Data source and participants
Data for this study are drawn from an epidemiological study of mental disorders in the province of Ningxia located in Western China, where Muslims makes up 35% of the total population (6.4 million) [34]. Inclusion criteria were age 18 years or older and residence for at least six months or longer at the current address. Exclusion criteria were significant impairment caused by brain injury, brain tumor and/or craniotomy or dementia, being in the acute phase of a stroke, any severe illness, any obvious cognitive disabilities, or the presence of deafness, aphasia or other language barriers.
Sample selection and procedure
Participants were identified in three stages. First, 62 primary sample units (PSU) were selected from 2,209 villages and 393 neighborhood communities using a probability proportionate to size (PPS) method [35]. Second, depending on the total number of households in the selected PSU, 60 to 210 households were systematically identified from each PSU resulting in a total of 6,890 households being selected. Third, interviewers visited sample households and used a Kish selection table [36] to identify one eligible participant from each households. A total of 6,476 participants were approached to conduct a face-to-face interview. 414 households were not found participants because no one could be located during the period of study.
Subsampling was used in most surveys to reduce respondent burden by dividing the interview into two stages. In the Stage I interview, which was administered to all respondents, information was collected on demographics and tobacco use. The World Health Organization Composite International Diagnostic Interview (WHO-CIDI) used for mental disorders assessment during this stage. A total of 5,810 participants (89.7%) completed the Stage I interview. Stage II included assessments of risk factors, services sought, religious involvement, and identification of additional disorders that were either of secondary importance or were too time consuming to assess in the full sample. A computer program was used to select participants who completed Stage I interviews to take part in the Stage II survey. The program which divided respondents into three groups based on their Part I responses. First, all respondents who (1) met lifetime criteria for at least one mental disorders assessed in Part I, or (2) met sub-threshold lifetime criteria for a mental disorders and sought treatment for it at some time in their life, were selected to complete Part II of the evaluation. Second, a probability sample was selected of 59% of respondents who did not meet criteria for membership in the first group, but gave responses in Part I indicating that they (1) ever met subthreshold criteria for Part I disorders, or (2) ever sought treatment for any emotional or substance abuse problem, or (3) ever had suicidal ideation, or (4) used psychotropic medications in the past 12 months to treat emotional problems. Third, a 25% random sample of respondents without mental disorders or emotional problems was selected to receive the Part II evaluation [37]. The present study consisted of 2,770 participants who completed the Stage II interviews.
Face-to-face computer assisted personal interviews (CAPI) [38] were carried out by lay interviewers from Ningxia Medical University. Interviewers were trained in a 7-day session by our research team. The training covered general interviewing techniques, review of the questionnaire, post-interview editing, and in- and out-of-classroom exercises. Ninety trainees passed the final test and were selected as interviewers, forty-one of them are male, and forty-nine of them are female. These interviewers then were sent into the field to administer the survey. The survey was designed as anonymous. The potential risks and benefits of the survey were described by the interviewer and the participant was asked to provide their consent by checking a box on computer screen with the response (1 = I agree to participate in the study; 5 = I do not agree to participate in the study). If the response was “I do not agree”, the CAPI program was immediately terminated automatically. The institutional review board of the Ningxia Medical University approved the study.
Dependent variables
Cigarette smoking was assessed in terms of (1) current smoking, but without a tobacco use disorders, (2) past history of smoking (but not current use), and (3) tobacco use disorders (tobacco abuse/tobacco dependent). Smoking was measured by asking, “Are you a current smoker, ex-smoker, or have you never smoked?” here we defined the smoking as “lasting at least two months when you smoked at least once per week”. ICD-10 of Tobacco Use Disorders were diagnosed using the WHO-CID, a structured diagnostic interview that is widely used to identify mental disorders in community populations [39]. A Chinese version of the CIDI was produced by translating and back-translating the English version using the standard WHO protocol. Culture adaptation and modification research have verified the validity of this version [40].
Independent variables
Religious involvements
Religious involvement was determined using measures of religious importance, attendance, and affiliation. Religious importance was measured with a single question that asked, “In general, how important are religious or spiritual beliefs in your daily life?” Responses options ranged from not at all important (1) to very important (4). Religious attendance was assessed using a single question that asked, “How often do you usually attend religious activities?” Responses ranged from never (1) to more than once a week (5). Finally, religious affiliation was determined by asking, “What is your religious preference?” Religious affiliation was coded for analysis into four categories: 1 = none, 2 = Chinese religion (i.e., Buddhist, Daoist, etc.), 3 = Western religion (i.e., Christian, Catholic, etc.), and 4 = Muslim.
Participants were divided into high and low religiosity in the following manner. High religiosity was defined as (1) attending religious activities at least 2–3 times per month and (2) indicating that religious or spiritual beliefs were very important in daily life. All other participants were placed in the low religiosity category.
Health variables
Anxiety disorders and mood disorders were assessed using the WHO-CIDI. Anxiety disorders include agoraphobia, generalized anxiety disorder, obsessive-compulsive disorder, panic disorder, social phobia, specific phobia, and neurasthenia. Mood disorders assessed were unipolar depressive disorder and bipolar disorder.
Physical health characteristics assessed were overall self-rate physical health (good vs. poor), self-rated chronic body pain (yes vs. no), type II diabetes (yes vs. no), and hypertension (yes vs. no).
Socio-demographic variables
Socio-demographic information collected included age, gender, education, marital status (married vs. unmarried), residence (rural vs. urban), ethnicity (Han vs. Hui), experience of migration from other areas of China (yes vs. no), and geographical region (developed vs. undeveloped).
Statistical analyses
Analyses were performed using the Statistical Analysis System (SAS) 8.2 software (SAS Institute Inc). Differences in socio-demographic, physical, mental, religious, and smoking characteristics between males and females were examined using the Student’s t-test for continuous variables and the chi-square statistic for categorical variables. Differences by demographic characteristics, physical and mental health, and religious group and level of involvement across the three smoking categories (current, past, smoking disorder) were examined using one-way-analysis of variance (ANOVA) for continuous variables and the chi-square statistic for categorical variables. Three separate logistic regression models were used to examine correlations between religious involvement and smoking status. The final logistic regression model was then repeated in males to compare Muslims and non-Muslims (smoking exposure rate in females was too low for this comparison). Unstandardized beta and standard errors were calculated for all models. Given the exploratory nature of these analyses, statistical significance was set at 0.05 without correction for multiple comparisons.