Second hand smoke exposure in public venues and mental disorder: a representative nationwide study of China

This study employed a cross-sectional multistage sampling design. In Stage 1, 21 cities, with over 1.5 million in population, were purposefully selected from across China and differentiated by regional location: Nine were located in eastern China, five in central China, and the remaining seven in the west. About 68 % of Chinese provinces were covered by the 21 cities in this study. Most of the sample cities were provincial capitals, but we included a few non-provincial capital cities because of their size (over 1.5 million population) in order to achieve better regional representation. Stage 2 involved selection of residential districts within each city. Two residential districts were randomly selected in the main urban zone of each city. This study included urban residents, the main urban zones were those areas where most permanent residents live, excluding new building districts and sub-districts where many rural migrants reside. In Stage 3, four communities were randomly selected within each residential district. In Stage 4, a family household registration (“hukou”) list was used to randomly sample households in each community. Individuals aged 15 years and older, who had lived in their home for at least a year, were identified within each household [5, 40‐42, 46]. Finally, one eligible participant was randomly selected from each family, with eligibility being determined by birthdate closest to the contact date [46].

Smoking, SHS exposure in Public Venues, mental status, and other variables were assessed through a self-administered questionnaire. The questionnaire was personally delivered to each participant by a trained field staff member who also answered questions, collected the completed questionnaires and reviewed them to make sure they were complete. This is the conventional data collection procedure for population-based tobacco control surveys such as the GATS tobacco control survey [5, 40]. The appropriateness of this data collection process is reinforced by evidence that self-report bias in smoking research is minimal [14, 29, 37]. Since smoking represents normative behavior for adults in China, any social expectation of accurate reporting about smoking and related information is plausible but a minor concern [42]. Field staff members were fourth-year medical students from local medical colleges, who had received a 1-day training on study protocol and interviewing procedures. The questionnaire was administered privately in participants’ homes or in a designated quiet place, such as a backyard or community park. Survey were conducted on Saturdays, Sundays, weekday evenings or other times when participants were available. After receiving oral instructions by staff about the survey and questionnaire, the questionnaire was given to each participant to complete. Completion took approximately 30 min. Participants could ask staff if they were confused about any items on the questionnaire. For participants with low educational levels or physical limitations, a survey staff member read the entire questionnaire. Each participant was given an opportunity to clarify confusing questions and adequate time to complete the questionnaire. Participants were requested to resolve any omissions, as appropriate, and were given a token of appreciation (toothbrush and toothpaste, or other small gifts valued at approximately US$ 1.00) for their time.

The same research protocol was utilized across all 21 cities to assure homogeneity of interview and data collection. The survey protocol including the use of small gifts and the questionnaire was approved by the Ethics Committee at the Medical Center, Zhejiang University, and verbal consent was obtained from all participants or their guardian prior to the commencement of the study. Our methods have been extensively employed in smoking research in China, and they possess acceptable validity [5, 46].

A current smoker was defined as someone who smoked cigarettes at the time of the interview. Smokers were excluded from the study. Only non-smokers participated in this study.

Sociodemographic information was collected on the variables of: age, gender, ethnicity, educational level, occupation, person income, and regional location. Person income was estimated as the average income per person in a household during the prior year.

For the purpose of this study SHS exposure included exposure in public venues such as workplaces and public places [23, 46]. With regard to workplaces respondents were asked how often they were exposed to SHS in all indoor work areas. Possible responses included: never, rarely, sometimes, often, and always. For retired or unemployed respondents, “workplace” referred to the places where they spent much of their time such as part-time employment sites, and places where they may participate in leisure activities in their community or elsewhere. For students, “workplace” covered such places as classrooms and libraries [46]. Concerning SHS exposure in public places, participants were asked two questions: (1) “Have you been in restaurants, hospitals, shops, buses, and other public places in the past 6 months?” (yes/no), and (2) If “yes”, they were asked how often they were exposed to SHS in these facilities. Response options were never, rarely, sometimes, often, and always [5, 43, 46].

Mental health status was measured by the Chinese Health Questionnaire (CHQ), which was derived from the General Health Questionnaire (GHQ), and has been used to screen for mental disorders in community settings [8, 45]. This questionnaire has been widely used to assess mental disorders in China and has been shown to be an appropriate indicator of mental health status [12, 18, 27, 45]. The CHQ is a self-administered, 12-item instrument designed for detecting mental disorders in the community and among primary health care patients. It has acceptable reliability and validity and is commonly utilized for research purposes in community and primary care settings in China [12, 18, 27]. It has a four-point scale for responses: “not at all” and “same as usual” both = 0 and “rather more than usual” and “much more than usual” = 1. The total score, obtained by summing up the scores for the individual items can range from 0 to 12, and measures the severity of mental disorders. A cut-off score of 3 or more on the CHQ was classified as a higher score, which signified mental disorders [1, 45].

All data were entered into a database using Microsoft Excel. The dataset was then imported into SAS (9.3 version) for the statistical analyses using the SAS 9.3 survey procedure [32]. Descriptive statistics were calculated for SHS exposure status. Both unadjusted and adjusted logistic regression methods were considered in the analyses. The unadjusted method used only the key factors of interest as independent variables in the analyses, while the adjusted method added all of the possible confounders listed in Tables 1 and 2 as covariates in the logistic models. For this analysis, we operationalized our dependent variable, mental disorder, as a binary response (no = 1, yes = 0). The independent variables in this analysis were those emerging as statistically significant in the unadjusted method. All categorical, they are listed in Table 2. The first category in each variable served as the referent in the multiple logistic regression analysis. Backward stepwise regression is a preferred method for exploratory analyses, where analysis begins with a full or saturated model and variables are eliminated from the model in an iterative process. P-values less than or equal to 0.05 indicated statistically significant differences.

All analysis were weighted [3, 10]. Weights included (1) sampling weights as the inverse of the probability of selection, which were calculated at regional, city, district, and community levels, and they were multiplied together for each level. We did not consider households in our weighting, since family size was rather uniform given population policy and family planning (approximately 3 people per household); thus, the probability of participant selection was essentially the same across households. (2) The non-participation weight covered household and individual aspects. Household-level non-participation weights were calculated at the city level. Individual-level non-participation weights were based on a combination of city, age, and sex, and the corresponding weight was the inverse of the participation rate. The overall non-participation weight was the product of the household and individual weights (3) post-stratification weights reflected a combination of sex (male, female) and age (<25 year, 25–34, 35–44, 45–54, and 55+) based on estimated distributions of these characteristics from a national survey [24]. The final overall weights were computed as the product of the above three weights.