Epidemiology of depressive disorders among youth during Gaokao to college in China: results from Hunan Normal University mental health survey

This is a cross-sectional two-stage survey of DDs among youth enrolled at Hunan Normal University (HNU). HNU is one of the leading universities and a member of “Project 211” in China. The “Project 211” is the Chinese government’s endeavor aimed at strengthening approximately 100 institutions of higher education (approximately 6%) and key disciplinary areas as a national priority for the 21st century [59]. The “Project 211” universities have been established to meet the national standards of the institutions in terms of overall quality and have adopted a unified admission policy to enroll college students who achieve good CEE scores across China. HNU ranks 89–106 among the ranked 1,211 universities in China [60]. It has 20 colleges and has an enrollment of over 40,000 students for all grades [61], which accounts for 0.24% of Chinese general undergraduates [62].The percentage of the entrants in 2017 (n = 6,922) represented 0.17% of the total entrants of undergraduates (n = 4,107,534 [62]) in China.

The participants in the survey were 6,922 freshmen coming from 31 provincial-level administrative regions across China, which included 22 provinces (excluding Taiwan), 3 autonomous regions (i.e., Inner Mongolia, Xinjiang Uyghur, and Guangxi Zhuang ethics), 2 special administrative regions (Hong Kong and Macao) and 4 municipalities directly under the Central Government (Beijing, Shanghai, Tianjin, and Chongqing). In the province distribution (Supplemental Table 1), 43.3% of the youth came from Hunan Province, 33.2% of the youth came from the other 30 provincial-level administrative regions, and 23.5% were missing this information. In the survey, all freshmen were invited to participate. To increase accessibility to the students, we sent a letter of introduction to the student administrative heads across the 20 colleges for their assistance in the fieldwork. Another letter of introduction was sent to all the students with basic information about the survey and the interviewers’ contact information. Participants were informed that the survey was totally voluntary and they had right to self-determination. All the eligible freshmen (n = 6,922) responded (100% response) and agreed to participate. As 104 participants unfinished the Beck depression inventory-II (BDI-II) (i.e., more than 3 items missing), the effective response rate for screening in the first stage was 98.5% (N = 6,818, 71.4% female, age range: 16–25 years, mean age = 18.6, standard deviation (SD) = 0.82). Among the effective respondents, 42.5% were from cities, 21.7% were from suburban areas, and 35.8% were from rural areas of China. Furthermore, 90.5% were of the Han race, and 9.5% were of other races. For the age distribution and demographic features of the sample, see Table 1. Notably, on the proportion of urban and rural populations and the Han and other races, the sample is consistent with the distributions of the national census [63].

For the two-stage survey, we used the census sample in the first screening stage, and then adopted stratified and randomly selected sampling to obtain a representative interview sample in the second stage. Specifically, all the incoming students (n = 6,922) from the 20 colleges sampled and completed the demographic information, Beck Depressive Inventory-II [64, 65] and Mental Health Screening Scale for College Students (MHSSCS) [66] in the first stage. In the second stage, we used the randomized stratification sampling based on the risk of depression. Because depression is a dimensional, not a categorical construct [67‐69], depressive symptom severity can provide an optimal way to characterize the quantitative and qualitative aspects of depressive disorders and the overall degree of dysfunction [67‐70], so as to provide an effective standard for stratification of sampling. Additionally, as mental health problems are the most common symptoms among youth with subthreshold or remitted depression [71, 72], we combined mental health problems and minimal depressive symptoms to construct the potential and low risk strata of depression. Thus, respondents were classified into five risk strata: (a) the high risk (i.e., with severe depressive symptoms, BDI-II scores 29–63); (b) the moderate risk (i.e., with moderate depressive symptoms, BDI-II scores: 20–28); (c) the mild risk (i.e., with mild depressive symptoms, BDI-II scores: 13–19); (d) the potential risk (i.e., with minimal depressive symptoms, BDI-II scores < 13, and with mental health problem, MHSSCS standard scores ≥ 2); and (e) the low risk (i.e., with minimal depressive symptoms, BDI-II scores < 13, and without mental health problems, MHSSCS standard scores < 2). To maximize the use of relatively limited clinical psychological manpower, we increased proportion of the youth at the potential to high risk among individuals administered the K-SADS in the second stage for sampling design [73, 74]. Specifically, on the basis of depression risk, 52 (74%) youth with severe depressive symptoms were randomly selected from the high risk stratum by simple random methods, 187 (74%) with moderate depression were randomly selected from the moderate risk stratum, 262 (31%) with mild depression were randomly selected from the mild risk stratum, 198 (64%) with minimal depression and mental health problems were randomly selected from the potential risk stratum, and 227 (4%) with minimal depression and without mental health problems were randomly selected from the low risk stratum. Finally, 926 (13.6%) participants were randomly selected, and all agreed to complete the face-to-face clinical interview using the Chinese version of the Kiddie Schedule for Affective Disorder and Schizophrenia for School-age Children-Present and Lifetime (K-SADS-PL) [75‐79]. A flow chart describing the stratified sampling and selection and diagnostic interview results is presented in Fig. 1.

