Prevalence and influence factors of suicidal ideation among females and males in Northwestern urban China: a population-based epidemiological study

The data used in this study were derived from the third wave of the Chinese Urban Social Protection Survey [49‐51], a cohort study of a randomized community sample in Lanzhou City and Baiying City of Gansu Province (northwestern China), with the cooperation of the Civil Affairs Department of Gansu Province (northwestern China). The baseline survey was completed in 2005, and subsequently two follow-up rounds of data collection were conducted in April 2006 and January 2008, respectively. This study was approved by the Committee of Ethics of School of Social Development and Public Policy (SSDPP) at Beijing Normal University. Gansu Province is a typical low-income region in western China. In 2012, the per capita disposable income of urban and rural residents was 17,237 RMB and 4,495 RMB respectively, both of which ranked in the bottom of the 31 provinces [52]. Lanzhou City, the capital of Gansu Province, is home to a population of 2,063,800, with a per capita disposable income of 18,442 RMB in 2012 [53]. Baiying City, which is 60 km away from Lanzhou, has a population of 714,200, with a per capita disposable income of 18,532 RMB in 2012 [54].

In urban China, there is a four-tier administrative system, including provinces, cities, counties/districts, towns/sub-districts, and communities [23]. Based on the administrative system, a three-stage cluster sampling process was employed, which was showed in Fig. 1. In the first two stages, the “probability proportional to size” (PPS) sampling method was used to select districts and communities based on their population size, and the detailed information can be found in our published articles [49, 50, 55]. During the third stage, 100 households were set to be the equivalent of one sampling unit; the final number of households included in the sample was determined according to the number of sampling units in each community [49]. Thus, using simple random sampling, the final sample includes a range of 100 to 300 households from each selected community. A total of 4,661 households from 32 communities throughout the two cities were eventually selected. In each household, one representative, who aged 20 years and over and was familiar with the familial situation, was responsible for the household survey and received the psychological assessments. All contacted family members provided the information about the social-demographic characteristics, medical service utilization, social protection, and employment status. As a result, the surveyed households and household respondents totaled 4,661 and 13,051, respectively. Of them, 4,291 household representatives, who successfully completed the psychological assessments, were involved in this study, yielding a response rate of 92.1 %. For the remaining household representatives, 201 people refused to participate in this study, 60 people could not been reached for more than three times, and the reminding 109 questionnaires were discarded because of the bad quality.

The survey was completed via face-to-face interviews by the trained interviewers. 50 graduate students, with a major in psychology, sociology or social policy, were selected as the interviewers from Northwest Normal University (NNU), Lanzhou City. All the interviewers participated in a 7-day systematic training at NNU, which was taught by experienced researchers from the School of Social Development and Public Policy (SSDPP) of Beijing Normal University (BNU). After finishing the training, interviewers were supposed to acquire following knowledge: (1) the purpose of the survey, (2) the standard interviewing protocol and techniques, and (3) the explanations of the items in the questionnaire. Finally, 40 qualified interviewers were recruited after a face-to-face assessment.

To ensure the quality of survey, a guideline for investigation, which covered detailed information about the research objective, question definitions, and a step-by-step interviewing protocol, was sent to each interviewer. Furthermore, prior to the interview, a letter was delivered to each selected family to inform them of the purpose of the survey and their right to refuse it. Prior to the start of interview, an inform consent document was signed once the respondent agreed to join the survey. The whole interview took about 30 to 60 minutes, which depended on the family size and the understanding capacity of the respondents. The entire survey took 12 days in January 2008. In addition, a quality control group that included five experienced researchers from SSDPP were responsible for monitoring the process of interview, and checking the questionnaires at the end of each working-day. The repeated interview was conducted through the telephone when over 30 % of data were missed in a questionnaire. Eventually, 120 questionnaires were completed by the supplementing interviews.

The major questions about the variables used in this study are described in Table 1. Participants were asked the following question about general suicidal ideation: “During the past 12 months, have you ever considered suicide?” (yes/no) [23, 56]. Depression status in this study was measured using the 20-item Center for Epidemiologic Studies Depression Scale (CES-D) Chinese version [57], with a score of 21 serving as the cut-off point for moderate to severe depression status [58, 59], and the Cronbach’s Alpha of 0.87 for 20 items. The data processing methods has been explicated in our previous articles [49, 50, 55]. In brief, social-demographic characteristics included gender (male, female), age (20–34, 35–44, 45–54, 55–64, 65 plus), education (junior high school or less, high school, college or above), currently married (yes, no). BMI was calculated by weight in kilograms divided by height in meters squared (Kg/m²), and recoded as a three-category variable (<19, ≥19& < 25, ≥25). The self-rated health was an ordinal five-point variable, ranked as very good, good, fair, bad, and very bad. Responses of “very good” and “good” were collapsed into the “good” category, and responses of “poor” and “very poor” were recoded as “poor”. Additionally, the hope for the future was categorized as very hopeful or hopeful, fair, hopeless or very hopeless, and unknown. Based on the health insurance structure in urban China [60], the variable of health insurance included the following categories: Basic Employee Medical Insurance (BEMI), Government Insurance Scheme, Medical Aid, private medical insurance, other programs related to medical services, and uninsured. As the BEMI is the basic health insurance provided for urban employees in China, and covers both the outpatient and inpatient services [60], so the variable was recoded as BEMI, other insurances, and uninsured. The household annual expenditure was derived from the summation of the 16 items listed in Table 1, and was categorized into three groups: ≤10,000 RMB, >10,000 & ≤ 20,000 RMB, and >20,000 RMB. Debts, smoking status and drinking status were all dichotomized variables (yes/no); employment status was grouped into employed, retired, and unemployed.

