Table 3.

Multilevel multinomial logistic regression model for the discordant and concordant categories of family income, neighbourhood income and deprivation in relation to underweight, overweight and obesity categories compared with normal weight in children aged 0–12 years

Exposure variable	Model 1 (adjusted for child age, sex, number of family members, maternal ethnicity) OR (95% CI)
	Underweight (zBMI<−2)	Overweight (1<zBMI≤2)	Obesity (zBMI>2)
Family income and neighbourhood income
Low and low	0.94 (0.58 to 1.52)	1.36 (0.99 to 1.86)	3.28 (2.08 to 5.18)
Low and higher	0.77 (0.48 to 1.22)	0.92 (0.68 to 1.25)	2.05 (1.32 to 3.19)
Higher and low	0.95 (0.62 to 1.46)	0.76 (0.56 to 1.04)	1.37 (0.84 to 2.24)
Higher and higher	1.00	1.00	1.00
Family income and neighbourhood deprivation
Low and low*	0.83 (1.49 to 1.42)	1.61 (1.15 to 2.24)	3.39 (2.09 to 5.50)
Low and higher	0.83 (0.54 to 1.27)	0.91 (0.68 to 1.21)	2.14 (1.41 to 3.23)
Higher and low	0.81 (0.46 to 1.42)	0.91 (0.63 to 1.31)	1.61 (0.90 to 2.91)
Higher and higher	1.00	1.00	1.00

Underweight=zBMI<−2, normal weight=−2≤zBMI≤1, overweight=1<zBMI≤2. Normal weight is the reference category for the dependent variable.

*Low was defined as Q1 and higher was defined as Q2–Q5, where low deprivation score represents the most deprived. The multiplicative interaction between family and neighbourhood income had a p value of 0.09 and for family income and deprivation the p value for interaction was 0.12.

BMI, body mass index; zBMI, BMI z-scores.