Contextual Inuences on Deliberate Self-harm Among Chinese University Undergraduates: A Representative Nationwide Study

Background: Numerous studies have identied factors associated with deliberate self-harm (DSH), but environmental inuences have been largely neglected. This study explored regional and university contextual factors impacting DSH among undergraduate students in China. Methods: In this observational cross-sectional study, subjects were 5,016 undergraduate students, who were identied through multistage survey sampling in 22 Chinese universities. Individual-level data were obtained through a self-administered questionnaire, and environmental variables were extracted from the National Bureau of Statistics database. Multilevel logistic regression models were used to examine regional correlates of DSH. Results: The overall prevalence of self-reported DSH in the study sample was 7.5% (95% CI: 4.1%, 10.9%). The full multilevel logistic model showed university rank and city size were inversely associated with DSH prevalence (OR: 0.24 and 0.55). Regional unemployment rates were positively associated with DSH prevalence (OR: 1.98). Conclusion: Results yield important insights on the role of context in DSH among Chinese undergraduates, and indicate the need to consider environmental factors in order to ameliorate the problem.


Background
Deliberate self-harm (DSH) is de ned as the deliberate, direct destruction or alteration of body tissue without conscious suicidal intent and suicidal behavior [1]; alternatively described as nonsuicidal selfinjury. Research has con rmed that DSH is not only identi ed frequently among patients with psychiatric diagnoses, but also occurs in nonclinical populations, especially adolescents and young adults [2,3].In addition to causing physical pain, DSH impacts affect, interpersonal relationships, and academic performance [4]. DSH has emerged as a major social and public health issue among young adults. Many studies have reported characteristics associated with DSH, including sociodemographics, behavioral and mental problems, and social support [5][6][7]. These studies, however, mainly focused on individual-level factors. Although some studies discerned environmental variation in the prevalence of DSH, they were con ned to a very small number of relevant predictors [8][9][10]. No studies have examined how both individual and environmental factors contribute to DSH variation. The current study examined the independent in uence of environmental factors on DSH, especially regional factors. More speci cally, it identi ed variables related to DSH among Chinese undergraduate students, and assessed the extent to which DSH was affected by contextual factors independent of personal characteristics. According to the 'breeder hypothesis,' exposure to environmental factors and physical, social, and behavioral factors may directly or indirectly impact personal health [11,12].

Hypotheses
Several studies associated individual socioeconomic status with DSH [13,14],but no previous research has examined the association between regional socioeconomic status and DSH. In this study, we evaluated the hypothesis that both individual-level and regional socioeconomic status were associated with DSH among Chinese undergraduates. We also explored the independent in uence of city size and university rank upon DSH among this group. Large cities generally have richer social resources and better technology and public facilities than do smaller ones and more geographically remote cities or smaller communities in China [15]. We measured city size by absolute population rather than by population density or areally. It is plausible that students residing in highly populous and generally opportunity-rich cities endure less stress than do other students, which in turn could translate into a lower prevalence of DSH. Similar to large cities, high-rank universities feature more social and nancial resources, better provision for knowledge and skills acquisition, and more employment opportunities than do low-rank universities. Students attending such institutions may therefore experience lower stress levels, and hence engage in less DSH than counterparts in other universities. We hypothesized that both city size and university rank were positively associated with DSH.
Unemployment was another contextual variable examined in this study. Many studies have reported an association between personal psychological stress and DSH [5,16], and a study by Morgan and colleagues found personal unemployment was also associated [17]. However, these studies excluded regional factors. An unemployment designation requires not only that a person is without work but is also actively seeking it [18]. High unemployment re ects discontinuities in economic growth and structure, a con uence of forces that may stress residents. One study found an association between regional unemployment and violence among women [19]. In concert, the aforementioned ndings indicate that high unemployment may be a determinant of DSH.

