Page 83 - IJPS-8-1

International Journal of
Population Studies Urbanization and body weight

Table 1. Descriptive statistics of the control variables. method. For BMI, the overall association was driven
by both between- and within-community differences,
Control variables Women Men although the latter played a relatively minor role and was
Age in years (mean [SD]) 42.5 (12.6) 42.1 (13.0) only marginally significant at α = 0.1 level. For example,
Marital status (%) according to the standard disaggregation, BMI would
Never married 9.8 15.1 increase by 0.024 unit for every one-unit increase in
Married 84.9 82.0 between-community difference in urbanicity index, but
Divorced/widowed 5.3 2.9 only by 0.005 unit for every one-unit increase in within-
Educational attainment (%) community difference. The results were almost identical
according to the growth curve disaggregation.
No school 16.3 4.1
Elementary school 29.6 25.8 4.4. Regression results for overweight and
Middle school 31.9 40.3 abdominal obesity
High school 17.6 22.9 Table 3 reports multilevel regression disaggregation results
College or above 4.6 7.0 for the longitudinal associations of urbanization with the
Per capita household income (%) overweight and abdominal obesity measures. In the female
1 quartile 25.2 24.5 subsample, urbanicity index was positively associated
st
with both overweight- and WC-based abdominal obesity
2 quartile 25.0 24.9 measures, but unrelated to WHpR or WHtR (Model 1).
nd
3 quartile 25.0 25.0 The standard disaggregation showed the association
rd
4 quartile 24.8 25.6 between urbanicity index score and the overweight
th
Province (%) measure being driven by both between- and within-
Liaoning 9.7 9.3 community differences (Model 2), whereas the growth
Heilongjiang 8.7 8.8 curve disaggregation suggested that within-community
Jiangsu 12.0 11.6 difference did not played any significant role (Model 3).
In contrast, the two disaggregation methods consistently
Shandong 11.0 10.7 showed that the association between urbanicity index score
Henan 11.6 11.4 and the WC-based abdominal obesity measure was entirely
Hubei 11.4 11.6 driven by between-community difference in urbanization.
Hunan 11.4 11.3 In the male subsample, urbanicity index score was again
Guangxi 12.4 13.3 significantly and positively related to all the four measures
Guizhou 11.8 12.0 of overweight and abdominal obesity. For abdominal
Wave (%) obesity, the two disaggregation methods found that the
1991 11.0 10.7 association was attributed to between- but not within-
1993 10.3 10.1 community difference, regardless of which measure was
1997 12.0 13.0 used. For the overweight measure, between-community
difference again played a much stronger role than within-
2000 11.0 11.0 community difference (i.e., log odds = 0.048 vs. 0.011
2004 11.4 11.4 according to the standard disaggregation, and 0.051 versus
2006 11.2 11.0 0.01 according to the growth curve disaggregation).
2009 11.3 11.3 4.5. Sensitivity analysis
2011 10.7 10.4
2015 11.1 11.2 Two sensitivity analyses were conducted. First, Table S5 in
N of person-year observations 32,573 30,880 Supplementary File shows that similar results were obtained
when alternative cutoff points were used to classify overweight
2
(BMI ≥25 vs. 24 kg/m ) and abdominal obesity (WC ≥80 vs.
The patterns were different for men. Urbanicity 85 cm in women). For both men and women, urbanicity index
index was significantly and positively related to all the was positively related to the risks of being overweight and
four continuous measures of body weight status. For having abdominal obesity, and these associations were mainly
WC, WHpR, and WHtR, the overall associations were driven by between- rather than within-community difference.
completely attributable to between-community difference Second, Table S6 in Supplementary File reports coefficient
in urbanization, regardless of the choice of disaggregation estimates from growth curve models of the continuous

Volume 8 Issue 1 (2022) 77 https://doi.org/10.36922/ijps.v8i1.334