The survey was carried out from October 10 to December 13, 2017. In the first stage, all participants were invited to complete the BDI-II in paper-pencil, and then completed the MHSSCS online in computer classrooms in each college during a specified class period. Based on the depressive symptom and mental health problem scores, 926 (13.6%) respondents from 20 colleges were stratified, randomized and invited to participate in the interview in the mental health growth room in each college. During the survey, the safety procedure (i.e., risk management) was conducted in case of adverse events, such as suicidal ideation or suicidal attempt (See Supplemental material 2). Written informed consent for participation was obtained from each participant or the participant’s legal guardian when the participant’s age was less than 18 years before the interview. Supervised by two licensed clinical psychologists, the diagnostic interview was administered by eleven clinical psychological graduates who completed a 40-hour intensive training program for the K-SADS-PL, and two weeks of clinical training, and passed both the diagnostic exams of the DSM-IV and practical tests of the K-SADS-PL. All procedures were performed in accordance with the 1964 Declaration of Helsinki and its later amendments. Ethical approval and the safety procedure were granted by the Research Ethics Committee of Hunan Normal University (protocol code 073).

To ensure a high level of consistency among interviewers, the intensive training program included a detailed course on the use of the K-SADS-PL, the DSM-IV diagnostic classification and criteria, interview skills, role play, mock interviews, patient interviews in hospital, and scoring and interview tests. The training program was closely monitored and supervised by licensed clinical psychologists.

During the survey period, clinical recalibration meetings were held regularly, and all the interviews were audio recorded to estimate interrater reliability. All the interviewers were required to submit the interview records on each working day. The data check on completeness and logicality was completed before the interviewer’s submission. Errors and other flaws were solved both at the scene and at the clinical recalibration meetings.

After the interview, the team leader randomly selected 5% of the records (n = 46) and assigned two blinded clinical psychologists to reassess them with the K-SADS-PL independently. The reliability of diagnoses for current and past mood disorders was excellent (κs = 1.00).

All analyses were conducted using procedures for complex samples. The results of participants who completed the K-SADS-PL assessment were weighted up to the census sample. The initial weight (i.e. sampling weights) for each of the five risk strata (severe, moderate, mild, potential and low risk of depression) were multiplied with the reciprocal of their sampling probabilities for the interview [73, 74]. Poststratification was used on basis of sex (male or female) using data of incoming youth enrolled at Hunan Normal University (Supplemental Table 3). Sex adjustment was used to ensure that the sex distribution of the sample matched the population of college students admitted in the 2017 academic year in China (i.e., 58.0% female) [85]. Cross-tabulation was used to estimate the prevalence, cumulative incidence, the proportional distribution of age of onset across age periods, and during the 9-month period involving the CEE and matriculation. The cumulative incidences of the new-onset DDs were estimated among youth who were at risk. Cumulative incidence and prevalence were reported as a weighted proportion (%) and associated standard error (S.E.). The standard errors were estimated with Taylor-series linearization [86] to adjust for unequal sampling fractions within each risk stratum and possible homogeneity within sampling clusters. The direct standardization method [87] was used for the sex-adjusted incidence and prevalence of DDs, distribution of age of onset, and proportion of new onset across the CEE and matriculation. The associations between DDs and sociopsychological variables were investigated with weighted univariate logistic regression analyses, followed by weighted multivariable logistic regression with variables that showed significant univariate associations. All analyses were conducted with Stata (Stata Corp. 2017, Version 15.0), and tests were performed by a two-sided test (P < 0.05). The data are provided in the supplemental materials.