Procedure, subjects and data source
This was an observational cross-sectional, multilevel study with a multi-staged sampling design. Subjects comprised 5,016 undergraduate students, who came from 22 universities across China. This study reports individual data from the Global Health Professions Student Survey (GHPSS) on Tobacco Control in China GHPSS (Extended version). Compared to the original survey, the extended version included extra health, mental stress, and behavioral items. A detailed description of study methods is reported by Yang et al [20].The individual-level variables were augmented by regional variables, which were extracted from the National Bureau of Statistics database [15,21].

Dependent variable
Deliberate self-harm (DSH): Survey respondents were asked "In the past year have you ever intentionally harmed yourself without suicidal intent and behavior?" A rmative answers were classi ed as DSH [22][23][24][25].

Individual-level independent variables
All respondents provided the following sociodemographic information: age, sex, race/ethnicity, grade, major, parental occupation, annual household income, and family residence. Mental stress was measured using the Chinese version of the Perceived Stress Scale (CPSS) [25,26].This scale comprises 14 items that address perceptions of stress during the month prior to the survey. Items were rated on a 5-point Likert-type scale, and ranged from 0 (never) to 4 (very often). We summed item scores to yield a total stress score, with higher scores indicating higher perceived levels of stress. Severe stress was operationalized as a score greater than 25, following prior practice [25,26].

Contextual independent variables
University rank was distinguished using the China University Ranking System ("high," "medium," and "low"), as established by the National Ministry of Education [27]. Differential levels implied variable resources, opportunities, and learning environments. There were several independent variables that re ected potential regional variation. The rst was degree of economic development, as measured by Gross Domestic Product (GDP) per capita in Yuan. Also included in our analyses were GDP per capita in the province of origin (home province GDP), and GDP per capita in the city where the students resided (university city GDP). Additional contextual variables were the population of the city in which the university was located, and the unemployment rate expressed as the number of unemployed persons per one million population. We extracted the preceding data from the National Bureau of Statistics [16,22].

Data Analysis
All data were entered into a database using Microsoft Excel. The dataset was then imported into SAS (9.3 version) for the statistical analyses. Our descriptive outcome measure was prevalence of DSH. A logistic model was utilized to assess associations between the independent variables and the outcome, selfreported self-harm versus no harm. Both unadjusted and adjusted multivariable analyses were presented. SAS survey logistic procedures were applied in the unadjusted analysis. Associations were con rmed through application of a multilevel logistic regression model using the SAS GLIMMIX procedure [27]. All analyses accounted for a within-clustering correlation employing the university as the clustering unit. Series models were built for each primary predictor, with adjustment for potentially confounding sociodemographics. We started with the Null Model, a three-level (individual, university, and city) model with random intercepts, in building the DSH multilevel logistic regression model. To the null base, we added individual, university, regional variables as xed main effects to form the "individual model", "university model," and "region model" for evaluating environmental impact on DSH. For this analysis, we operationalized the variable "DSH" as a binary response (no = 1, yes = 2). The rst category enumerated for each variable served as the referent in the logistic regression analysis. We assessed the signi cance of the random parameter variance estimates using the Wald joint t test statistic [28].
All statistical analyses were weighted. Weights comprised (1) sampling weights, as the inverse of the probability of selection, calculated at the university level, and (2) post-strati cation weights, calculated relative to sex, based on estimated distributions from a national survey [29]. The nal overall weights were computed as the product of the preceding weights [28]. We did not consider using a non-response weight since non-response rates were very low. Unadjusted logistic regression analyses were weighted using the overall participant-level weights, and multilevel analysis was weighted using sampling in regional level, subject-level weights were used post-strati cation weights, respectively [28].

