Perfectionism and Suicidal Ideation in Outpatient Depressed Adults: The Moderating Role of Self-Esteem, Loneliness and Rumination

Perfectionism is defined as the setting of extremely high standards for performance for oneself as well as for others (Hassan et al., 2014), accompanied by overly critical self-evaluation (Shafran et al., 2002). Perfectionism can be divided into two dimensions: perfectionistic strivings (PS) and perfectionistic concerns (PC; Frost et al., 1993; Stöber & Otto, 2006). PS represent setting high personal standards and is considered adaptive, because PS are related to higher levels of conscientiousness, motivation and active coping styles (Stöber & Otto, 2006). PC involve socially prescribed perfectionism. Individuals scoring high on PC believe that exceedingly high standards are imposed on them by others (Stöber & Otto, 2006; Frost et al., 1990) and that failing to meet these expectations may lead to criticism (Hewitt & Flett, 2004). This dimension has been labeled as maladaptive, because it is associated with excessive self-criticism, maladaptive coping styles and various psychological disorders, including depression and anxiety (Limburg et al., 2017; Sommerfeld & Malek, 2019). PC have been identified as a predictor of SI (Flett et al., 2014; Shanaz et al., 2018; Smith et al., 2018), even after controlling for powerful suicide predictors as depression and hopelessness (Flett, et al., 2014; Smith et al., 2017). There are several potential reasons why PC are associated with SI and suicidal acts. First, individuals high in PC hide behind a self-representation of apparent invulnerability and exhibit low levels of self-disclosure (Flett et al., 2014; Kawamura & Frost, 2004), which is known to be a risk factor for SI ( Friedlander et al., 2012). Second, there is a strong association between PC and hopelessness (Flett et al., 2022; O’Connor et al., 2007); another important psychological risk factor for SI (Qiu et al., 2017; Ribeiro et al., 2018). Finally, PC are associated with a rigid cognitive style including overgeneralization, dichotomous (Jahromi et al., 2012) and striving for the absolute (Flett et al., 2014). This style may lead individuals scoring high on PC to interpret failures as catastrophes and to perceive the world as rejecting (Stöber & Otto, 2006; Moscardini et al., 2022), making them prone to feelings of entrapment when experiencing interpersonal stressors (O’Connor & Kirtley, 2018; Smith et al., 2018). Entrapment occurs when a person feels distressed or defeated, without any prospect of escape or rescue (O’ Connor & Nock, 2014), and is central to the formation of SI and the amplification of existing suicidal thoughts (O’ Connor & Kirtley, 2018).

Of course, PC are not a sole predictor of SI. There is a broader biopsychosocial context that contributes to vulnerability for and development of SI. In this context, factors interact to either increase or decrease the vulnerability for the development of SI. Self-esteem, loneliness and rumination are frequently proposed as potential factors in theoretical frameworks of suicidality (O’Connor & Kirtley, 2018; O’Connor & Nock, 2014; O’Connor et al., 2012).

Self-esteem is regarded as a stable personality characteristic and refers to a person’s global evaluation of one’s own worth (Rosenberg, 1965). Existing research has reported strong correlations between self-esteem and SI (Bhar et al., 2008; Wild et al., 2004). Individuals with low self-esteem report more negative affective, cognitive and behavioral reactions to failure or criticism, anticipating social disapproval and rejection (Dunkley et al., 2012), which may reinforce negative self-perceptions and thoughts that life is not being worth living (Sommerfield & Malek, 2019).

The self-critical evaluations associated with perfectionism (e.g. “I have disappointed others”, “I’m not satisfied with myself”; Dunkley et al., 2006; Fearn et al., 2021) maintain the perceived gap between an individual’s ideal self and the actual self, resulting in experiences of low self-esteem. Low self-esteem, in turn, may amplify the critical self-evaluative tendencies of perfectionistic individuals and their tendency to expect disapproval or rejection when performance expectations are not met (Dunkley et al., 2012). According to Baumeister (1990) a strong desire to escape aversive self-awareness is an important motivation for suicide. Furthermore, Blankstein et al. (2007) reported that low self-esteem led to a stronger positive association between PC and SI in men, and a stronger positive association between PC and hopelessness in women. We therefore expect that PC are more strongly associated with SI in individuals who have lower levels of self-esteem.