Using diagnostic interviews, the 9-month cumulative incidence and prevalence estimates of DDs are presented in Table 2. The sex-adjusted 9-month cumulative incidence of new-onset DDs was 2.3% (Standard error [S.E.] 0.3%). The sex-adjusted lifetime, 6-month, and 1-month prevalence rates were 7.5% (S. E. 1.3%), 1.7% (S. E. 0.2%), and 0.7% (S.E. 0.3%), respectively. The sex-adjusted 9-month incidence of MDD was 1.4% (S.E. 0.3%), and the lifetime, 6-month, and 1-month prevalence rates of MDD were 5.0% (S.E. 1.0%), 1.0% (S.E. 0.2%), and 0.4% (S.E. 0.2%), respectively. The sex-adjusted prevalence of dysthymic disorder was 0.05 (S.E. 0.04%). The sex-adjusted 9-month incidence of DD NOS was 0.9% (S.E. 0.2%), and the lifetime, 6-month, and 1-month prevalence rates were 2.4% (S.E. 0.5%), 0.7% (S.E. 0.1%), and 0.3% (S.E. 0.1%), respectively.

Of the 122 youth with current or remitted DDs, all had available age-of-onset data, especially the specific time point of the first onset of DDs during CEE and matriculation. The median age of onset (i.e., the 50th percentile of the age of onset) of DDs was 17 years, and the interquartile range (i.e., the range of ages between the 25th and 75th percentiles) was 16–18 years.

Critically, 39.4% (48/122 cases) of youth with DDs experienced first onset during the 9-month period: 15.6% (19/122 cases) at 3 months pre-CEE, 13.1% (16/122 cases) at 3 months post-CEE, and 10.7% (13/122 cases) at 3 months post-matriculation. The sex-adjusted standardized distribution of the first-onset DDs in the 9-month transition was 36.5% (S.E. 0.6): 15.8% (S.E. 0.4) at 3 months pre-CEE, 10.8% (S.E. 0.4) at 3 months post-CEE, and 9.8% (S.E. 0.4) at 3 months post-matriculation [see Fig. 2 (A)].

Regarding the age of onset, 1.6% (2/122) of youth with DDs had their first onset in childhood (ages 10–11), 6.5% (8/122) in early adolescence (ages 12–13), 14.8% (18/122) in mid-adolescence (ages 14–15), 37.7% (46/122) in late adolescence (ages 16–17), and 39.3% (48/122) in early adulthood (ages 18–22). The sex-adjusted standardized distribution of age of onset was 1.3% (S.E. 0.1) in childhood, 5.8% (S.E. 0.3) in early adolescence, 14.4% (S.E. 0.4) in mid-adolescence, 34.0% (S.E. 0.6) in late adolescence and 38.7% (S.E. 0.6) in early adulthood [see Fig. 2 (B)].

In the multivariate analysis (Table 3), the 9-month incidence, 6-month, and lifetime DDs occurrence were all positively correlated with (a) mothers educated for more than 13 years and (b) having experienced major life events. In addition, the occurrence of 6-month DDs and 9-month incidence were both positively correlated with the female sex, whereas the 6-month and lifetime of DDs were both correlated with parental loss. Lifetime of DDs were positively correlated with parental divorce but negatively correlated with fathers educated for more than 7–9 years or 13 years.

Only 11 (9.0%, 11/122) of youth with DDs had received medication treatment at some point in the past, and greater than 90% (111/122) of the youth with any type of DDs had never received any professional help. The weighted care-seeking behaviors for any DDs was 8.7% (Table 4).

The lifetime prevalence rate for DDs in the youth sample was 7.5%, which was comparable to those reported in the recent China Mental Health Survey of community residents aged 18 years or older [90], in which the weighted lifetime prevalence of DDs was 6.8%. In addition, the lifetime prevalence rate of MDD in the sample is similar to that in previous surveys among college students in China [16, 91]. Notably, our data specifically showed that the lifetime prevalence of MDD in male students (6.4%) was significantly higher than that in female students (3.2%). In contrast, the 6-month prevalence and the 9-month cumulative incidence rates in females were significantly higher than those in males. Similar results have been found among college students in China [16, 92]. Further investigation is needed to determine the probable reasons for the higher lifetime prevalence of MDD in male students, but higher 6-month prevalence and 9-month incidence rates in female students in the sample population.

