Population and sampling
Participants were rural adults in Suixi County, Guangdong, China. The Suixi County population has been documented to be some 1.03 million (47% female), rural (>90% rural residents), and 77% of the working population are employed in farming or farming-related business/industry [21]. Households were randomly selected based on village, street, block, and house numbers. A list of 70 villages was enumerated based on information from the local government website http://cwgk.zhanjiang.gov.cn/. Ten villages were selected randomly. In each selected village, a map was obtained through local sources, and streets and blocks were randomly selected accordingly. In each selected household, the adult with the most recent birthday was invited to participate. If no one was home during the first visit, interviewers would return on the next two consecutive days for interview attempts. Inclusion criteria were 18 years or older and having lived in the village for at least three years. Written informed consent was obtained and small monetary incentives were provided. The San Diego State University Institutional Review Board approved the study.
In July 2009, 454 local residents were contacted and 287 participated in the survey (63% participation rate) during 14 days of data collection, which took place on both week days and weekend days depending on participants' availability. Interviewers kept notes of the gender and estimated age (young, middle-age, old) of each individual they approached and found no differential refusal rates by gender or age. Participants ranged from 18 to 82 years of age (mean = 40, SD = 16), 53% were women, 73% were married, and 65% completed middle school (nine-year mandatory education in China). About 32% were farmers, 18% were employed in non-farming occupations, 23% were self-employed shop keepers, and 26% were unemployed. The unemployed category included those who were retired, those not actively seeking employment, and those between jobs; the unemployed were usually not financially independent and were taken care of by families. It was possible, however, that they worked episodically for money.
Procedures
Face-to-face interviews were conducted by trained interviewers, who were local college students speaking Mandarin and at least one of the two local dialects (Cantonese, Leizhou Dialect). Interviews were conducted in participant's preferred language. About 62% of interviews were conducted in Cantonese, 30% in Leizhou Dialect, and 8% in Mandarin. Verbal explanation and visual aids were provided if participants had problems understanding certain questions. For quality-control purposes, two interviewers were paired for each interview: one asked questions and wrote down answers while the other checked progress with the interview protocol and the accuracy of recorded answers.
Questionnaire design
The questionnaire was developed from existing measures, informant interviews and focus groups. An iterative process of questionnaire development, informant interviews/focus groups, and questionnaire revision was used. Chinese-translated items were back-translated to English to compare with original questions. The final questionnaire was shortened, simplified, and was considered culturally appropriate by focus group participants, informants and interviewers.
Measurement
Physical activity
Physical activity was measured in three domains: occupation, household, and leisure-time, using questions modelled after the Global Physical Activity Questionnaire (GPAQ) [22]. Occupational moderate-to-vigorous physical activity was defined as "activity that causes at least small increases in breathing and heart rate" through occupation; examples were provided such as lifting loads, digging, or farm work. Household physical activity was defined as any housework that involves physical activity, such as cleaning and maintaining the yard. Leisure-time physical activity (LTPA) was defined as any physical activity for the purpose of recreation and/or fitness, such as leisure walking, playing basketball, and martial arts. For each type of activity, two questions were asked about the number of days and hours/minutes per day in a typical week. A weekly time for each activity was calculated from the two questions.
Sedentary behaviour
This was defined as "sitting or reclining". Questions were modelled after GPAQ [22] regarding the days per typical week and time per day spent on each of the sedentary behaviours. Specific behaviours listed were suggested by focus groups as being "the most common sedentary activities" in rural Zhanjiang. These included TV viewing (when this was the primary behaviour and did not include doing other non-sedentary activities while TV was switched on [23]), recreational computer use (prevalent among young people), sitting chatting, playing Mahjong (a popular board game), and driving/riding a car/bus.
Seasonality of activity
Activity patterns of farmers are seasonal. Farmers do moderate-to-high-intensity, long-hour planting and harvesting activities during the farming season and less-intense field maintenance during the non-farming season. In the study sample, the length of the farming season ranged from two weeks to six months, depending on the type of crops under cultivation. To capture seasonality of activities, those in the farmer subsample was asked to estimate the length of farming season every year, and to recall physical activity and sedentary behaviours separately for a typical week during the farming and non-farming seasons.
Neighbourhood characteristics
The Neighborhood Environment Walkability Scale (NEWS) [24] aesthetics and safety subscales were translated and tested in formative interviews and focus groups. Based on their feedback, several items were deleted due to lack of relevance or appropriateness in rural China (e.g. crosswalks, posted speed limits). Items were simplified in language and response categories were reduced from four to two ("agree" vs. "disagree"). The final safety subscale included six items concerning walking/biking safety, crime rates, day-time safety, night-time safety, and street lights. The aesthetics subscale included three items regarding sidewalk condition, trees/shades, and cleanness.
Social modelling
Participants were asked "how many of your family members participate in physical activity for recreation or exercise?" Parallel questions were asked about friends and neighbours. Response options ranged from "none" (1) to "all" (5). Due to a large proportion of the "none" response, variables were dichotomized into "none" and "any".
Data analyses
Activity data were examined for distribution and outliers. Outliers were identified and recoded to the 95th percentile of the distribution. Most of the physical activity and sedentary behaviour variables were highly skewed, and were therefore log transformed.
Analysis of covariance (ANCOVA) compared hours of physical activity and sedentary behaviours across occupations, adjusting for age as a covariate, and separately for men and women. Additional ANCOVA compared hours of activity between men and women for each employment category. For farmers, weighted physical activity hours were calculated based on the lengths of farming/non-farming seasons and activities in each season.
Participation in LTPA was dichotomized as "none" and "any" due to a large proportion of zero values (66%) and the lack of variance in LTPA time for those who reported any (around 80% reported an hour or less per week). TV time was median-split at 12 hours/week. χ2 tests and t-tests examined bivariate associations between independent variables and dependent variables (LTPA, TV time).Variables with a significance level of p < 0.05 were included in the multiple logistic regression model. Analyses were first conducted for men and women separately, and then combined if similar associations were observed for both genders. Social-modelling variables were summed as an overall index in the logistic regression model.
Informant consultation
Member checking [25], a qualitative research method, was used to obtain community feedback and to validate findings. Informants (n = 10) were rural adults who did not participate in the previous survey, and were randomly selected from the villages where the survey interviews were conducted. Key findings from quantitative data analyses were abstracted into simple sentences and graphs, and were presented to informants both individually and as a group. Informants were encouraged to provide feedback and interpretation of findings, which were audio taped and transcribed.