Statistical analysis in this study involved in four steps. First, cross-tabulations were computed to estimate the prevalence of suicidal ideation in all of the respondents and among different groups. P values were obtained to test the difference between independent variables and suicidal ideation. As suicidal ideation is a dichotomized variable, logistic regression model was used in this study. Univariate logistic regressions were then performed separately for all independent variables. Next, all the procedures above were conducted in the female and male subgroups. Then all variables used in the univariate analysis entered into a multivariate logistic regression simultaneously. The adjusted odd ratios (AORs), 95 % confidence intervals (95 % CI) and P values of influence factors were obtained. Finally, two separate logistic regressions were performed to determine associations between influence factors and suicidal ideation among males and females, respectively. In addition, the listwise deletion was employed to address the missing data in all statistical analyses. The level of significance was set as 5 % throughout the study. Data analyses were performed using Stata12.0 (StataCorp LP, College Station, Texas, USA).

The self-reported 12-month prevalence of suicidal ideation was 4.29 % (3.62 % for male, and 5.04 % for female) among urban adults (≥20 years) in northwestern China. The 12-month prevalence in this study was lower than that in rural Chinese aged 16–34 years (5.2 %) [23] and that in older adults aged 50 years or older in rural China (8.8 %) [26]; however, it was higher than that in rural residents from developed Zhejiang province [24]. Theoretically, the lifetime prevalence of suicide ideation was higher than the 12-month prevalence [1]. Therefore, we can conclude that the 12-month prevalence of suicidal ideation in undeveloped urban districts may be higher than that in developed metropolises, and lower than that in undeveloped rural areas. In other words, this study confirmed the protective effects of economic development and urbanization against suicide and suicidal ideation in China [23, 61, 62].

In this study, all influence factors examined in univariate analysis were found to be significantly different in subjects with and without suicide ideation, except for age, education, drinking and smoking (Table 2). Subsequently, multivariate logistic regressions further confirmed several risk factors for suicidal ideation including being unmarried [23, 25], depression [21, 36, 37], feeling hopeless [21, 25], having debts [38, 63] and currently drinking [27, 34]. It is important to point out that there were disparities between results of univariate regression analysis and multivariate regression analysis. These disparities may be led by the difference of the statistical methods, the univariate regression analysis help us investigate whether a relationship exists between two paired variables, however, multivariate regression analysis aims to determine which variables influence the outcome among a set of variables.

It was intriguing to find that BEMI served as a protective factor against suicidal ideation when compared with subjects who were uninsured and those with other insurances. A recent study, conducted by our research team, has also found a protective function of basic health insurance against depression [49], and this study further confirmed the protective impact of basic health insurance on suicidal ideation in the general population. However, the protective role of insurance, after achievement of universal coverage, should be further examined by longitudinal study in the future.

Age was not significantly associated with suicidal ideation, even those aged 65 or older did not show higher prevalence of 12-month suicidal ideation in this study. It was analogous to the situation of finding a similar prevalence among working-age adults and older adults in Hong Kong [21, 37], but was inconsistent to the results of rural residents in Sichuan Province [23, 26]. Even though, we would like to point out that there has been no large-scale study on this topic by using standardized assessment tools among both rural and urban areas in mainland China. Therefore, the epidemiology of suicidal ideation in China needs to be explored through further researches.

Furthermore, education, employment status, and BMI were not significantly associated with suicidal ideation in multivariate regression model. Previous studies, conducted in China [23, 64] and France [65], showed that lower education levels and being unemployed were significantly associated with increased odds of reporting suicidal ideation. However, another study, conducted by Bromet, E. J. et al. [66], showed that low education were not significant risk factors for the lifetime ideation. These discrepancies may be due to the characteristics of sample, and need to be further studied. It is important to point out that the association between BMI and suicidal ideation still remains unclear. Carpenter et al. concluded that increased BMI was associated with higher suicide ideation for women, however, lower BMI was associated with suicide ideation for men [40]. The overall significant positive associations between BMI and suicidal ideation were also found by Mather et al. [41], Wagner et al. [42], and Dutton, G. R. et al. [43]. Among U.S. youth, perceived overweight was a statistically significant risk factor for suicide attempts [67]. However, Goldney, R. D. et al. [68] revealed that the assumption “increased BMI is necessarily associated with suicidal ideation” was not tenable.