Loneliness represents a distressing psychological experience resulting from a perceived discrepancy between an individual’s desired and actual interpersonal relationships (Peplau & Perlman, 1982). It is a primary indicator of thwarted belongingness: a low sense of social belonging (Joiner, 2008). Thwarted belongingness serves both as a vulnerability factor for the emergence of SI (Arango et al., 2016) and a stress factor that can exacerbate SI (O’ Connor & Kirtley, 2018). Existing research indicates that individuals high in PC experience more interpersonal problems and distress and tend to feel socially isolated (Harper et al., 2020; Roxborough et al., 2012; Stöber et al., 2021), as they fear other’s conditional acceptance based on achievements (Hewitt et al., 2017). Moreover, many individuals scoring high on PC seek to avoid potential public criticism and the appearance of defect (Blatt, 1995), leading to high self-concealment (Besser et al., 2010; DiBartolo et al., 2008; Kawamura & Frost, 2004), which contributes to feelings of alienation, isolation and loneliness, and may also increase the risks of SI (Flett et al., 2014). A recent meta-analysis showed that PC have a large positive relationship with social disconnection (r = 0.30; Smith et al., 2020). Moreover, loneliness moderates the association between PC and depressive symptoms (Chang et al., 2008), which can amplify existing SI (Hewitt et al., 2006). We therefore expect that perfectionistic individuals who report high levels of loneliness will report more SI.

Rumination is defined as repetitive thoughts about oneself and one’s own condition when facing a problematic situation, generating a negative mood (Nolen-Hoeksema, 1991). Treynor and colleagues (2003) identified three components of rumination. The first component is depression-related rumination, which involves the repetitive cognitions about the depressive symptoms. The second is brooding, referring to the passive comparison of one’s current situation with some unachieved standard. The third component is reflection, involving cognitive problem-solving. Rumination is known to intensify negative affect and increase the risk of engaging in maladaptive behaviors, such as self-injury, as a means to regulate intense emotions (Selby et al., 2008). A meta-analysis of Rogers and Joiner (2017) reported that global rumination (including the three components of rumination), brooding and reflection are positively associated with SI, with moderate to large effect sizes (respectively Hedge’s g = 0.74, Hedge’s g = . 63 and Hedge’s g = 0.38).

Rumination is also associated with perfectionism (DiBartolo et al., 2001; Flett et al., 2002) particularly with PC (Frost et al., 1997; Xie et al., 2019). Perfectionistic individuals are characterized by an excessive tendency toward evaluation and criticism (Di Schiena et al., 2012; Patterson et al., 2021) and by low decision-making skills (Frost & Shows, 1993). They easily engage in repetitive negative thinking, which may enhance the risk of experiencing defeat and hopelessness (Flett et al., 2014) and may increase vulnerability to SI. We therefore expect that perfectionistic individuals reporting high levels of depression-related rumination, brooding and reflection will report more SI.

In this study, we focus on the association between PC and SI, and the potential moderating role of self-esteem, loneliness and rumination (depression-related rumination, brooding and reflection) in a psychiatric outpatient sample of adults reporting depressive mood and SI. To our knowledge, there are few existing studies on the relationship between PC and SI in adults (Flett et al., 2014; Robinson et al., 2022; Smith et al., 2018; Shanaz et al., 2018). Most existing studies have focused on adolescents and in many studies a community or college samples was used (Smith et al., 2018). It is unclear whether the association between PC and SI observed in these populations is generalizable to a clinical population and warrants attention in treatment.

Based on previous research, we expected that adults with depressive symptoms reporting higher levels of PC would also report higher levels of SI. Furthermore, we expected that the relationship between PC and SI would be stronger in individuals with lower self-esteem and higher levels of loneliness, brooding, reflection and depressive-related rumination.

The study used data collected as part of a randomized controlled trial (RCT) testing the effectiveness of an Eye Movement Dual Task treatment to decrease suicidal intrusions in adults with elevated levels of depression (see study protocol Van Bentum et al., 2019). The sample consisted of 110 adults, aged 18–70 years, (M = 33.54, SD = 12.03). Characteristics and psychiatric diagnoses of the subjects can be found in Tables 1 and 2. Individuals could participate in this project when they satisfied the following inclusion criteria: (a) age ≥ 18; (b) reporting moderate to high levels of depressive symptoms (not necessary a depressive disorder) measured by a score of 20 or higher on the Beck Depression Inventory Scale (BDI-II; Beck et al., 1996) and a score of 1 or higher on the Suicidal Ideation Attribution Scale (SIDAS; Van Spijker et al,. 2014); (c) adequate proficiency in the Dutch language; (d) currently receiving treatment as usual (TAU) at a mental health institution.

