The Mediating Roles of Rumination and Resilience in the Association Between Borderline Personality Features and Suicidality in College Students
Article information
Abstract
Objective
This study investigates the prevalent characteristics of suicidality in college students with borderline personality (BP) features. In addition, it explores the relationship between BP features and suicidality and analyzes related psychological mechanisms.
Methods
The study encompassed 6,854 college students from 12 universities in Southern China from October to December 2020. The subjects completed a general information questionnnaire, Personality Diagnostic Questionnaire 4+, ruminative responses scale, 10 items in the Connor-Davidson resilience scale, and an assessment for suicidality.
Results
Among all participants, subjects with BP features reported higher rates of suicidal ideation and suicidal attempt than those without BP features, and the differences were statistically significant (all p<0.05). Logistic regression analysis demonstrated that resilience served as a significant protective factor against both suicidal ideation and suicidal attempt in both females and males (all p<0.05). Higher borderline personality disorder (BPD) score emerged as risk factor for suicidal ideation and suicidal attempt in both females and males (all p<0.001). Symptom rumination was a risk factor for suicidal ideation in both females and males (all p<0.001). Reflection was a risk factor for suicidal attempt in both males and females (all p<0.05). Resilience and symptom rumination mediated the relationship between BPD score and suicidal ideation in both females and males. Resilience mediated the relationship between BPD score and suicidal attempt only in males.
Conclusion
College students with BP features exhibit an elevated rate of suicide-related behaviors. The relationship between BP features and suicidality is mediated by rumination and resilience.
INTRODUCTION
According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [1], borderline personality disorder (BPD) is characterized by a pervasive pattern of instability in interpersonal relationships, identity, and affect. Individuals with BPD also tend to demonstrate marked impulsivity in regard to potentially risky behaviors (i.e., risky sexual behavior, substance abuse, gluttony) and even self-harm and suicidal behaviors [2]. Epidemiologically, BPD has a lifetime prevalence of approximately 6% and accounts for 15%–28% of psychiatric inpatients and outpatients [3], although conservative estimates suggest a general population prevalence of roughly 1% [4,5]. The disorder typically emerges in early adolescence and peaks during mid-to-late adolescence [6]. Although symptoms often diminish with age, significant social impairment still persists [7]. Critically, personality pathology exists on a continuum rather than a strict categorical divide [8]. Thus, individuals in the general population may exhibit subclinical borderline personality (BP) features: traits falling below the diagnostic threshold but still clinically significant. Evidence indicates that these features are prevalent among college students, with a meta-analysis estimating that 0.5% to 32.1% of college students exhibit such characteristics [9]. The college stage is a crucial period for individual psychological development and thus represents a vulnerability window, during which pressures from new environments and uncertainty about future prospects can destabilize personality structures. Consequently, focusing on subclinical BP features in college students is vital for early intervention, promoting mental health resilience, and mitigating long-term social dysfunction.
Globally, suicide is an important public health problem that seriously affects social and economic development. There are indications that every adult suicide death is accompanied by more than 20 adults who attempt suicide [10]. Alarmingly, suicidal ideation and attempts are both common among college students [11]. Huang et al. [12] found the suicidal ideation rate in this population to be 13%, with suicidal ideation positively correlated with impulsivity. Crucially, as previously noted, impulsivity in risk-taking behaviors is a characteristic BP feature. Furthermore, one prospective study indicated that the severity of BPD symptoms is correlated with an increased likelihood of suicidality [13]. However, there is a lack of data on suicidality rates in Chinese college students with BP features. Given the high co-occurrence of BP features and suicide vulnerability within this demographic, investigating suicidality prevalence in college students with BP features is important. More importantly, investigating whether other psychological factors influence the relation between BP features and suicidality is critical for implementing interventions targeting suicidality in college students with BP features.
