Page 79 - IJPS-8-1

International Journal of
Population Studies Urbanization and body weight

40% of China’s population, or 549 million people. Survey ≥0.85 for women (the World Health Organization, 2008);
communities were drawn through a stratified, multistage or a WHtR >0.5 (Browning et al., 2010). As a robustness
random sampling process from cities, suburbs, towns, and check, overweight is also defined as a BMI ≥ 25 kg/m ,
2
villages designated by China’s National Bureau of Statistics. using the World Health Organization guidelines (1998)
In each community, 20 households were randomly and abdominal obesity is WC ≥90 cm for men and WC
selected and all household members were interviewed. The ≥80 cm for women according to the International Diabetes
response rate at the individual level is 88%. In addition Federation guidelines (IDF, 2006). The CHNS did not
to individual-level data, the CHNS collects background collect data on WC until 1993, and thus, all measures of
characteristics of the survey communities. Details on the body weight status that involve WC are from solely the
design and sampling of the CHNS are available elsewhere 1993–2015 period.
(Popkin et al., 2010).
The primary independent variable of interest is a
This study draws on data from nine waves of the survey community-level urbanicity index designed specifically for
(1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011, and 2015) the CHNS data (Jones-Smith & Popkin, 2010). Capitalizing
that span more than 20 years. Three municipal cities were on the rich community-level data in the CHNS, the
added to the sampling frame in 2011 and three more urbanicity index captures 12 dimensions of urbanization for
provinces were added in 2015. Respondents from these each community in each wave, including population density,
new sites were excluded, because they might obscure the economic activity, traditional markets, modern markets,
long-time trends in the other provinces since 1991. A total transportation infrastructure, environmental sanitation,
number of 69,582 (93.5%) out of 74,419 age-eligible communications, housing conditions, average education
person-year observations had valid values of body weight level, socioeconomic diversity, health infrastructure, and
and height between 1991 and 2015 (69,175), or valid values social services. Each of these dimensions is measured by
of waist circumference (WC) between 1993 (see the next one or multiple variables and assigned ten possible points,
section) and 2015 (60,760), or both (60,353). Among them, resulting in a maximum value of 120 points summed
1818 (2.6%) observations were excluded due to missing across the 12 dimensions, with higher values indicating
data on covariates. The overall analytical sample consisted greater urbanization. Detailed information on this index
of 67,764 person-year observations (35,255 females and is available elsewhere (Monda et al., 2007; Jones-Smith &
32,509 males) contributed by 21,029 individual adults Popkin, 2010).
(10,983 female and 10,046 male) from 241 communities.
To maximize the statistical power, the final sample size was To account for biological differences in body weight
allowed to vary depending on the number of valid values status and weight gain trajectory, the full sample is divided
of each outcome variable. into female and male subsamples. Other demographic
control variables include age (in years), age-squared, and
3.2. Measures marital status. Socioeconomic backgrounds are captured by
The dependent variable, body weight status, is captured respondents’ educational attainment (no formal education,
in two ways: (1) BMI, calculated from body weight (in primary school, middle school, high school, or college and
kilograms) and height (in centimeters), which taps general above) and annual household per capita income (adjusted
obesity and (2) WC, waist-to-hip ratio (WHpR), and WHtR, for inflation and divided into quartiles). Dummy variables
all of which tap abdominal obesity. All the anthropometric indexing provinces and survey waves are included to adjust
measures were taken by experienced healthcare workers. for spatial and temporal fixed effects, respectively.
While widely used as an indicator for measuring whole- 3.3. Methods
body obesity, BMI is not the best measure for abdominal
fat accumulation. In several populations, measures of To assess the robustness of the findings, two methods that
abdominal obesity, such as WC, WHpR, and WHtR, were vary in their strengths and limitations were employed to
found to be superior to BMI for predicting cardiovascular disaggregate between- and within-community components
disease risk and for obesity screening (Yusuf et al., 2004; Li of the urbanization-body weight associations. The first,
et al., 2006; Knowles et al., 2011). standard method is to rescale the time-varying urbanicity
index by community-specific mean centering. Borrowing
We also examine dichotomous outcomes of body the notations from Curran and Bauer (2011), let z be
jt
weight status. Overweight status is defined as a BMI the urbanicity index score for community j at time t, the
≥24 kg/m , using the Chinese Center for Disease Control mean-centered score, denoted by  z , can be calculated as:
2
and Prevention guidelines (2003). Abdominal obesity jt
is defined as a WC ≥90 cm for men and WC ≥85 cm for  z
women (JCDCG, 2007); a WHpR ≥0.9 for men and WHpR jt z z j (1)
jt

Volume 8 Issue 1 (2022) 73 https://doi.org/10.36922/ijps.v8i1.334