Patients were excluded when they (a) were diagnosed with at least one of the following disorders: DSM-IV (American Psychiatric Association, 2000) psychotic disorder, bipolar disorder, depressive disorder with psychotic symptoms; (b) they currently received treatment for Posttraumatic Stress Disorder (PTSD) (since the RCT focused on evaluating a treatment that may resemble eye movement desensitization and reprocessing, an evidence-based PTSD treatment).

The sample size is 110. For the regression of SI on PC the power will be 0.91 according to G*Power (Faul et al., 2009). The power calculation is based on an effect size of r = 0.28 (as reported in the meta-analysis study of Smith et al. (2018) and a Type I error of 0.05.

Depressive symptoms were measured with Dutch version the Beck Depression Inventory (BDI-II-NL-RBeck et al., 1996; Van der Does, 2002). The BDI-II consists of 21 items, each comprising a list of four statements about a particular symptom of depression. Scores on the items range from 0 to 3 (e.g. “I do not feel sad’’ to “I am so sad and unhappy that I can’t stand it”. The level of depressive symptoms was calculated as the sum score across all items. A sum score of 29 or above is indicative of severe depressive symptoms. Reliability of the BDI-II has been supported (α = 0.92;Evers et al., 2005). Cronbach’s alpha was 0.85 in this study.

Perfectionistic Concerns was measured with the Dutch version of the Frost Multidimensional perfectionism scale (FMPS;Boone et al.,; 2014; Frost et al., 1990). The FMPS consists of 35 items measured on a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). This questionnaire consists of six subscales: Personal Standards, Concern over Mistakes, Organization, Doubts about Actions, Parental Expectations and Parental Criticism. For the purpose of this study we only focused on the dimension PC (13 items), which consists of the mean scores of the items of the subscales Concern over Mistakes (9 items; e.g. ‘I hate being less than the best at things’) and Doubt about Actions (4 items; e.g. “I usually have doubts about the simple everyday things I do”). Higher scores indicate higher levels of PC. The dimension PC of the Dutch version of the FMPS has demonstrated good internal consistency (α = . 85; Schoofs et al., 2010). Cronbach’s alpha was 0.90 in this study.

Self-Esteem was measured with the Dutch version of the Rosenberg Self-Esteem scale (RSES; Rosenberg, 1965). The RSES is designed to measure global self-worth and consists of 10-items (e.g. ‘On the whole I’m satisfied with myself’) measured on a 4-point Likert scale ranging from 0 (strongly disagree) to 3 (strongly agree). After reversed-coding of the negative items, self-esteem was calculated as the mean score across all items. Higher scores indicate higher levels of self-esteem. The Dutch version of the RSES has demonstrated good internal consistency (α = 0.86; Franck et al., 2008). Cronbach’s alpha was 0.86 for the present sample.

Loneliness was measured with the Dutch translation of a short version (RULS-8; Roberts et al., 1993) of the Revised UCLA Loneliness Scale (R-ULS; Russell et al., 1980). This scale consists of 8-items (e.g. “I feel left out”) measured on a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). Loneliness scores were calculated as the mean score across all items. Higher scores indicate higher levels of loneliness. RULS-8 has demonstrated good internal consistency (α = . 82; Goossens, et al., 2014). Cronbach’s alpha was 0.82 in the present sample.

Rumination was measured with the Dutch version (RRS-NL, Raes et al., 2003) of the Ruminative Response Scale. This scale measures ruminative thoughts concerning depressive and sad feelings. The 22-item version we used is the first Dutch version based on the original RRS constructed by Treynor et al. (2003). Each item contains a statement about thoughts and acts when feeling down, sad or depressed. Participants report on a 4-pointscale from 0 (almost never) to 3 (almost always) the frequency of these thoughts and acts when they feel sad or depressed. The Reflection Subscale consists of five items of the RRS (e.g. “I Analyze recent events to try to understand why I am depressed”). The internal consistency reliability of the reflection subscale of the Dutch version has been demonstrated acceptable (α = 0.75; Raes et al., 2009) and was also acceptable in this study (α = 0.73). The brooding scale consists of five items (e.g. ‘I think: “why do I always react this way?”). The internal consistency of this five-item brooding scale of the Dutch version has been demonstrated acceptable (α = 0.78; Raes & Bijttebier, 2012). However not all five brooding items are part of the widely circulated first Dutch version of the RRS we used in this study (Raes et al., 2003). Accidentally, in this version of the RRS only three of the five items of the brooding scale were included (Raes et al., 2009). In the current sample the internal consistency of the three items has demonstrated poor (α = 0.56). The other 14 items represent the depression-related rumination subscale (Treynor et al., 2003; “I think about how alone I feel”). Cronbach’s alpha for this subscale was 0.88 in the current sample. Scale scores were calculated as the mean across the items of a scale. Higher scores indicate higher levels of rumination.