Rumination and resilience, as psychological variables, have always been widely studied in the field of suicide research. Rumination is a mode of thinking that repeatedly focuses on negative emotions and the corresponding events [14]. Resilience is a person’s ability to successfully adapt to change, resist the negative effects of a stressful situation, and thereby avoid significant dysfunction [15,16]. The theory of suicide cognition posits that repeated thinking (a form of ruminative thinking) about suicide increases the availability of suicide-related thoughts, lowers the threshold of suicidal ideation, and increases the probability of suicidal behavior [17]. However, resilience has been shown to be a protective factor against suicidality [18]. Notably, higher resilience is associated with reduced suicidal ideation in college students [19]. Prior studies have demonstrated that the severity of BPD symptoms is associated with increased rumination in undergraduate college students [20,21]. Meanwhile, resilience appears to be closely linked to personality pathology [22,23]. For example, Fonagy et al. [23] proposed that BPD may be associated with a notable absence of resilience. Building on empirical evidence that higher rumination [17,20] and lower resilience [18,23] are inversely correlated with both suicidal risk and BPD symptom severity, one can infer that rumination and resilience are influenced by BP features and also affect suicidality. Therefore, we hypothesize that rumination and resilience are mediators in the relationship between BP features and suicidality in college students.
Several studies have suggested that BPD symptoms manifest differently by sex [24,25]. For example, Hoertel et al. [24] found substantial sex-related differences in the manifestations of BPD symptoms. Sher et al. [25] found that, compared to females, male BPD patients have higher impulsivity and aggression. Moreover, the rate of suicidal attempt is significantly higher in female BPD patients than in males [26]. Sher et al. [25] also found that male BPD patients who are suicidal attempters have a higher mortality rate than females. In addition, sex differences have been observed in rumination [27] and resilience [28]. Therefore, this study further conducted analyses among different sex subgroups.
In summary, this study aims to: 1) investigate the prevalence of suicidality in college students with BP features and 2) analyze the influencing factors and indirect pathways linking BP features to suicidality in different sex subgroups of college students.
METHODS
Participants
Participants for this study were recruited from 12 universities in three provinces of South China (Hunan, Guangxi, and Jiangsu) from October to December 2020 using the convenience sampling method. The questionnaire was distributed using an online platform. All participants signed informed consent forms online prior to filling out the questionnaire. The participants were free to choose whether to participate in the study or not, and were assured of their right to withdraw from the study at any time. Subjects were instructed to answer the questionnaire according to their own situation. The Shenzhen Kangning Hospital Ethics Committee granted approval for the study (ethical approval number: 2020-K001-01), and the study conforms to the provisions of the Declaration of Helsinki.
A total of 7,695 college students participated in this research, and 841 college students were excluded because the subjects did not pass the three “trick” questions or did not meet the inclusion criteria for being under 18 years old. For example, the place of residence question was repeated in two different places on the questionnaire, and the questionnaire was regarded as invalid if two different answers were given. Ultimately, 6,854 valid questionnaires were included in the analysis, of which 2,379 were male (34.71%) and 4,475 (65.29%) were female. The average age was 19.69±1.40 years old. Of the cohort, 3,772 (55.03%) respondents lived in the city, while 3,082 (44.97%) lived in the countryside. There were 5,922 (86.40%) subjects who reported that their parents had stable marriage status, while 932 (13.60%) did not. There were 566 (8.26%) participants who had smoking behavior and 4,078 (59.50%) who had drinking behavior during the past year.
Instruments and procedure
General information
Basic information including age, sex (female or male), and place of residence (city or countryside) was collected. The marital status of parents was assessed by the question, “What’s your parents’ current marital status? (marital harmony, divorced, separated, remarried, one or both parents have passed away, any other unstable condition).” Of these response options, “divorced,” “separated,” “remarried,” “one or both parents have passed away,” and “any other unstable condition” were regarded as unstable, while “marital harmony” was classified as stable. The family economic condition was assessed by one question, “Currently, does your family have economic difficulties?” The response choices were “very wealthy, generally wealthy, general, poor, very poor.” “Poor” and “very poor” situations were taken to indicate family economic difficulties, while other situations were taken as indicating no family economic difficulties. Behaviors associated with smoking and drinking were assessed by the questions, “Have you ever smoked during the past year?” and “Have you ever drunk any alcoholic beverages or alcohol during the past year?,” respectively. The response choices were “Yes” and “No.”