Results
Valid questionnaires were completed by 97.7% of targeted students, resulting in a sample of 5,016 subjects representing 22 universities. Of the sample, 12% were younger than 20 years of age, 38% were 20 or 21, and the remainder 22 years or older. Forty-one percent were male. The majority of subjects (68%) were freshmen and sophomores, and 32% juniors or seniors. Ninety-one percent were Han Chinese (Table 1). ). The unadjusted logistic analysis showed grade level, mental stress, smoking, university rank, city size, and unemployment rate were associated with self-reported DSH. The full multilevel logistic model showed that individual grade level, mental stress, and smoking rates were associated with DSH. Turing to the contextual variables, undergraduates residing in cities with fewer than one million people manifested a 1.81 times higher likelihood of reporting DSH than counterparts residing in cities of four million or more. Undergraduates at low-rank universities had 4.14 times the likelihood of reporting DSH than did students in high-rank institutions. Undergraduates residing in regions with a high unemployment rate had approximately twice the self-reported prevalence of DSH as counterparts residing in areas with a low rate ( Table 2).

Discussion
Approximately 8% of university undergraduates in this study reported DSH, a lower prevalence than found in other Chinese studies [30,31] but similar to that reported in more economically developed countries [32,33]. Further research is necessary to enlighten our nding. We found DSH prevalence varied widely among cities and universities across China, variation that may partially be accounted for by environmental and other contextual variables.
Addressing a gap in the literature, we also found that several contextual variables were associated with self-reported DSH. University undergraduates, who resided in cities of less than one million, approached an almost two-fold higher likelihood of reporting DSH than counterparts in cities of four million or more. Large Chinese cities usually have more social and nancial resources than do midsize and small cities [16].Thus, as consequence, it seems plausible undergraduates in large cities have less stress, and in turn the smallest prevalence of DSH. Similarly, undergraduates studying at low-rank universities showed 4fold higher likelihood of reporting DSH than did students in high-rank institutions. Low-rank universities usually have less government nancial support, and fewer quality courses and student opportunities [27]. Higher stress experienced by students in response to these de cient living and learning environments may in turn induce more DSH [6, 17]. Neither regional socioeconomic status nor personal income was associated with DSH. However, university rank and city size showed inverse associations, pointing to the importance of social and economic resources as determinants of DSH [34,35].These resources are inequitably distributed across China. The more elite universities concentrate in large cities, particularly along the eastern coast. Our research results argue for the need to augment resources for undergraduates in lower rank universities, and contract intercity differences resources in order to reduce DSH.
Undergraduates residing in high unemployment regions approached twice the self-reported prevalence of DSH as counterparts residing in low unemployment regions. Regional unemployment rates are inversely associated with level of economic development, labor absorption capacity, and social stability. Characteristics associated with unemployment, including lack of jobs, low pay, and excess frustration and hopelessness about future employment prospects[36], likely foster a stressful environment for undergraduates. We also found high personal mental stress was associated with self-reported DSH, a rming ndings from other studies that examined individual-level determinants [6,17]. Parenthetically, we report a signi cant interaction between individual mental stress and regional employment status (Estimate: 0.2141, SE: 0.0889), World Chi-Square: 5.80, p: 0.0160).

Study limitations
Study limitations included the cross-sectional design, which precluded us from assessing causation between individual and regional variables, on the one hand, and DSH, on the other. Serving as mitigating forces, we employed a large sample and our ndings met several criteria for inferring causality, including the strength of some associations, their consistency, and plausibility of effect. Future studies need to compile longitudinal surveillance data on DSH. Our subjects were con ned to university undergraduates, thus inhibiting generalizability of results to the wider student population and to the Chinese population.

Conclusion
Contextual disparities appear to contribute to DSH among Chinese undergraduates. Study results point to the need to consider environmental factors in DSH prevention and reduction. Preventive initiatives must focus on redressing imbalances in the allocation of social and economic resources across universities and regions, but include teaching undergraduates how to reduce and manage their mental stress. The study was approved by the Ethics Committee at the Medical Center, Zhejiang University, and verbal consent was obtained from all participants prior to data collection.

Consent for publication
The authors listed the manuscript participated in the work and have agreed to publish this manuscript.
Availability of data and material