Risks for Suicidality in Major Depressive Disorder KATHLEEN MYERS , M.D., M.P.H., ELIZABETH McCAULEY, PH.D., ROSEMARY CALDERON, PH.D. , JEFFREY MITCHELL , M.D ., PATRICK BURKE, M.D. , PH.D., AN D KELLY SCHLOREDT, B.A. Abstract. This investigation developed a hierarchical multiple regression model to assess the potential risk factors for suicidality in youths 7 to 17 years old. Variables were assessed in three domains: self-perceptions, demography and diagnosis, and homelenvironment. The model controlled for major depressive disorder (MDD), which has confounded previous investigations, by evaluating potential risks in a diagnostically heterogenous sample, and then evaluating these risks in a subsample with MDD. Conduct problems and depressive thinking emerged as the most powerful predictors in both samples. Hopelessness, life stress, and maternal psychopathology predicted suicidality only in the total sample. Separation anxiety protected MDD youths. These results suggest that suicidal MDD youths may comprise a distinct subgroup of depressed youths. J, Am. Acad. Child Adolesc. Psychiatry, 1991, 30, 1:86--94. Key Words: major depressive disorder , suicidality, conduct disorder, children's depression inventory, hopelessness scale for children.

Children and adolescents in both clinical (Tishler et al. , 1981; Carlson and Cantwell , 1982; Cohen-Sandler et aI. , 1982; Garfinkel et al. , 1982; Kazdin et al., 1983; Robbins and Alessi, 1985; Pfeffer et aI., 1987) and nonclinical (Pfeffer et al., 1984; Carlson et al. , 1987; Velez and Cohen , 1988; Levy and Deykin , 1989) samples frequently report suicidal thinking , plans , or attempts, and the frequency of these behaviors may be increasing (Shaffer, 1974; Shaffer and Fisher, 1981). Many studies have identified depression as a risk factor for suicidality (Pfeffer et aI. , 1979, 1980, 1986; Carlson and Cantwell, 1982; Cohen-Sandler et al. , 1982; Myers et al. , 1985; Apter et al. , 1988; Hoberman and Garfinkel, 1988; Levy and Deykin, 1989), similar to the risk observed in depressed adults (Guze and Robins , 1970; Minkoff et aI., 1973). However , Shaffer (1982) expressed skepticism that suicidal behaviors in children and adolescents have any diagnostic specificity. His view is supported by the association of multiple diagnoses with suicidality in youths, and by Carlson and Cantwell's (1982) finding that increasing severity of depression correlates highly with severity of suicidal ideation but less so with suicide attempts . By contrast , Brent and colleagues (1988) have found a continuity between increasing suicidality and depression . Many of the risk factors identified for suicidality and for depression overlap , and depression may, therefore , confound the determination of risk factors specific to suicidality. Although suicidality in youths may not be specific to any diagnosis, Accepted February 13, 1990 . From Group Health Cooperative of Puget Sound (Dr. Myers); Children' s Hospital and Medical Center (Drs. McCauley, Calderon, Mitchell, Burke, andMs. Schloredt), and the Department ofPsychiatry and Behavioral Sciences, University of Washington School of Medicine, Seattle, WA. Dr . Mitchell is now with Lovelace Medical Center, Albuquerque, NM ; Dr. Burke is now with Children's Hospital and Western Psychiatric Institute, University of Pittsburg School of Medicine, Pittsburg, PA ; Ms. Schloredt is now with the Department of Psychology , University of Utah, Salt Lake City, UT. Reprint requests to Dr. Myers, Division of Child Psychiatry, Children's Hospital and Medical Center, 4800 Sand Point Way, NE, Seattle, WA 98105. 0890-8567/9113001-0086$02 .0010© 1991by the American Academy of Child and Adolescent Psychiatry .

