Prevention of suicide in youth has been a major health focus in the 1990s (1). Recent attention has been directed toward cognitive factors related to suicidal behavior in children and adolescents. Cognitive variables such as attributional style, hopelessness, and self-esteem appear to be pertinent to understanding suicidality in youth.
There has been more investigation of these cognitive variables in adolescents than in children. With regard to adolescents, attributional style (how a person explains the causality of events) has shown conflicting findings with regard to suicidal behavior. Two studies (2, 3) found no relationship between attributional style and adolescent suicide attempts, whereas Keinhorst et al. (4) found that adolescents who attempted suicide had a cognitive style that promoted a negative evaluation of events. Hopelessness, defined as negative expectancies about oneself and one’s future, has been shown to be correlated with suicidal ideation and suicide attempts in adolescents in some studies (4–8), but not in a study by Spirito et al. (2). Low self-esteem has been associated with suicidal ideation in adolescents (9), even after the addition of control for hopelessness (10).
To our knowledge, no studies have examined differences in attributional style between children with and without suicidal ideation. Hopelessness has been shown to be correlated with suicidal ideation (11, 12), and suicidal intent has been found to be more highly correlated with hopelessness than depression in children (13). One study (13) investigated self-esteem in children with suicidal ideation and failed to find a difference between children with and without suicidal ideation.
Given the disparity of these findings in adolescents with suicidal ideation and the paucity of investigation in children with suicidal ideation, further study of these cognitive factors is warranted. The present study investigated the cognitive factors of attributional style, hopelessness, and self-esteem in psychiatrically hospitalized children and adolescents. Moreover, this study expanded on prior research by addressing whether suicidal ideation fades as a function of improvement on cognitive indices. That is, does attributional style become more optimistic as youth with suicidal ideation become nonsuicidal, their hopelessness lessens, and their self-esteem improves? This issue of cognitive shifts as a means of understanding the resolution of suicidal ideation has not been addressed in children or adolescents, to our knowledge. Since these cognitive variables have been found to be associated with depression in children (11, 13) and adolescents (9, 10), the extent to which these variables are influenced by depression was also investigated. Finally, children and adolescents were asked an open-ended question at hospital discharge about their understanding of why they no longer had suicidal thoughts.
A total of 100 subjects were selected from patients admitted to the child and adolescent inpatient units at the University of Texas Medical Branch in Galveston. On the basis of consecutive admissions and willingness to participate, 50 subjects with and 50 subjects without suicidal ideation were selected at admission. Informed consent was obtained from a parent, and assent was obtained from the minor for participation in the study. Subject diagnoses were made by means of the DSM-IV criteria on the basis of direct interviews of the subject and a parent by means of the Schedule for Affective Disorders and Schizophrenia (K-SADS) for School-Aged Children (6–18 years)—Lifetime (14), which assesses both current and lifetime psychiatric disorders in youth.
Suicidal ideation was assessed during clinical interview by means of the Lifetime K-SADS for Adolescents, by documentation of suicidal ideation in the patient’s admission note, and by the response to item 9 on the Children’s Depression Inventory (15). Item 9 asks whether the child has thoughts of killing himself or herself or wants to kill himself or herself. The group with suicidal ideation had documentation of suicidal ideation on the K-SADS for School-Aged Children, their admission notes, and the Children’s Depression Inventory. The comparison group had no evidence of suicidal ideation on the Lifetime K-SADS for Adolescents, their admission notes, or the Children’s Depression Inventory.
The subjects ranged in age from 7 to 17 years (mean=13.38, SD=2.67). Subjects with suicidal ideation were significantly older than the subjects without (mean=14.64 and 12.12 years, respectively) (t=5.32, df=98, p<0.01). Fifty-one percent of the subjects were male, and 49% were female. The subjects with suicidal ideation were more likely to be female (Yates’s χ2=7.84, df=1, p=0.005). The majority of the subjects were Caucasian (64%); others were African American (24%), Hispanic (10%), and Asian (1%). One subject’s ethnicity was not classified. Chi-square analyses revealed no significant differences between the subjects with and without suicidal ideation regarding ethnicity.