Suicidal ideation was measured using with the Suicidal Ideation Attributes Scale (SIDAS; Van Spijker et al., 2014). This questionnaire is designed to screen individuals for presence of suicidal thoughts and to assess the severity of those thoughts. It consists of 5 items, each targeting an aspect of suicidal thoughts: frequency (e.g. “In the past month, how often have you had thoughts about suicide?”), controllability (e.g. “In the past month, how much control have you had over these thoughts?”), closeness to attempt (e.g. “ In the past month how close have you come to making a suicide attempt?”), level of distress associated with the thoughts (e.g. “ In the past month, to what extent have you fell tormented by thoughts about suicide?”) and impact on daily functioning (e.g.” In the past month, how much have thoughts about suicide interfered with your ability to carry out daily activities, such as work, household tasks or social activities?”.

Responses on these 5 items are measured with 10-point scales varying from 0 to 10 (e.g. frequency of suicidal varied form 0 (never) to 10 (always)). All participants in our sample scored above 0 on the first item. SI was calculated as the mean score across all items. A mean score of 4.2 or above (or a sum score of 21 or above) indicates high risk of suicidal behavior. The SIDAS has demonstrated excellent internal consistency (α = 0.91, Van Spijker et al., 2014). Cronbach’s alpha was 0.80 in this study.

First, means, standard deviations and correlations between the research variables were computed. Hierarchical moderated regression analysis was used to test moderation effects, loneliness and rumination (brooding, reflection and depression related rumination) in three steps. An analysis was conducted for each of the five moderators separately. In the first step of the analysis, the control variables gender, age and education were included as predictors of SI. In the second step PC and a moderator variable were included and in the third step the interaction term of PC x moderator. Because our first hypothesis was about the relationship between PC and SI, we first tested the effects of PC (Step 2A, Table 5) before including the moderator variable (step 2) in the moderation analysis of self-esteem. PC and the five moderator variables were centered (raw scores minus the mean of the variable) to facilitate the interpretation of the regression outcomes and to prevent multicollinearity problems. Interaction terms are the product of centered predictors. For the analyses we used IBM Statistical Package for the Social Sciences (SPSS) version 27 (IBM Corp., Armonk, NY USA).

The mean BDI-II sum-score in this sample was 37.2 (range 10–57, SD = 9.76). A score of ≥ 29 indicates severe depressive symptoms. The mean SIDAS sum-score was 29.44 (range 3–45, SD = 10.39). Most individuals (80,9%) reported a SIDAS sum-score above the cut-off of 21, indicating a high risk of suicidal behavior. Table 3 displays the descriptive statistics of the research variables..The mean score on the SIDAS (M = 5.89) is also high and lies far above the cut off score of 4.2.

Table 4 includes intercorrelations among all study measures. PC have a strong negative significant correlation with self-esteem (r = -0.64) but a positive significant correlation with loneliness (r = 0.43) and the three rumination variables (brooding r = 0.40; reflection r = 0.22; depression–related rumination r = 0.51). Self-esteem is negatively correlated with loneliness brooding and depression-related rumination, loneliness is positively related with the rumination variables and the rumination variables are significantly positively correlated with each other. SI has only a significant correlation with depression-related rumination.

4.2 Moderation Effects of Self-Esteem, Loneliness and Rumination on the Relationship Between PC and SI. Table 5 summarizes the results of the hierarchical moderated regression analyses. Two significant interactions were found: PC x self-esteem and PC x depression-related rumination. In addition, education level appeared to have a significant negative effect on SI. PC are not significantly associated with SI (see step 2a of model 1 and the non-significant correlation in Table 2).