86

depression is an important risk , and its relationship to suicidality remains to be elucidated . One problem with previous studies on suicidality is that risk factors have been identified in diagnostically heterogeneous groups . Increasing age has been one of the most consistently reported risks. Both the frequency and severity of suicidal behaviors correlate with increasing age (Shaffer , 1974; Shaffer and Fisher, 1981; Orbach et aI., 1987). It is unclear whether this correlation represents a developmental difference , or whether older youths experience more depression. Other identified risk factors include: lower flexibility in thinking (Orbach et al. , 1981); impulsive and aggressive behaviors (Cohen-Sandler et al. , 1982; Pfeffer et al. , 1982, 1983a, 1983b, 1984); feelings of hopelessness and worthlessness (Pfeffer et aI. , 1979; Kazdin et al., 1983; Asamow et al., 1987; Velez and Cohen, 1988); negative self-perception (Carlson and Cantwell, 1982); recent stressful life events (Cohen-Sandler et al., 1982); inflexibility in parenting style (Corder et al. , 1974); family conflict (Williams and Lyons, 1976); and parental depression and parental suicidality (Pfeffer et al., 1984, 1986; Myers et aI., 1985). However, specific risk factors for homogeneous major depressive disorder (MDD) samples have not been assessed. Diagnosis may influence which risk factors are operative for different groups of youths, just as age (Pallis et aI., 1982), ethnicity (O' Carrol, 1987), and gender (Trautman and Shaffer , 1984; Shaffer et al., 1987) determine differential risks for both completed and attempted suicide. Therefore, diagnostic homogeneit y in suicidal samples appears importantto understanding how and why certain MDD youths become suicidal. The current investigation assessed potential risks of suicidality, while controlling for MDD . Method

Subjects

The current study was part of a 3-year longitudinal evaluation of depression in children and teens. A total of 138 males and females, 7 to 17 years of age, were enrolled between 1983 and 1986. All subjects were recruited from the inpatient and outpatient services of a children's teaching l.Am.Acad. Child Adolesc .Psychiatry, 30:1 , January 1991

RISKS FOR SUICIDALITY IN MAJOR DEPRESSION

hospital affiliated with the University of Washington School of Medicine . Recruitment procedures and sample characteristics have been previously described (McCauley et aI., 1988; Mitchell et al., 1988, 1989). Subjects were enrolled in the MDD group if they had experienced an episode of MDD within the 12 months preceding the interview. A control group was recruited from those patients who did not have depressive symptoms. A total of 100 (72.5%) subjects were enrolled in the MDD group and 38 (27.5%) youths in the psychiatric control group . There were no significant differences between the MDD and control samples for age, sex, racelethnicity or socioeconomic status (SES). The sample was ethnically representative of the area from which it was drawn: 88% white, 9% black, 2% Hispanic, 0 .5% native American, and 0.5% Asian . The subjects came predominantly from middle- and upper middle-class families. Families whose Hollingshead (1975, unpublished) SES categories were I to III constituted 77.5 % of the sample. PROCEDURES AND DATA COLLECTION

Data for Children and Adolescents Youths' diagnoses. Study category was determined using the Schedule for Affective Disorders and Schizophrenia, Children's Version for 6 to 16 year-olds (Kiddie-SADS) (Chambers et aI., 1985). If the subjects met the research diagnostic criteria (Spitzer et aI. , 1978) for MDD, with or without endogenicity, within the past year, they were placed in the MDD group. Those who did not meet Kiddie-SADS criteria for MDD were classified as controls. Three of the authors (E. M., J. M., P. B.) conducted the admission Kiddie-SADS interviews. The kappa statistic for interrater reliability of the diagnosis between pairs of interviewers ranged from 0.50 (separation anxiety disorder) to 1.00. The kappa for MDD was 1.00 for both pairs of interviewers . DSM-III criteria were used to determine diagnoses other than MDD . The diagnoses for the control group included: 11 adjustment disorders; six overanxious disorders; four avoidant disorders; four attention deficit disorders; three conduct disorders; three parent-child problems; two separation anxiety disorders; two schizoid disorders; one oppositional disorder ; one panic disorder; and one gender identity disorder. Sixty (60%) MDD youths had multiple Axis I diagnoses. The most common concurrent diagnoses were anxiety disorders , particularly separation anxiety disorder (40%), and conduct disorder (15%) . Twenty-five percent of the MDD subjects had preexisting dysthymic disorder. Severity of depression . The severity of depression was determined from the Kiddie-SADS interview. The severity ratings from each of the eight criterion items were summed to give the final total depression severity score. Suicidality status. Suicidality was determined for the past year and for the MDD sample in relation to the depressive episode. This variable was extracted from the Kiddie-SADS item that assesses presence and severity of suicidality on a seven-point scale: 1 = not at all; 2 = slight , i.e ., thoughts of death but not specifically of suicide; 3 = mild, i.e., occasional thoughts of suicide but no thought of a specific l .Am.Acad. Child Adolesc. Psychiatry ,30:1 ,January1991