The principal diagnoses of the group with suicidal ideation were major depression (36%), bipolar disorder (20%), depressive disorder not otherwise specified (14%), conduct disorder (12%), substance use disorder (4%), psychotic disorder not otherwise specified (4%), organic mood disorder (4%), posttraumatic stress disorder (PTSD) (2%), acute adjustment disorder (2%), and schizoaffective disorder, bipolar type (2%). The principal diagnoses of the group without suicidal ideation included major depression (8%), bipolar disorder (32%), depressive disorder not otherwise specified (6%), conduct disorder (4%), psychotic disorder not otherwise specified (6%), PTSD (2%), acute adjustment disorder (2%), attention deficit hyperactivity disorder (ADHD) (28%), intermittent explosive disorder (4%), undifferentiated somatoform disorder (2%), obsessive-compulsive disorder (2%), and anxiety disorder not otherwise specified (2%).
The group with suicidal ideation contained a larger number of subjects with major depression than the group without suicidal ideation (36% versus 8%) (Yates’s χ2=9.85, df=1, p=0.002). The group without suicidal ideation included several subjects with ADHD (N=14), whereas the group with suicidal ideation included none (Yates’s χ2=14.04, df=1, p=0.0002). Forty-six percent of the group with suicidal ideation had previously attempted suicide, whereas 14% of the comparison group had prior suicide attempts, which was a significant difference (Yates’s χ2=10.71, df=1, p=0.001).
During the course of hospitalization, the patients received standard treatment, including individual and family psychotherapy, group treatment, and pharmacotherapy, if clinically indicated. The mean length of hospitalization for the group with suicidal ideation was 7.98 days (SD=2.92; median=7, minimum=3, maximum=24), and the mean for the comparison group was 9.38 days (SD=3.01; median=8, minimum=2, minimum=49). This difference was nonsignificant (t=1.32, df=98, n.s.).
Questionnaires and rating scales were administered to the subjects at admission to the psychiatric unit. These were readministered at discharge to determine whether there were changes in the cognitive factors of hopelessness, self-esteem, and attributional style when the subjects’ suicidal ideation had faded. Resolution of suicidal ideation was verified by a score of 0 on item 9 of the Children’s Depression Inventory, a patient interview, and no evidence of suicidal ideation on the patient’s chart. Subjects were then asked an open-ended question about suicidal thoughts.
The Children’s Attributional Style Questionnaire (16) is a 48-item, forced-choice measure of causal explanations for 24 positive and 24 negative events. Hypothetical events are presented, and children are requested to select the response that explains why the event happened to them. Sixteen questions pertain to each of the three attributional dimensions of internality, stability, and globality. A score of 1 is assigned to each internal, stable, or global response, and a score of 0 to each external, unstable, or specific response. There are six subscales on the Children’s Attributional Style Questionnaire, which yield a positive composite score (positive events: internal, stable, and global) and a negative composite score (negative events: internal, stable, and global). An overall score is determined by subtracting the negative composite score from the positive composite score. The lower the score, the more depressive the attributional style. The coefficient alphas for the composite positive score, composite negative score, and overall composite score are 0.71, 0.66, and 0.73, respectively. This is the most commonly used measure of attributional style in children; it is aimed at the first-grade reading level.
The Hopeless Scale for Children (17) is a 17-item self-report questionnaire that measures feelings of hopelessness and pessimism about the future. Items are identified as either true or untrue for the patient. Scores range from 0 to 17, with a higher score reflecting greater hopelessness and pessimism about the future. This measure of hopelessness was selected because of its reasonable psychometric properties. Satisfactory internal consistency (alpha=0.97) and test-retest reliability (r=0.52) after a 6-week interval have been reported (17).
The Rosenberg Self-Esteem Scale (18) is a 10-item self-report measure of global self-worth. Items are rated on a 4-point scale ranging from "strongly agree" to "strongly disagree." Internal consistency (alpha=0.81) and temporal stability (r=0.75) are satisfactory (19). The Rosenberg Self-Esteem Scale is a psychometrically sound measure for use with school-age children (20, 21) and is aimed at the first-grade reading level.
The Children’s Depression Inventory (15) is a 27-item self-report measure of depression designed for school-age children and adolescents. Its items include a wide range of depressive symptoms. Each item has three choices that are scored 0, 1, and 2, with the higher number indicative of greater severity. The total score ranges from 0 to 54. The child selects the sentence that best describes himself or herself for each item. The instrument is aimed at the first-grade reading level. This measure of depressive symptoms was selected because it has well-established reliability and validity (22).