The two significant interaction effects are visualized in Figs. 1 and 2.

The relation between PC and SI differs under condition of high or low self-esteem, see Fig. 1. Under condition of low self-esteem SI increases if PC are increasing. Under condition of high self-esteem SI decreases if PC are increasing.

This study has several strengths. First, there are few existing studies examining the relationship between PC and SI in clinical populations of adults (Smith et al., 2018). Our study focused on psychiatric outpatient adults characterized by higher levels of depressive mood and SI. Moreover, to the best of our knowledge this study is among the first to examine how self-esteem moderates the relationship between PC and SI (Blankstein et al., 2007), as well as the first to explore how loneliness and rumination moderate this relationship. By exploring these factors, this study provides valuable insights to the profiles of individuals at risk for SI.

However several limitations of this study should be acknowledged. First, the use of a cross-sectional design precludes establishing causal relationships between PC, self-esteem, rumination, loneliness and SI. Second, despite strong evidence linking hopelessness with SI and PC (Qiu et al., 2017; Ribeiro et al., 2018; Flett et al., 2022; O’Connor et al., 2007), hopelessness is not included as a control variable in this study. Third, the study did not account for the participant’s treatment phases, or subgroups of participants based on symptom severity or diagnosis. Including this information could have provided an opportunity to examine the PC-SI association in greater detail across various subgroups. Fourth, a repeatedly suggested concern with the depression-related rumination scale is its overlap with items assessing depression severity, as indicated by its high correlation with the BDI-II (Treynor et al., 2003). Post-hoc analysis revealed a moderate correlation between depression related rumination and depressive symptoms (BDI-II) in our study (r = 0.44, p = 0.00), suggesting that depression-related rumination may be partly redundant with the depressive symptoms in terms of predicting SI. Fourth, we employed an outdated version of the RRS in our study, that included only three of the five brooding items (Raes et al., 2009), resulting in poor reliability of the brooding items in our study. Finally, we used a sample with individuals reporting high levels of both SI and PC. The mean score on PC (M = 3,36) in this sample was much higher than the mean scores in other studies with clinical samples examining the relationship between PC and SI with the F-MPS (M = 1,8—2,6; Portzky et al., 2014; Adkins & Parker, 1996). Therefore our results may have limited generalizability to the general psychiatric population.

Future research should replicate and extend these findings using longitudinal designs across non-clinical, subclinical, and clinical populations to elucidate and establish causal relationships between PC and SI. Additionally, exploring the moderating effects of other well-established predictors of SI such as coping, social support and resilience, in the relationship between PC and SI, could further identify and clarify the risk factors for SI development in depressed adults and inform prevention and treatment strategies.

First, our findings suggest that therapists should screen for SI in depressive patients who report high levels of PC, low-levels of self-esteem, and high levels of rumination. Second, this study identifies as a specific psychological profile for individuals high in PC and SI. Including self-esteem and rumination in the initial case conceptualization at the start of the treatment may be beneficial allowing treatment to be tailored accordingly.

Our findings indicate that techniques aimed at reducing PC may benefit depressed adults, reporting SI and low self-esteem or high levels of depressive rumination. Cognitive restructuring techniques, that broaden patients’ self-evaluation schemes and enhance self-compassion have been effective in reducing perfectionistic cognitions and associated emotions (Patterson et al., 2021). Furthermore, integrating depression-related rumination (repetitive thinking about depressive symptoms) into treatment could be advantageous. Rumination-focused behavior therapies such as metacognitive therapy and mindfulness based cognitive therapy have shown promise in altering depressive rumination patterns and reducing SI in depressed adults (Normann & Marina, 2018; Watkins, 2016; Schmelefske et al., 2020; Jelinek et al., 2021; Hashemi, 2020). Lastly, interventions aimed at enhancing self-esteem could be beneficial for perfectionistic individuals. For instance, COMET, an intervention utilizing counterconditioning, has demonstrated effectiveness in addressing self-esteem issues and reducing rumination (Korrelboom et al., 2012).

In sum, our study highlights that depressive adults reporting high levels of PC, coupled with low self-esteem or high levels of depression-related rumination are more likely to report SI. Clinicians should be vigilant for this specific profile in the treatment of depressive adults and interventions targeting PC, self-esteem or depression-related rumination may be instrumental in reducing SI. Replicating and extending these findings using longitudinal designs could further clarify the causal relationship between PC and SI.