Personality Diagnostic Questionnaire 4+
The Personality Diagnostic Questionnaire 4+ (PDQ-4+) comprises 107 items, which correspond to 12 different types of personality disorders in the DSM-IV [29]. If subjects believe that an item does not conform to their own situation, then they answer “no” (coded as “0”). However, if the item does conform to their own situation, they answer “yes” (coded as “1”). The PDQ-4+ has been translated into Chinese by Yang et al. [30]. The PDQ-4+ has been shown to have high sensitivity and low specificity with regard to the diagnosis of personality disorders [31]. In the current study, BP features were assessed using the borderline subscale of the PDQ-4+, comprising nine items in total, of which eight items answered “yes” scored “1,” and the remaining one item comprises six sub-items and is scored “1” if “yes” responses are given to more than two sub-items. Total scores on the borderline subscale of the PDQ-4+ of ≥5 are taken to indicate the presence of BP features [32]. The PDQ-4+ has been proven to be a useful tool for BPD screening [33]. In the current study, the Cronbach’s α coefficient of borderline subscale was 0.828.
Ruminative Responses Scale
The Chinese version of the Ruminative Responses Scale (RRS), as translated and calibrated by Han and Yang [34], was used in this study. The RRS comprises 22 items divided into three factors, namely brooding (the passive comparison between a person’s current situation and their unsatisfied standards), reflection (a purposeful turn to internal cognitive problem solving to alleviate a person’s depressive symptoms), and symptom rumination (symptoms of repeated thinking about the current situation) [35]. Each item is rated on a four-point scale, with 1=almost never, 2=sometimes, 3=often, and 4=almost always. The higher the score, the more serious the rumination. The Cronbach’s α coefficient has been found to be 0.84 when used with a Chinese sample [36]. In the current study, the Cronbach’s α coefficient of this scale was 0.966.
Ten-item Connor-Davidson Resilience Scale
The Connor-Davidson Resilience Scale (CD-RISC) was initially developed by Connor and Davidson [37]. The scale comprises a total of 25 items, each of which is scored on a scale from 0 to 4, with 0=never, 1=very seldom, 2=sometimes, 3=often, and 4=always. The CD-RISC-10 comprises of 10 items taken from the CD-RISC. Higher total scores on these scales indicate greater resilience. The Chinese version of the CD-RISC-10 has demonstrated good reliability and validity in use with Chinese college students and depressed patients [38], and its widespread applicability for use with Chinese populations has been further supported by recent research [39]. In the current study, the Cronbach’s α coefficient of this scale was 0.951.
Assessment for suicidal ideation and suicidal attempt
Suicidal ideation was assessed by one item, “Have you ever seriously thought about committing suicide?” [40] Suicidal attempt was assessed by one item, “Have you ever attempted to kill yourself?” [41,42] The respondents were asked to answer each item “yes” or “no.”
Statistical analysis
Statistical analysis was conducted using SPSS version 21.0 (IBM Corp.). Descriptive statistics included classified variables, the prevalence of suicidal ideation and attempt presented as percentages and 95% confidence intervals (CI), and the continuous variables presented as means and standard deviations. The chi-square test was used to assess differences in rates of suicidal ideation and suicidal attempt across the demographic subgroups. Independent samples t-testing was used to compare resilience and rumination scores between groups stratified by suicidal ideation (yes vs. no) and suicidal attempt history (yes vs. no). Logistic regression models were fitted to evaluate BPD score, resilience, and rumination as predictors of suicidal ideation and suicidal attempt. Mediation analyses were performed using the PROCESS macro (Version 3.2) for SPSS, with dichotomous outcomes (suicidal ideation or suicidal attempt) specified as dependent variables [43] and BPD score specified as the independent variable. Significant independent variables from the logistic regression (excluding BPD score) were included as mediators, while non-significant variables were included as covariates. Model 4 was selected in the mediation model, bootstrapping with 5,000 resamples was used, and statistical significance was inferred when the 95% CI from the bootstrapping method did not include 0. All hypothesis tests were two-tailed, with a significance level set at p<0.05, unless otherwise specified in the mediation analysis.