To test the study’s hypothesis that suicidal ideation fades as a function of improvement in cognitive variables, we conducted three repeated measures. Analyses of covariance (ANCOVAs), with the attributional style, hopelessness, and self-esteem indices as repeated measures, suicidality (with or without suicidal ideation as a between-subjects factor), and time (before and after hospitalization as a within-subjects factor) as independent variables and change in score on the depressive symptom items of the Children’s Depression Inventory as a covariate. The depressive symptom items of the Children’s Depression Inventory were covaried to demonstrate that any effects found existed beyond a mere variation in depressive symptoms. In each case, assumptions were met for the use of ANCOVAs (e.g., within-class regression coefficients were homogenous). Of importance, we also examined age as an additional between-subjects factor and (separately) as a covariate to determine if findings differed for children versus adolescents. There were no significant differences between children and adolescents on any of the measures listed. Also, we examined gender and number of suicide attempts as covariates; the magnitude, direction, and significance of our original findings were unchanged.
Consistent with our prediction, the repeated measures ANCOVA produced a significant effect for the suicidality-by-time interaction (t1). An examination of means, displayed in t1, indicated that patients with suicidal ideation at admission possessed negative attributional styles, which, at discharge had improved considerably (for scores on the Children’s Attributional Style Questionnaire, higher scores mean more positive attributional style). We controlled for changes in depressive symptoms; accordingly, it was not only that subjects with suicidal ideation became less depressed over time. Rather, independent of the diminution of their depressive symptoms, the subjects with suicidal ideation no longer were suicidal, partly as a function of an improved attributional style.
A similar, although less pronounced, pattern of findings emerged with regard to hopelessness. The effect produced by the repeated measures ANCOVA for the suicidality-by-time interaction term approached, but did not reach, statistical significance (F=3.19, df=1, 97, p<0.08). Neither the main effect for suicidal ideation (F=1.88, df=1, 97, p=0.17) nor the main effect for time (F=0.02, df=1, 97, p=0.88) reached statistical significance. The means on t1 indicate that patients with suicidal ideation at admission scored high for hopelessness, which, at discharge had improved considerably. Again, because we controlled for changes in depressive symptoms, there was some evidence that the subjects with suicidal ideation reached readiness for discharge partly as a function of becoming more hopeful, independent of the diminution of their symptoms. We reemphasize that the effect of the suicidality-by-time interaction did not reach conventional levels for statistical significance and should be interpreted cautiously.
A different pattern of findings emerged regarding self-esteem. The effect produced by the repeated measures ANCOVA for the suicidality-by-time interaction did not reach statistical significance (F=0.82, df=1, 97, n.s.) (t1), and the main effect for suicidal ideation was not statistically significant (F=0.52, df=1, 97, p=0.47), whereas the main effect for time was (F=10.01, df=1, 97, p=0.002).
Subjects’ self-reports regarding their explanations for the resolution of their suicidal ideation were categorized into three major groups. A total of 56% (N=28) of the responses involved factors related to the self. These included statements regarding such topics as feeling worthy to live, recognizing positive qualities in oneself, wanting to have a family, not feeling at fault for bad outcomes, recognizing that positive events can happen, realizing that problems can be resolved, and obtaining better control over emotions. Twenty percent (N=10) of the responses involved factors related to others. Comments included recognizing that parents, relatives, and friends care about them, that suicide would be hurtful to parents and friends, and that family and friends need them. Fourteen percent (N=7) of the responses were related to treatment. Medication, therapy, psychiatrists, nurses, and hospital setting were viewed as important factors. The remainder of subjects did not have an explanation for their improvement.
In the group of psychiatrically hospitalized children and adolescents with suicidal ideation, we found that attributional style was significantly more positive and there was less hopelessness when suicidal ideation faded. These findings were independent of any diminution of depressive symptoms. Changes in these cognitive factors were not seen in the group without suicidal ideation. Moreover, there were no significant age differences (children versus adolescents) found in the pattern of results.
Although low self-esteem has been associated with suicidal ideation in adolescents (9), after control for depression in our study, there were no significant changes in the level of self-esteem when suicidal ideation was resolved. Since self-esteem is a depressive symptom, this finding is not particularly surprising. This result is consistent with the findings of Marciano and Kazdin (23), who reported that self-esteem did not discriminate between children with and without suicidal ideation when depression was controlled for.
The demographics of the group with suicidal ideation were similar to those reported in the literature (24). The subjects with suicidal ideation were older, had a greater preponderance of mood disorder diagnoses, and had more suicidal attempts than the group without suicidal ideation. In fact, nearly one-half of the group with suicidal ideation had previously attempted suicide. This is similar to the repeat rate of 58% of the suicide attempts reported in adolescent psychiatric inpatients (25). The risk factors identified were severity of suicidal ideation and having a family member or friend who attempted or committed suicide.
