Page 12 - IJPS-3-1

Exposure to urban life and mortality risk among older adults in China

(2.3%) survived to 2014, 8,179 (24.4%) had 2+ interviews but were lost to follow-
up afterwards, 18,944 (about 56.5%) died between 2002 and 2014, and 5,606 (16.7%)
had only one interview and were lost to follow-up afterwards. Because the survival
status and the mortality exposure for those 5,606 respondents were unknown, they
were excluded from the study. The total valid sample size is 27,906 (= 33,512 - 5,606)
individuals who were recruited from 2002 to 2011/2012 and exposed to mortality risk
from 2002 to 2014. For those who had 2+ interviews and then were lost to follow-up,
information after their last interview was excluded from modeling since their survival
status and the length of mortality exposure were unknown. The sampling procedures
and assessments of data quality of the CLHLS can be found elsewhere and thus are not
detailed here (Gu, Brown and Qiu, 2016; Gu, 2008).
2.2 Measurements

2.2.1 Urban life exposure
Urban life exposure was measured in reference to three stages throughout the life
course of the respondent. Routine measures for urban life exposure include being born
in an urban area (early-life exposure, yes vs. no) and living in an urban area at the time
of the interview (late-life exposure, yes vs. no). We added a measure of urban-related
primary lifetime occupation (PLO) (mid-life exposure, yes vs. no), a binary variable of
primarily doing a non-agricultural job before age 60 (u-PLO) versus agricultural sector
or unemployment before age 60 (r-PLO). We further added migration experience by
asking whether the respondent had a geographic movement beyond their county/city
administrative boundary of birth. The urban-rural definition comes from the National
Bureau of Statistics of China (NBSC, 2002; 2008).
Based on different combinations of these measures of life course urban status, we
obtained three classification schemes for urban exposure with 4, 8, and 14 categories
reflecting degrees of exposure to urban life (see Table 1). These specific categories
are all meaningful types of urban exposure in the context of China. Those who were
born in a rural area, reported a rural PLO, and lived in a rural area at the first CLHLS
interview were defined as no exposure, while those who were born in an urban area,
reported an urban PLO, and lived in an urban area at the first CLHLS interview were
defined as full exposure. Besides these two extreme ends of urban life exposure,
combinations of the measurements above further created a series of categories,
reflecting a spectrum of urban life exposure in the life course (See Table 1).

2.2.2 Outcome

The outcome variable is mortality risk, measured by the duration of exposure (in days)
from the date of the first CLHLS interview in 2002–2011 until the date of death (for
those who died in 2002–2014), the date of the 2014 interview (for survivors), or the
date of the latest CLHLS interview (for those who had at least two interviews but were
lost to follow-up afterwards). The dates of death for those who died between 2002 and
2014 were collected from official death certificates when available; otherwise, they
were collected from the next-of-kin of the deceased respondents and local residential
committees. The mortality data in the CLHLS were of high quality (Gu and Dupre,
2008).
2.2.3 Covariates
To ensure robust results, we controlled for a wide range of covariates that are
associated with mortality (Wen and Gu, 2011; Woods, 2003; Zhao, Sautter, Qiu, et
al., 2017; Zimmer, Kaneda and Spess, 2007). The covariates included age (single
year), sex (men vs. women), educational attainment (no formal education, received
1–6 years of schooling, and received 7+ years of schooling), economic independence
(main financial resource was from own or spouse’s retirement wage/pension or self-
employment vs. from children or other sources), adequate medical services (yes vs.

6 International Journal of Population Studies 2017, Volume 3, Issue 1