Study Population
We derived data from the China Third National Health Services Survey, which collected data through face-to-face interviews from September 18 to October 20, 2003. Of the 193,689 respondents surveyed, we included individuals who were 18 years of age or older and excluded 3,677 respondents with missing values, resulting in a total of 139,831 (69,748 males and 70,083 females) respondents in our analysis.
The national survey employed a multiple stage cluster sampling method to select the sample randomly. The mainland of China was clustered according to the government administrative geographic system (i.e., county, town and village in rural areas, and city, community, and neighbourhood in urban areas). Firstly, 95 counties and cities were randomly selected from rural and urban areas. Secondly, 5 towns and 5 communities were randomly selected in each county and city, respectively. Thirdly, 2 villages in each town and 2 neighbourhoods in each community were randomly selected. Fourthly, 60 households were randomly selected in each village and neighbourhood, respectively, resulting in about 57,000 households. All family members aged 15 years or older were invited to participate in the face-to-face interview.
Data collection
Medical doctors and nurses conducted the survey. Before the survey, interviewers were trained and practiced interviewing; their understanding and knowledge about the survey method and content were examined through testing. During the survey, interviewers visited each household up to three times on different days and times. Interviewers explained the purposes and confidentiality of the survey, and then invited family members to participate. Respondents could choose not to participate and their participation in the survey was accepted as oral consent. The completeness of questionnaires was checked by a district survey manager at the end of every day. If there was missing information on the survey, individuals would be re-surveyed if possible. After the survey, 5% of households were randomly selected and re-surveyed on 14 questions to examine survey quality; the agreement was 95%. The survey response rate for adults was 77.8% [5].
Demographic Characteristics
Demographic variables included age, sex, marital status, education; rural/urban residence, and geographic region. Educational level was categorized into five categories, illiterate (it was defined as people who could not read newspaper or magazines, or write a short note), elementary school (i.e. those who attended up to 6 years of schooling or were not illiterate for those without schooling), junior high school (i.e. schooling 7 – 9 years), senior high school (i.e. schooling 10–12 years), and college or university or higher (i.e. complete or incomplete of post-secondary school). Residence was divided based on rural and urban area and then economic development. Rural area included towns and villages. Based on economic development, Eastern China, the most developed region, included 11 provinces and metropolitans such as Beijing, Shanghai, and Liaoning. Middle China included 8 provinces, such as provinces of Heilongjiang, Shanxi, and Hunan. Western China, the least developed region, included 12 provinces such as Yunnan, Tibet, In-Mongolia, and Ningxia.
Health Status Indicators
Self-perceived overall wellbeing was assessed using a five -point Likert-type scale of being excellent, good, fair, poor or very poor. Presence of illness in the last two weeks and physician-diagnosed chronic disease in the last six months was recorded. The two-week illness was surveyed by asking: "Have you had any physical and mental discomforts during the last two weeks?" Chronic disease referred to disease diagnosed by medical doctors and occurring in the last 6 months prior to the survey, or chronic disease that was diagnosed more than 6 months prior to the survey but reoccurred within the last 6 months and received treatment. Non-physician diagnosed chronic disease was not included because the validity of self-diagnosed medical conditions depends on the level of the respondent's knowledge and their perceptions on the definition of 'disease' and 'health'. Physician diagnosed chronic disease was further confirmed by asking diagnosis location including community clinics, county hospital, city hospital, provincial hospital, military hospital, and others. Respondent reported up to three specific chronic diseases. The reported diseases were coded and classified using the disease classification scheme designed by China Ministry of Health for the survey.
Quality of life was measured using a seven-item instrument. Respondents were asked about presence and level of severity of their dysfunction and disability in the last 30 days in 1) ability about washing or dressing themselves, 2) ability to do job work or housework, 3) feeling of pain or physical discomfort, 4) ability of concentration on work or study and memory, 5) ability of recognizing familiar people from 20 meters away (with glasses for those wearing glasses), 6) emotional discomfort due to restlessness, and 7) anxiety or depression. Under each item, five itemized answers about presence and severity were provided, including: none, mild, moderate, severe, and extremely severe.
Health determinant
Information about smoking, alcohol consumption and physical exercise was collected. For smoking, the survey asked: "Are you currently smoking?" (with answer: Yes, No) Under the survey question of "Do you drink alcohol?", the three answers were provided: "No or rarely", "Sometimes" (defined drinking < 3 times per week), and "Frequently" (defined drinking ≥ 3 times per week). For exercise, a question of "What is the sport or exercise that you have been regularly doing in the last 6 months?" was asked with providing a list of recreational physical activities, such as running, Tai Chi, Wushu, dancing, and playing balls. Regularity of exercise was not defined in the survey and determined by respondent's perception.
Statistical Analysis
Proportion was employed to describe respondents in demographic characteristics, health status and health determinants. Because of the large sample size and multiple categories in some variables, the P-value for sex difference was not reported. Frequencies of variables in the survey were not weighted because sampling weight was not available. The same sampling method had been used in the previous two National Health Services Surveys in China. Analyses of previous surveys suggest that this sampling method is adequate to generate a nationally representative sample [5]. The survey respondent age and sex composition was comparable with the 2000 census. Finally, multiple logistic regressions were used to generate risk adjusted P-value for gender difference in health indicators after adjustment for demographic characteristics and correction of clustering of individuals within family using the repeated measure [6, 7].
The data were analyzed at the health information centre of the Ministry of Health in Beijing. Confidentiality of the survey was protected through storing the data on password protected computers at the Ministry, and removing personal identifiable information (such as name and address) from the database available for researchers and examining analysis outputs for release of aggregated data by the centre staff.