Regarding the age of onset of DDs, the study showed that the sex-adjusted standardized distribution is 1.3% of the depressed youth onset in childhood, 5.8% in early adolescence, 14.4% in mid-adolescence, 34.0% in late adolescence and 38.7% in early adulthood. Similar patterns have been found in large clinical sample investigations [50] and epidemiological surveys [57], as well as longitudinal studies and the meta-analysis on depression continuities between adolescence and young adulthood [12, 13]. The results may indicate that the period from late adolescence to early adulthood might be high-risk times for first-onset depression across cultures.

Our demographic-sociopsychological correlates of depressive disorders are consistent with those of many previous studies. These results showed that family composition and parental loss during childhood predict later depression [58, 93]. Stress during the past 12 months is the strongest predictive factor for the severity of MDD in college freshmen [94]. Moreover, our results showed that offspring of mothers with more than 13 years of education are associated with an elevated risk for DDs. This finding is in contrast to the results from Western cultures, which indicated that the offspring of mothers with less than a secondary-school level of education were twice as likely to experience a major episode of depression in early adulthood relative to those whose mothers had more education [54]. This might be a cultural characteristic of risk factors for adolescent depression [53]. As noted, parents in Mainland China have been found to be more directive in their parenting and downplay the expression of warmth (i.e., authoritarian), in line with traditional Confucian beliefs in emotional reservedness [95‐97]. The probable reason may involve cultural factors impacting the maternal patterns of parenting [98], such as high parental expectations, authoritarianism [53] and control [99], among mothers with higher education in China. The results have a similar pattern as the data of the WMH survey on the association of education years with the risk of depression in China. In Shenzhen, the least educated individuals had the lowest risk of major depressive episodes (MDEs), whereas in high-income countries, the least educated had the highest risk of MDEs [100].

Interestingly, our findings simultaneously showed that lifetime DDs were negatively correlated with fathers educated for more than 7–9 years or 13 years. This might reflect cultural factors impact on patterns of parenting of mothers and fathers in China, and their association with DDs. As a series of cross-cultural comparisons between Chinese and Western parents [101‐104] showed that Chinese mothers and fathers were more authoritarian (i.e., physical coercion, verbal hostility, and nonreasoning oriented regulation) than their Western counterparts. Somewhat surprisingly, researchers found that Chinese fathers were rated as less authoritarian than Chinese mothers [103]. In addition, on perceived parenting and risk for major depression in Chinese women[53], researchers also found that the pathogenic effect of maternal authoritarianism was stronger than that of paternal authoritarianism for risks of MDD. Furthermore, for protectiveness (i.e., an overprotective and controlling parental style), the researchers surprisingly found that paternal protectiveness was negatively associated with risk for MDD while maternal protectiveness was positively associated with risk for MDD[53]. High parental protectiveness is generally pathogenic in Western countries but protective in China, especially when received from the father [53]. This further supported cultural factors impact on patterns of parenting of mothers and fathers, and their risk for DDs in China. Replication is needed.

Unfortunately, we show that the treatment rate of depressive disorders was considerably low in youth in the sample, which suggests that the low service utilization for depression is a greater problem in young people in China. Again, these findings are consistent with the previous data from four provinces [74], the national epidemiological survey from China [90], and the WMH survey in developing countries [15, 105]. Possible reasons for this may include the stigma associated with depression, cultural response bias partly due to stoicism—a relatively high tolerance for or denial of emotional suffering of depression [106], and the scarcity of access to service due to the fact that available services are concentrated in urban-based specialty psychiatric hospitals, and most suburban and rural areas have little or no access to mental health services [74, 107, 108].

Transitioning from high school to college is an important developmental milestone that holds the potential for personal growth and behavioral change [109]. The evidence showed that negative life experiences in this transition period are uniquely associated with depressive symptom trajectories even after adjusting the effects of adolescent characteristics [110]. Specifically, negative and stressful life experiences maintain or rapidly elevate depressive symptoms during this developmental period [111]. As the period from high school to college involves a number of significant life-transition experiences, reducing negative life experiences and increasing positive life experiences is vital. However, the evidence showed that the vast majority (86.1%) of young people during 4–10 weeks before taking the CEE experienced moderate to extreme stress and 43.6% had depressive symptoms in Taiwan [19]. Numerous studies have shown that 16.7–35.4% of Chinese freshmen experience moderate to severe depressive symptoms in the first year [21, 91, 112, 113], especially at 3 months after admission [21]. Our findings also showed 16.6% (S.E. 0.5%) sex-adjusted prevalence of self-reported depressive symptoms, 11.5% (S.E. 0.4%) self-reported suicidal ideation and 5.8% (S.E. 1.3%) suicidal ideation by interview among the youth sample (Supplemental Tables 4 and Supplemental Fig. 1). Given the far-reaching consequences of adolescent and young adult depression, timely and effective intervention need be taken before the new onset.