Kienhorst et al. (26) examined adolescents’ reasons for attempting suicide. The majority of the responses indicated that circumstances were viewed as unbearable or impossible or that the adolescent wanted to end his or her pain. In a similar vein, many children and adolescents in the current study identified beliefs that they could change or solve their problems or knew they had things to live for as reasons for no longer experiencing suicidal ideation.
The cognitive model of psychopathology described by Beck (27) has led to characterization of the negative thinking that typifies depressed individuals (28). A negative cognitive shift occurs in which a person disregards positive information and focuses on negative information. This results in negative beliefs and assumptions. Beck and colleagues (29) have examined these dysfunctional attitudes and their relationship to suicidal ideation in adult psychiatric outpatients. They found that although dysfunctional attitudes such as a need for approval were related to suicidal ideation, they were not as significantly related as a history of a suicide attempt and the degree of hopelessness about the future. Pinto and Whisman (30) reported that negative views of oneself and others led to negative affect and suicidal ideation in a sample of psychiatrically hospitalized adolescents.
Our study expanded on the cognitive view of suicide by specifically addressing a child’s attribution for these negative expectancies. Improvement in attributional style was shown to be a significant factor related to the resolution of suicidal ideation after control for depression. The majority of the patients viewed themselves as able to effect positive changes and outcomes in their lives.
The findings of our study may have important evaluation and treatment implications for youth with suicidal ideation. The assessment of attributional style may be useful in evaluating suicidal ideation. In a review of risk factors for suicide attempts, Rudd and Joiner (31) listed cognitive rigidity. On the basis of this study, it would be reasonable to include negative attributional style as a risk factor as well.
Since attributional style was shown to be a significant factor related to the resolution of suicidal ideation, this cognitive style could be specifically addressed in psychotherapy. Children with a negative attributional style tend to attribute bad outcomes to factors involving the self that will occur in all future situations, whereas good outcomes are viewed as externally derived and unlikely to occur in any future settings. This view of oneself as contributing causally could be targeted in cognitive behavior therapy. Brent et al. (32) have demonstrated the importance of cognitive behavior therapy in reducing depression and suicidality in adolescents. Furthermore, they found that a high level of cognitive distortion predicted poorer treatment outcome (33) and argued for increased attention to cognitions in youth with suicidal ideation.
The results of our study may be relevant to the effectiveness of treatment and outcome. Bologna et al. (34) argued for the redesign of behavioral health care in addressing the demands of managed care for effective, cost-contained care. In our study, significant improvement in attributional style occurred within a short period of hospitalization, i.e., mean=7 days. This shift in attributional style occurred in the absence of a specific targeted intervention to address cognitions. One may speculate that suicidal ideation may have faded earlier if therapy had been directed at fostering a more positive attributional style. Eddy et al. (35) addressed issues related to the measurement of change during therapy, including the assessment of multiple variables, which were evaluated at multiple times. It may be useful to assess attributional style at multiple points during hospitalization.
A limitation of our study is that it did not address whether the changes in cognition with the resolution of suicidal ideation were temporary or persistent. That is, although attributional style was significantly more positive at hospital discharge, there was no follow-up of these patients. King et al. (36) reported that baseline indices of adolescent functioning before psychiatric hospitalization were most predictive of later outcomes. They found that the course of psychopathology was difficult to alter by psychiatric hospitalization. Future research should determine the persistence of improved attributional style and whether this serves as a protective factor against suicidal ideation.
Presented in part at the 44th annual meeting of the American Academy of Child and Adolescent Psychiatry, Toronto, Oct. 14–19, 1997. Received April 15, 1999; revision received Jan. 6, 2000; accepted June 12, 2000. From the Division of Child and Adolescent Psychiatry, Department of Psychiatry and Behavioral Sciences, Medical Branch, University of Texas at Galveston; and the Department of Psychology, Florida State University, Tallahassee. Address reprint requests to Dr. Wagner, Division of Child and Adolescent Psychiatry, Medical Branch, University of Texas, 301 University Blvd., Galveston, TX 77555-0425; firstname.lastname@example.org (e-mail). Supported by an American Suicide Foundation Institutional Award.Erin Silvertooth, the recipient of a James Comer Minority Research Fellowship from the American Academy of Child and Adolescent Psychiatry, helped with questionnaire scoring and data entry.