The higher prevalence of suicidal ideation among females (female vsmale: 5.04 % vs 3.62 %) was consistent with previous findings in Chinese population [4, 23]. However, being female was not a significant variable in the regression model after controlling other variables (AOR = 0.83, p = 0.351). The disappearance of gender-difference in suicidal ideation is contrary to previous studies, in which females experienced a significantly greater risk of reporting suicidal ideation than males [23, 26]. A recent meta-analysis also estimated that AOR of the pooled suicidal ideation for females was 1.43 (95 % CI = 1.25-1.64) in a Chinese population [34]. The disappearance of gender-differences in this study may be partly explained by the characteristics of this sample. For example, females in our sample came from urban community, and their social-economic statuses, compared with males, were relatively better than those from rural areas. Actually, it was found that the female–male suicide ratio has decreased sharply [12], because of the improvement in females’ social-economic status in recent years [23]. However, this interpretation need to be further examined in the future.

This study revealed the different influence factors of suicidal ideation between females and males, which were less emphasized by previous studies. The only shared risk factor of suicidal ideation was depression, a well-known risk factor for suicide and suicidal ideation [34], suggests that mental illness is still a major challenge for preventing suicide in China. In this sample, debt was only associated with suicidal ideation in females but not in males. In addition, household annual expenditure of >10,000& ≤ 20,000 RMB was a protective factor against suicidal ideation in males (marginal difference, p = 0.05), but not in females. These results indicated that males benefit more from better economic status than females, while females were easier influenced by negative economic situation (i.e., having debts). Females may be more vulnerable when they experienced the life-events including being unmarried, depression, uninsured, and having debts, compared with males. This finding is inconsistent with the results of studies from western countries, which found debts were significantly correlated with higher suicidal ideation among general population. A study [69] in European and American countries found that the 2008 global economic crisis mainly influenced the suicide rate in men. Another study, conducted by Turvey C et al., revealed that financial loss rather than low income remained a significant correlate of suicidal ideation after controlling for depression [29]. Similarly, Hintikka et al. concluded that difficulty in repaying debts is an independent factor that is associated with suicidal ideation [70]. A more recent study also indicated that the population in debt were twice as likely to have suicidal ideation after controlling for the socio-demographic, economic, social and lifestyle factors [63]. Another more recent study also indicated that poorer financial perception was associated with more prevalent suicidal ideation [23]. One possible explanation for the difference may be that the traditional cultural values in China were different from western countries. Of course, this hypothesis should be further explored.

Finally, smoking and drinking status were not significant influence factors of suicidal ideation in the full model; however, smoking was a significant risk factor in females, and drinking was a significant factor in males. This difference may be explained by the much higher smoking rate in males than females in China (52.9 % vs. 2.4 %) [71]. Two studies showed that current smoking were associated with suicidal ideation [25, 27], another two studies further indicated that daily smoking was associated with suicidal ideation in women but not in men [65, 72]. However, there were no significant differences between suicidal ideation and smoking status in other studies [66, 73]. Now, China is the largest consumer of tobacco in the world, with an estimated 301 million current smokers. Therefore, the inter-relationships of smoking and suicide ideation need to be considered in the future. On the other hand, drinking was a risk factor for males in this study, which was similar with previous studies which confirmed alcohol abuse disorders are more common in males suicides [46, 74].

Several limitations of this study should be acknowledged. First, as a cross-sectional study, this study was unable to prove a causal-effect relationship between risk factors and suicidal ideation. Second, as the sample was selected from two cities of Gansu Province, a conservative generalization of the findings of this study should be applied when considering the diversity of social-economic status on other populations. Third, suicidal ideation was assessed by using a single question, which could not distinguish the severity of the suicide risk. The independent measuring tools should be utilized in epidemiological or clinical study in the future. Fourth, more valid measures, such as Beek Depression Invertory Beck (BDI-21) and Patient Health Questionnaire (PHQ-9), should be used in the future study, although the CES-D scale has been used extensively in previous studies. Fifth, social support and personality were not measured in this study. Sixth, although gender differences were observed in this study, the underlying mechanisms for those influence factors remain unclear.

Regardless of those limitations, this study with a large sample size provided several contributions to current knowledge about suicidal ideation. 1) This study may be the first study conducted within a general population in an undeveloped urban area of China. 2) Detailed risk factors of suicidal ideation were investigated among a large number of populations. 3) The gender-specific differences were justified by using statistics analyses, which will be beneficial for the design and implementation of suicidal prevention strategies.