RESULTS
The prevalence of suicidal ideation across demographic distributions in groups with and without BP features
Of the 2,379 male participants, 363 (15.26%) reported having BP features, and of the 4,475 female participants, 787 (17.59%) reported having BP features; the difference between male and female was statistically significant (χ2=6.029, p=0.014). Of the students with BP features, 62.35% had suicidal ideation; this was higher than the rate for students without BP features (18.00%), and the difference was statistically significant (p<0.001). Among the students with BP features, females (65.57%) reported having a higher rate of suicidal ideation than males (55.37%). Similar results were found for the non-BP features group, in that the females (19.36%) reported higher suicidal ideation than the males (15.53%). Both differences were significant (p<0.001). Across all demographic subgroups examined, participants with BP features exhibited significantly higher rates of suicidal ideation compared to their non-BP counterparts (all p<0.001). Within the BP features group, those with family financial difficulty had a higher suicidal ideation rate (68.94%) than those without family financial difficulty (59.78%), and drinkers reported a significantly higher suicidal ideation rate (64.55%) compared to non-drinkers (56.51%), with all differences reaching statistical significance (all p<0.05) (Table 1).
The prevalence of suicidal ideation in the BP features and non-BP features groups across different demographic distributions
The prevalence of suicidal attempt across demographic distributions in groups with and without BP features
The prevalence of suicidal attempt was 14.61% in college students with BP features and 3.35% in those without BP features; the difference was statistically significant (p<0.001). Across all demographic subgroups, participants with BP features exhibited significantly higher rates of suicidal attempt than their non-BP counterparts (all p<0.001). Within the BP features group, those who lived in a city had a higher suicidal attempt rate (16.56%) than those who lived in the countryside (12.05%), and smokers reported a significantly higher rate (23.75%) compared to non-smokers (13.13%). All differences reached statistical significance (all p<0.05) (Table 2).
The prevalence of suicidal attempt in the BP features and non-BP features groups across different demographic distributions
Comparison and subgroup analyses of resilience, rumination, and BPD scores between college students with and without suicidal ideation/attempt
Across both the female and male subgroups, individuals with a history of suicidal ideation/attempt reported significantly higher symptom rumination, brooding, reflection, total rumination, and higher BPD score than those without such history (all p<0.05). Conversely, participants with suicidal ideation/attempt had significantly lower resilience scores than their non-suicidal ideation attempt counterparts (all p<0.001) (Table 3).
Comparison of resilience, rumination, and BPD scores between college students with and without SI/SA, with subgroup analyses by sex (Nmale=2,379 and Nfemale=4,475)
Logistic regression analysis for predicting suicidal ideation and suicidal attempt in college students
Logistic regression analysis revealed that psychological resilience was a significant protective factor against both suicidal ideation and suicidal attempt in both females (odds ratio [OR]=0.965, 0.980; p<0.05, respectively) and males (OR=0.979, 0.964; p<0.05 for both). Conversely, higher BPD score emerged as a risk factor for suicidal ideation and suicidal attempt in both females (OR=1.377, 1.343; p<0.001, respectively) and males (OR=1.372, 1.311; p<0.001 for both). Symptom rumination was a risk factor for suicidal ideation in both females (OR=1.068, p<0.001) and males (OR=1.063, p<0.001). Reflection was a risk factor for suicidal attempt in both females (OR=1.081, p=0.047) and males (OR=1.136, p=0.024) (Table 4).
Logistic regression analysis results for suicidal ideation and suicidal attempt in college students, with subgroup analyses by sex (Nmale=2,379 and Nfemale=4,475)
Rumination and resilience mediated the relationship between BPD score and suicidal ideation/attempt in college students
Based on the results of the regression model, two path analysis models were established to explore the mediating role of rumination and resilience in the link between BPD score and suicidal ideation/attempt in college students. In the mediation model, significant independent variables from the logistic regression (excluding BPD score) were included as mediators, while non-significant variables were included as covariates. In the suicidal ideation model, resilience and symptom rumination individually were found to mediate the relationship between BPD score and suicidal ideation in both females and males. In the suicidal attempt model, resilience and reflection were found to mediate the relationship between BPD score and suicidal attempt only in males (Table 5).
