270 adolescent adiposity. In adolescents of native born mothers, these associations were explained 271 by adjustment for parental education but not in adolescents of migrant mothers. Our finding of a lack of association between absolute income and adolescent BMI z-score 287 differs from previous reports, mainly from North America. 31-33 Hong Kong has highly 288 accessible and affordable health care, free universal education, strong family ties and a social 289 safety net, which might buffer the effect of income disparities between households and 290 neighborhoods. Extensive, affordable public transport also facilitates common usage of 291 societal infrastructure across districts.

293
To our knowledge, this is the first study distinguishing the role of absolute and relative 294 household income in adolescent adiposity. As in developed countries, 6 higher absolute 295 household income was associated with overweight (including obesity) in adolescents of native 296 born mothers. However, the association became null after adjusting for other income measures, 297 and was not observed in adolescents of migrant mothers, unlike the positive association in 298 minorities 7 and migrants. 8 Conversely, relative household income (greater relative deprivation 299 or lower relative rank) was associated with higher BMI z-score and overweight (including 300 obesity) in adolescents of native born mothers, but with lower BMI z-score in adolescents of 301 migrant mothers, with or without adjustment for absolute income, suggesting the underlying 302 mechanism may extend beyond access to material resources. The associations were similar in   Relative household income was associated with adolescent BMI z-score and overweight 358 (including obesity), independent of absolute household income, but the direction varied by        Bold indicates statistical significance; Abbreviation: CI, confidence interval; OR, odds ratio. a Model 1 is unadjusted association of each of the three income measures (absolute household income, relative household income deprivation or relative household income rank) with each outcome (adolescent BMI z-score or overweight (including obesity) status); Model 2 is mutually adjusted associations of absolute household income and relative household income deprivation with each outcome (adolescent BMI z-score or overweight (including obesity) status); Model 3 is mutually adjusted associations of absolute household income and relative household income rank with each outcome (adolescent BMI z-score or overweight (including obesity) status); Model 4 is Model 2 additionally adjusted for mother's and father's age, father's migrant status, and highest parental education; Model 5 is Model 3 additionally adjusted for mother's and father's age, father's migrant status, and highest parental education. b For native born mothers, absolute equivalized household income in Hong Kong dollar (mean+SD): $14,622+20,896; relative income deprivation: $3,958+2,224; relative income rank: 0.60+0.26.
Overweight (including obesity) was defined using the International Obesity Task Force cut-offs as equivalent to an adult BMI of 25 or more; Obesity was defined as equivalent to an adult BMI of 30 or more.
Appendix Bold indicates statistical significance; Abbreviation: CI, confidence interval; OR, odds ratio. a Model 1 is unadjusted association of each of the three income measures (absolute household income, relative household income deprivation or relative household income rank) with each outcome (adolescent BMI zscore or overweight (including obesity) status; Model 2 is mutually adjusted associations of absolute household income and relative household income deprivation with each outcome (adolescent BMI z-score or overweight (including obesity) status; Model 3 is mutually adjusted associations of absolute household income and relative household income rank with each outcome (adolescent BMI z-score or overweight (including obesity) status; Model 4 is Model 2 additionally adjusted for mother's and father's age, father's migrant status, and highest parental education; Model 5 is Model 3 additionally adjusted for mother's and father's age, father's migrant status, and highest parental education. b For native born mothers, absolute equivalized household income in Hong Kong dollar (mean+SD): $14,622+20,896; relative income deprivation: $3,958+2,224; relative income rank: 0.60+0.26.
Overweight (including obesity) was defined using the International Obesity Task Force cut-offs as equivalent to an adult BMI of 25 or more; Obesity was defined as equivalent to an adult BMI of 30 or more.
26 Appendix Bold indicates statistical significance; Abbreviation: CI, confidence interval. a Model 1 is unadjusted association of each of the four relative household income deprivation defined by neighbourhood of residence, mother's or father's migrant status and parental education, or each of the four relative household income rank defined by neighbourhood of residence, mother's or father's migrant status and parental education, with adolescent BMI z-score; Model 2 is mutually adjusted associations of absolute household income and each of the four relative household income deprivation, or absolute household income and each of the four relative household income rank, with adolescent BMI z-score; Model 3 is Model 2 additionally adjusted for mother's age, father's age, mother's migrant status (if appropriate), father's migrant status (if appropriate), and highest parental education (if appropriate).