Given that this study focused on young people who successfully passed the CEE and enrolled at HUN, the rates of new-onset depression might be higher on those who took the CEE but failed to enroll at the university. There are two aspects of this evidence. On the one hand, a series of studies demonstrated that returnees who failed the CEE had serious mental health problems compared to their peers in China [114‐117]. On the other hand, the WMH surveys, including a Chinese adolescent sample, demonstrated that college students had a significantly lower MDD prevalence than nonstudents in the same age range from 18 to 22 [15].

Given the early age of onset (median 17 years) and a high proportion of new-onset depression during the period from the CEE to matriculation, depression might have serious adverse effects on this critical developmental transition. For example, previous results from a WMH survey showed that mental disorders with pre-matriculation onset were more important than those with post-matriculation onset in predicting subsequent college attrition [15]. As depression is associated with a significantly elevated disease burden [9], post-dysfunctional harms [118, 119] and the risk of early death [10, 11], screening youth in high-risk times (ages14–22), from middle-to-late adolescence to early adulthood (Fig. 2B), might prevent the disorders and decrease harm early. Additionally, the subsequent offer of specific interventions such as evidence-based cognitive behavior treatment (CBT) [120, 121] and negative attentional bias modification training [122] to high-risk youth or those who have depressive symptoms (indicated prevention) might result in a more developmental approach to the prevention of depressive disorders before its peak onset. For those with moderate-to-severe depressive disorders, the strong evidence supported that psychotherapy (e.g., CBT) combining with pharmacotherapy (e.g., fluoxetine) seems to be the best choice [123, 124]. For the likely cost-effectiveness of expanded depression prevention and treatment from a societal perspective, an allocation of mental health resources may focus on innovative, low-threshold, inexpensive, and scalable interventions, such as computerized cognitive training [125], to prevent and treat adolescent depression [125] in large, scalable self-help procedures. This may ultimately help reduce this disorder’s large burden and alleviate its post-dysfunctional harms in youth.

Admittedly, the scarcity of access to and availability of some treatments (notably CBT) for depression in non-specialist contexts is a major concern in low-income and middle-income countries [126, 127]. Currently, the Action Plan to Develop Specialized Services for the Prevention and Treatment of Depressive Disorders [128] in China is a significant policy in the progress of mental health services. However, to implement the Plan, several major challenges need be addressed, such as lack of resources, lack of united action, and insufficient awareness of depression [129, 130]. Thus, prevention and treatment for depression, especially the improved access of services is a long-term process cross low-income and middle-income countries [131], including China [129, 130].

Although our study is the first investigation that uses the K-SADS-LP, a standardized diagnostic tool, to assess the incidence, first-onset age, distribution of depression, specific sociopsychological risk factors, and service of use for depression among Chinese youth in the pivotal developmental period, several limitations need be noted.

First, as the survey was conducted in one of the “211” universities, caution should be taken when generalizing the data, even if a previous study showed that no difference in the distribution of depressive disorders among youth across the key and ordinary universities in China [91]. Specifically, although it is an accurate estimate of the new-onset DD rates with high response (98.5%) and interview rates (100%) among the census sample of youth passing CEE to HUN, it might only represent depression distribution, its correlates, and treatment for the sample youth. Second, the incidence and age of onset were estimated through diagnostic interviews of participants’ reports, so recall bias is inevitable [132]. Even a salient marker of the CEE was used to improve the accuracy of recalling symptoms and disorder onset in critical periods. Third, using a retrospective approach, the distribution of age of onset among youth with depression might lead to definitive results [133]. In other words, there might be a potential bias on the age of onset of depression due to the restricted age of the youth sample. Fourth, although the results were adjusted for the sex ratio according to the national data, the sample population consisted of a large proportion of females. Fifth, the cross-sectional nature of the surveys makes it impossible to determine the temporal direction of associations between demographic-sociopsychological variables and depressive disorders.

Despite these limitations, the results present some valuable information on the depression distribution and onset in the pivotal epochs between late adolescence and early adulthood. It also reveals the specific high-risk period for the onset of depression and specific social-cultural risk factors for depression among youth in China. Meanwhile, the research reconfirms findings from previous epidemiological studies.