method; 4 = moderate, often thinks of suicide and has thought of a specific method; 5 = severe, i.e., often thinks of suicide and has thought of or has mentally rehearsed a specific plan, or has made a suicide gesture, or has heard a voice commanding suicidal behavior; 6 = extreme, i.e., has made preparations of a potentially serious attempt at suicide; 7 = very extreme, i.e., suicide attempt with a definite intent to die or potentially medically harmful. Using a cutoff of three or greater, 84 youths (60.9%) were classified as, at least, minimally suicidal. Of these, 69 (82.1 %) were MDD youths, and 15 (17 .9%) were controls. In the data analyses the sample was not dichotomized into suicidal and nonsuicidal youths; instead, the Kiddie-SADS suicidality scale was used as a continuous variable. Comorbidity. Comorbidity was determined from the Kiddie-SADS items for separation anxiety disorder and for conduct disorder. The Kiddie-SADS rating was used as a continuous variable in the data analyses. Comorbidity then refers to associated problems with separation anxiety and conduct, rather than a diagnosis of separation anxiety disorder or conduct disorder. Life-stress score. In order to assess the contribution of stressful life events to suicidality, a scale was developed incorporating both acute events and enduring circumstances. The scale expanded that described by DSM-III-R from 6 to 10 points . The additional stress rating points were as follows: mild to moderate; moderate to severe; severe to extreme; and extreme to catastrophic. Two of the investigators (K. M. and E. M.) independently determined a summary life stress score for each youth and then agreed on a consensus score . Youths' self-perception . The youths completed the following self-report questionnaires: Children's Depression Inventory (CD!) (Kovacs and Beck, 1977; Kovacs, 1980/81); Children's Hopelessness Scale (CHS) (Beck et aI., 1974; Garber , 1981; Kazdin et aI., 1986); Piers-Harris Children's Self-Concept Scale (SCS) (Piers, 1969); and the NowickiStrickland Children's Locus-of-Control (CLOC) (Nowicki and Strickland , 1973). Not all youths completed all of the inventories, particularly early in the recruitment. Ninetyfive youths completed all four instruments. Suicidality did not differ between completers and noncompleters on any of the questionnaires . Parental Data Parental diagnoses. Parents were interviewed at admission using the Schedule for Affective Disorders and Schizophrenia, Lifetime Version (SADS-L) (Endicott and Spitzer, 1978). Interviewers were blind to the diagnosis of the youth . The kappa scores for interrater reliability ranged from 0.63 (bipolar disorder) to 1.00 for most diagnoses. SADS-L data were obtained on 129 (93.5%) mothers and 68 (49.3%) fathers. Seventy (50.7 %) of the fathers were unavailable for an interview for multiple reasons. In these cases, data were then obtained from the youths' mothers regarding the following histories: MDD , bipolar disorder, schizophrenia, suicidal behavior, alcoholism, drug abuse, antisocial behavior, and psychiatric treatment. A diagnosis for a disorder was

87

MYERS ET AL. TABLE

Variable Youths' self-report inventories Children's depression inventory Hopelessness scale Piers-Harris self-esteem inventory Children's locus-of-control Demographic/diagnostic Study category Conduct problems Age Prior dysthymia Separation anxiety symptoms Sex Family/environment Life stress score Mothers' psychopathology Mothers' suicidality Fathers' psychopathology Fathers' suicidality

