Data source
This study reports data from the Global Health Professions Student Survey (GHPSS) on Tobacco Control in China [16]. The extended version added additional health, mental stress, and behavioral items. Respondents included 11,954 students from 50 universities in China’s mainland, differentiated by home regions. A more detailed description of the survey data and study methods is published elsewhere [16]. The research protocol was utilized across all 50 universities to assure homogeneity. This study was performed in accordance with the Declaration of Helsinki. The study was approved by the Ethics Committee at School of Medicine, Zhejiang University（2013-1-012）.
Measures
Dependent variable
Overall mental stress, life stress, and uncertainty stress
Overall mental stress (overall stress) was evaluated using the Perceived Stress Scale, Chinese version (CPSS), comprising 14 items that address perceptions of stress during the last month [ 17]. Items were rated on a 5-point Likert-type scale and ranged from 0 (never) to 4 (very often). Item scores were summed to yield a total, with higher scores indicating higher self-reported stress levels. This scale has been widely used to assess stress in China [12,18], manifest acceptable reliability and validity, and have been shown to be an appropriate indicator of mental health status [ 17,19,20]. Score>25 indicated severe stress verified by the Receiver Operating Characteristic Curve. This classification has demonstrated acceptable sensitivity and specificity [17]. The current study used a categorical variable coded dichotomously as signifying high stress = 2, and not high stress =1.
Life stress and uncertainty stress were measured respectively through standard questionnaires designed by Yang and colleagues [ 9,10]. Life stress questions, reflecting students’ daily worries, consisted of eight items including having “too much studying to do”, “no interest in major”, “poor study conditions”, “little support from others”, “frustration with romantic relationship”, “financial difficulty”, “poor relationship with family members”, and “poor health status among family members” [10 ].The uncertainty stress questionnaire included four items; current life uncertainty (“life is unstable and cannot be controlled”), social change uncertainty (“uncertain about what will happen in the future”), goals uncertainty (“uncertain about how to achieve goals”) and social values uncertainty (“cannot follow social values”). Resulting stress scores have acceptable validity, and have been used extensively in Chinese research [9]. This study also shows acceptable reliability, the Cronbach alpha coefficients of uncertainty stress and life stress being 0.74 and 0.79, respectively. All items pertaining to the perceived stress measures were rated on a five-point scale: feeling no stress (0); little stress (1); some stress (2); considerable stress (3); and excessive stress (4). For each individual a total stress score was obtained by summing the item scores; the higher the total score, the greater the perceived level of stress. Consistent with previous research, a cut-off score of 24 or more in life stress and 12 or more in uncertainty stress was classified respectively as a higher score with the binary outcome = 1 signifying high stress and 0 = not high stress [9, 10].
Independent variables
Individual variables
Sociodemographic questions including age, sex, ethnicity, parental occupation, and per capita family income were controlled for further determine the relationships between geographical variables and life and uncertainty stress.
University variables
University type was determined using the China university ranking system (high, middle and low level) as established by the National Ministry of Education [ 21]. Higher level universities receive more state resources than low-level universities and are more likely to attract non-local students from more privileged backgrounds [22]. However, given intense competition to enter elite universities, their high tuition fees, and pressures to succeed stress levels may be higher at such institutions.
Environmental variables
The first was the urban-rural location of the students. It was hypothesized that students from rural areas would be more likely to suffer from high mental stress [16]. Second, since mental stress has been shown to be higher among students from poorer provinces [16], one measure of regional economic development level was also included, per capital disposal income of households of the student's home province [23]. Comparing per capital GDP, this variable reflects the wealth of residents. Since the financial source of universities mainly comes from the family, their economic status reflects the area where the family is located. Urbanization is a regional industrialization process, and it reflects the active degree and level of social and economic development [24]. In this study we used province level proportion of urban population as the index for urbanization.
Analysis Strategy
All data were entered into a database using Microsoft Excel. The dataset was then imported into SAS (Statistical Analysis System) (9.4 version) for statistical analyses.
Though universities in this sample were located in 42 cities in China, they came widely from home provinces inclusive of all provinces in China’s mainland, the sample has good national representativeness. Home regions are where university students grew up. Before they go to university, their values and lifestyle was formed in home regions. In addition, university students depend on their families financially, and economic status of family is closely related to the economic level of the area where family is located. So the key analysis of this research is carried out at the home provincial level.
Weights for index are the basis of all analysis in this study. All analyses were weighted. Weights included: (1) sampling weights, as the inverse of the probability of selection, calculated at the university, and (2) post-stratification weights, calculated in relation to gender, based on estimated distributions of this characteristic from a national survey [25]. The final overall weights were computed as the product of the above two weights [26]. We did not consider using a non-response weight because non-response rates were low in this study.
Weighted overall stress, life stress and uncertainty stress prevalence were calculated, which is labeled at corresponding positions on China mainland to show the geographical distribution of them.
Geographical distribution variations were exanimated through application of a multilevel logistic regression model using the SAS GlIMMIX procedure [26]. Random parameters indicate whether there are significantly differences for different stresses across various original provinces.
In this base, we analyzed several individual and regional factors’ contributing to variations of different mental stresses. Though we estimated the prevalence of mental stress among university students by weighting several demographic variables, we didn’t know other variables associated with regional variations of mental stress. We examined whether there were regional variations in mental stresses, and know factors might contribute to the variations. Both unadjusted and adjusted methods were considered in the data analyses. The latter used SAS GlIMMIX procedure [26].
We applied SAS survey procedures in all analyses using university as the clustering unit to account for a within-clustering correlation attributable to the complex sample. In this study we built several models. We started with the Null Model, a two level (individual and home province) model with random intercepts, which did not include any predictors except a constant, in assessing variation of an individual experiencing mental stress. From this base we constructed three further models, overall stress, uncertainty stress and life stress models. The models thus enabled us to examine the relative impacts of home regional social and economic factors as predictors of mental stress.