Mediation analysis results for the indirect effect paths for suicidal ideation and suicidal attempt in college students by sex (Nmale=2,379 and Nfemale=4,475)
DISCUSSION
BP features is a risk factor for suicidal ideation and suicidal attempt in college students
Our study revealed that subjects with BP features had significantly higher rates of suicidal ideation and suicidal attempt compared to subjects without BP features, and this finding holds true across diverse demographic subgroups (e.g., living in the city/countryside, with stable/unstable parents’ marriage, with/without financial difficulty, with/without smoking behavior or drinking behavior). Specifically, 62.35% of participants with BP features reported suicidal ideation, a figure that exceeds the rates documented in prior studies [44,45]. For example, Knafo et al. [44] found that 59% of adolescents with BPD reported suicidal ideation, while Zhang et al. [45] reported a suicidal ideation rate of 55.2% among adults with BP features. Conversely, with regard to suicidal attempts, 14.61% of college students with BP features reported that they had attempted suicide, which is much lower than found in prior studies. Glenn et al. [46] found that 27.8% of adolescents with BP features had made at least one suicidal attempt during their lifetime, and Lee et al. [26] reported that 33.6% of adults with BPD had attempted suicide one or more times. Another study reported that 66.4% of youth with BPD had attempted suicide over the past 12 months [47]. These discrepancies likely stem from inter-study variations in sample age, clinical vs. non-clinical recruitment (e.g., our community-derived college sample vs. clinic-referred populations). Moreover, the measurement method used to assess suicide-related behaviors can also contribute to such differences. Here, suicide-related behaviors were assessed using question-and-answer items. However, Knafo et al. [44] used the Columbia-Suicide Severity Rating Scale to measure the severity of suicide-related behaviors, and a study by Glenn et al. [46] used the Suicidal Behaviors Questionnaire Revised.
Within the BP features group, females were found to have more suicidal ideation than males. Prior studies have reported higher rates of suicidal ideation for females among Chinese college students [12,48], which is consistent with the results for our non-BP features group. Moreover, in this study, females with BP features exhibited higher rates of suicidal ideation than those without BP features. These results indicate that the pattern of higher suicidal ideation rates among females relative to males persists irrespective of BP features, and that BP features are associated with further elevation of suicidal ideation levels in females. Notably, no sex differences in suicide attempt rates emerged in either the BP or non-BP groups. This contradicts the results of Lee et al. [26], who found that, among suicide attempters with BPD, there were significantly more females than males. The reason for this discrepancy is likely to be a difference in sample sources, as our study was conducted on a large sample of college students with BP features as determined by the PDQ-4+, not a small sample size of clinical patients diagnosed with BPD. Collectively, this study represents an early exploration of the demographic correlates of suicide-related behaviors in college students with subclinical BP features, highlighting the need for targeted prevention strategies in high-risk subgroups identified through population-level screening.
In the current study, logistic regression analysis demonstrated that higher BPD scores were associated with increased risk of suicidal ideation and suicidal attempt for both female and male participants. Prior studies have similarly reported an increased risk of suicide-related events among people with BPD [49,50]. This elevated risk may stem from core BPD pathology, particularly impulsivity, which amplifies vulnerability to suicidality. Supporting this, Greenfield et al. [51] reported that impulsivity specifically predicted suicidality in individuals with BPD. This finding provides a theoretical direction for elucidating the mechanism underlying the association between BP features and suicide-related behaviors. Building on this framework, the present study explores this mechanism from alternative theoretical perspectives.
The role of rumination in the relationship linking BP features to suicide-related behaviors among college students
In the present study, mediation analysis revealed that symptom rumination mediated the relationship between BP features and suicidal ideation in both females and males. Building on the logistic regression results, which confirmed that symptom rumination and BP features are risk factors for suicidal ideation, and combining them with established evidence that BPD symptom severity exacerbates rumination in college populations [20,21], we propose a mechanistic pathway by which BP traits elevate suicidal ideation in college students by potentiating symptom rumination. Therefore, when making interventions to reduce suicidal ideation in populations with BP features, one may consider the effects of symptom rumination. For example, interventions that aim to reduce symptom rumination may help reduce suicidal ideation in both females and males with BP features. Although reflection mediates the relationship between BPD score and suicidal attempt in male college students, its statistical difference is at the borderline, which may be influenced by sampling bias or other biases. Future research needs to further verify this finding.
