Surveillance and study sample
A cross-sectional study on chronic disease and its risk factors was conducted at 10 surveillance districts/counties (Lianhu, Huayin, Chencang, Meixian, Jingyang, Baota, Huangling, Shangzhou, Xunyang and Lueyang) in Shaanxi province from October to December 2015. The ethics committee of the Chinese Center for Disease Control and Prevention approved the implementation of the survey and written informed consent was obtained from each participant.
Complex multistage cluster sampling design was used to select the participants. In the first sampling stage, three subdistricts or townships were selected from each surveillance district/county with systematic sampling. In the second stage, two villages or communities were selected from each chosen subdistrict or township with the same sampling method. In the third stage, households were selected from each chosen village or community using simple random sampling method. Finally, all individuals at or above 18 years old from 45 chosen families were invited to participate the surveillance.
The sample size was calculated according to the China Chronic Disease and Risk Factor Surveillance (CCDRFS), which conducted in 302 surveillance districts/counties nationwide. , where u=1.96; p was 9.7%, which was the prevalence of type II diabetes in China in 2010; deff (the design effect) was 3; r=20%, which was the relative error, d=20%×9.7%。Taking into consideration the demographic differences and urban-rural disparity in China, the sample size was stratified by 31 provinces, autonomous regions, and municipalities of mainland China and urban/rural. A total of 185000 participants were designed to select for the surveillance nationwide. 6120 participants were assigned to investigate in Shaanxi province, which was a typical area of northwest China, environmentally and culturally [19]. Study has shown that the estimates were representative of Shaanxi population [20]. Details of establishment, history and good representativeness of CCDRFS were published elsewhere [21, 22]. In total, 6174 interviews were fully completed in Shaanxi province.All questionnaires, physical and laboratory examinations were developed by standard procedures. The qualified staff from local CDC and doctors from local hospital as investigators engaged in the field activities after training courses and standard exams. We collected the behavioral risk factors for NCDs by face-to-face interviews. Height, weight and blood pressure of each participant were measured. Blood glucose was conducted in local laboratories and blood lipids in KingMed Diagnosis, a third party testing company in Guangzhou. Details on surveillance design, laboratory examination procedures, blood sample transport and quality control has been published elsewhere [23, 24].
Measures
We invited participants to answer questions about their tobacco use, alcohol drinking, diet pattern, physical activity. Physical measurements including height, weight were obtained to calculate body mass index (BMI) and define individuals who were overweight or obesity. Blood pressure, blood glucose and blood lipids were tested at laboratories. Current smokers were defined as participants who reported that they smoked ‘every day’ or on ‘some days’ currently. Those who reported ‘no’ were defined as non-current smokers. We define harmful use of alcohol as consuming pure alcohol ≥ 61g per drinking day for men and ≥41g for women [25]. Food frequency questionnaires (FFQ) were used to collect information on fruit and vegetable intake. Consumption of less than 400 g of fruit and vegetables combined per day was considered insufficient [26]. We used Global Physical Activity Questionnaire (GPAQ) to assess physical activity of each participant [27]. According to the GPAQ analysis guide, Physical inactivity was defined as those with ≤ 150 minutes of moderate physical activity or 75 minutes of vigorous activity or an equivalent combination of moderate and vigorous PA achieving less than 600 MET-minutes (Metabolic Equivalents) per week [28].
Each participant had three times blood pressure measurements in succession using a calibrated electronic blood pressure device (HBP-1300, Omron Co, Kyoto, Japan), with at least 1 minute rest between measurements. The second and third measurement was used in our analysis. Blood samples of each participant were drawn to measure plasma glucose and serum cholesterol after fasting for ≥ 10 hours. Plasma glucose concentration was tested at the local laboratory under a standardized quality control. Serum samples were collected within 2 hours and tested in the laboratory of a third party company (KingMed Diagnosis, Inc., Guangzhou, China), which was certificated by China’s Ministry of health. Body mass index (BMI) ≥ 24 was defined as overweight and obese [29]. Blood pressure ≥ 140/90 mmHg was defined as raised blood pressure [30]. According to WHO 1999 criteria, fasting plasma glucose level ≥ 7.0 mmol/L or taking medication for diabetes mellitus was diagnosed as raised blood glucose. Total serum cholesterol ≥ 6.22mmol/L was defined as raised total cholesterol based on the Guidelines for Prevention and Treatment of Adult Hyperlipidemia in China [31]. The four behavioral risk factors and four biological risk factors are the main indicators investigated and measured based on the recommendations by WHO [32].
Statistical analyses
In our research, an analysis weight of each participant was considered due to the complex multistage sampling. It was obtained based on the surveillance sampling scheme and post-stratification factor, which adjustments for age and gender using 2015 Chinese population estimates from National Bureau of Statistics of China.
The overall distributions of participants’ characteristics, including demographic characteristics, socioeconomic status, rural/urban residence, were examined. Prevalence of each risk factor was presented and χ2 tests were performed to test its distribution difference. Distribution of number of risk factors (range, 0–8), as well as the prevalence of first 15 leading clustering patterns of the eight risk factors, were then presented. Next, the mean number of risk factors that each participant had was determined to manifest their clustering pattern within individuals. Finally, we explore the independent effects of demographic and socioeconomic covariates on risk factors’ clustering pattern by modeling an ordinal logistic regression (ordinal number of risk factors was the dependent variable). We carried out all statistical analyses using SPSS version 25.0 with weighted data.