Resilience mediated the relationship linking BP features to suicide-related behaviors in college students
In this study, we identified resilience as a significant mediator, with sex differences, in the relationship linking BP features to suicidality: resilience universally mediated the link between BP features and suicidal ideation in both females and males, but its mediation of the link between BP features and suicidal attempt was exclusively observed in males. Specifically, taking the results from our logistic regression analyses, which confirmed that resilience is a protective factor against suicidality for both males and females, in agreement with the existing literature [52,53] and combining them with established evidence for resilience deficits in BPD populations [23], we posit that resilience deficits may explain the elevated suicide risk in this population. In addition, the sex-specific mediating effects of resilience are partially consistent with the male-focused resilience buffering reported by Özbay and Bülbül [54]. The consistency of this sex-asymmetric pattern, observed in both university students with BP features and mental health workers facing chronic trauma [54], highlights the robustness of sex as a moderator of resilience pathways, in that males may more readily convert resilience resources into behavioral inhibition, particularly under high-adversity conditions [55]. These findings suggest that resilience-building interventions, particularly those enhancing social-emotional resources, could disrupt the trajectory from BP features to suicidality in collegiate populations, with male-specific protocols warranted to target suicidal attempt risk through resilience fortification.
Limitations
This study has several limitations that point to directions for future research. First, the present study was based on a cross-sectional design, and thus, it cannot establish any causal relationships between BP features and suicidal ideation/attempt. Second, the subjects in the study were college students and were thus in a special stage of psychological development. Our findings may, therefore, not be applicable to the whole population. Moreover, the convenience sampling strategy used at universities in three southern Chinese provinces may introduce sampling bias, limiting the generalizability of our results to the broader, national college student population. Therefore, more scientifically rigorous sampling methods are required to further validate the above findings. Third, the measurement of BP features was based on a screening questionnaire without clinical assessment, and thus, this result cannot be generalized to clinical practice. Fourth, although our findings revealed sex subgroup-specific mediation patterns of rumination and resilience in the relationship between BP features and suicidality in college students, the sociocultural determinants and neurobiological mechanisms underlying these differential pathways remain unexamined. Future research should explicitly examine these factors. Fifth, while some findings in this study reached statistical significance, their significance was at a marginal level (e.g., for the mediation path of BPD score → Reflection → Suicide attempt; 95% CI, -0.0099 to -0.0001). These results might be influenced by sampling bias and require further validation in future studies. Finally, suicidal ideation and attempt were assessed using question-and-answer items, which may have introduced some response bias, given that suicide-related behaviors are a sensitive topic for many people. Further structured interviews are needed to assess the risk of suicide-related behaviors.
Conclusions
College students with BP features exhibit elevated rates of suicidal ideation and suicidal attempt. The current study extends the existing evidence by demonstrating that rumination and resilience mediate the association linking BP features to suicide-related behaviors. These findings not only reinforce the need for targeted screening of BP features in college populations but also suggest that interventions designed to reduce rumination and enhance resilience may be promising strategies for mitigating suicide risk in individuals with subclinical BP features. It should be noted, however, that these conclusions require replication through more rigorous study designs, and specifically studies with reduced sampling bias, to validate the findings.
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are not publicly available due to privacy and ethical restrictions protecting patient confidentiality but are available from the corresponding author upon reasonable request.
Conflicts of Interest
The authors have no potential conflicts of interest to disclose.
Author Contributions
Conceptualization: Jianbo Liu, Xuerong Luo. Data curation: Yuqiong He. Formal analysis: Yuqiong He, Shuxian Wu. Funding acquisition: Jianbo Liu, Xuerong Luo. Investigation: Shuxian Wu. Methodology: Tingyu Yang. Project administration: Lin Chen. Resources: Jianbo Liu, Xuerong Luo. Supervision: Jianbo Liu, Xuerong Luo. Validation: Xilong Cui. Visualization: Yaru Zhang. Writing—original draft: Yuqiong He, Shuxian Wu. Writing— review & editing: Jianbo Liu, Xuerong Luo.
Funding Statement
This study was supported by the National Key Research and Development Program of China (No.2017YFC1309904), Guangdong Basic and Applied Basic Research Foundation (2019A1515110047), Medical Scientific Research Foundation of Guangdong Province of China (A2019155).
Acknowledgments
The authors would like to acknowledge all respondents who participated in the study and the research assistants who had contributed to the recruitment process and data collection.