I. Total Sample: Bivariate Regressions B-coefficients

Beta

R2

T-Test

p-Value

0.056 0.126 -0.370

0.336 0.410 -0.359

0.113 0.168 0.129

15.354 21.041 -15.430

0.0001 0.00005 0.0002 NS

-0.959 1.243 0.087

-0.267 0.254 0.135

0.071 0.064 0.020

-10.425 9.303 2.699

0.002 0.003 0.10 NS NS NS

0.229 -0.022

0.222 -0.102

0.050 0.026

7.080 -3.678

0.01 0.06 NS NS NS

not made unless the father had been treated for it, or, in the case of antisocial personality disorder, there had been contact with legal authorities. Using this approach, the authors conservatively assigned the best possible diagnoses to the "absent" fathers. However, this approach probably underestimates paternal pathology. In order to analyze all fathers as a single group in the statistical analyses, the paternal diagnostic data on the "available" fathers were reduced to this abbreviated format. A systematic bias in the data base for "available" and "absent" fathers was assessed. Compared to those youths whose fathers participated, youths with "absent" fathers demonstrated lower SES (43.43 versus 35.81; t (1,136) = 3.47; P < 0.001), and greater lifestress (3.29 versus 4.30; t(I,136 = 3.98;p < 0.0005). The youths with "absent" fathers, therefore, appeared to have experienced more deprived and chaotic lives, often with sociopathic fathers. There were no differences between youths with "available" and "absent" fathers with respect to youths' suicidality status, diagnostic category, age, or sex. Parental suicidality status. The mothers' history of suicidal behavior was determined directly from the SADS-L interview. The suicidality rating on the SADS-L for the "available" fathers was condensed to a shortened format that was developed for the "absent" fathers. This abbreviated scale included three points: 0 = no information; I = yes, made a known suicidal act, or was known to have expressed suicidal ideation; 2 = no, never made a known suicidal act or expressed suicidal ideation. Generalized psychopathology. A global summary score of psychopathology was constructed for parents who had completed the SADS-L interview. The intent of this scale was to assess the effect of overall psychiatric illness and parental disability on the youths' suicidality. The 75-point scale included whether the parents had a psychiatric diagnosis and, if they did, the number of acute episodes, the presence of chronic psychopathology, the types of treatment 88

including medication, electroconvulsive therapy, outpatient therapy, inpatient treatment, suicidality, drug or alcohol abuse, time missed from work or from other major life responsibilities. Psychopathology scores for the fathers who completed the SADS-L interview were reduced to an abbreviated 25point format that was developed for "absent" fathers. STATISTICAL ANALYSES

The main objective of this investigation was to determine the factors that predict suicidality in a diagnostically homogeneous sample of youths with MDD. A second objective was to determine how such predictors for an MDD sample might differ from a diagnostically heterogeneous sample. Variables were assessed in three domains: the youths' selfperception; demography and diagnoses; and family/environment factors. The youths' self-perception was evaluated with the cm, SCS, CHS, and CLOC. The demographic and diagnostic domain included: age, sex, diagnostic study category (MDD versus controls), severity of depression, concurrent problems with separation anxiety or with conduct, and a prior diagnosis of dysthymia. The family/environment domain included: parental suicidality scores, parental psychopathology scores, and lifestress scores. In order to determine the best predictors, a series of hierarchicallinear regression equations were modelled. First, each potential predictor was assessed individually in a bivariate regression equation to determine its predictive power for suicidality, independent of other variables. If the proposed variable failed to predict suicidality when assessed individually, then it was excluded from additional analyses. However, if the proposed variable did demonstrate significant predictive power on its own, it was retained for additional analyses in relation to other relevant variables. Next, those significant predictors from the bivariate regressions were entered into a multiple linear regression equation with other significant variables from the same dol.Am.Acad. Child Adolesc.Psychiatry, 30:1 ,January 1991

RISKS FOR SUICIDALITY IN MAJOR DEPRESSION TABLE

2. Total Sample: Multiple Regressions for Intra- and Cross-Domain Analyses

Variable

T-Te st

R'

0.079 0.039

0.255 0.248

0.167 0.205

2.210 2.143

-0 .908 1.131

-0.254 0.231

0.0 76 0.128

-3 .135 2.85 2

0.007 0.005 NS

0.237 -0 .023

0.230 - 0.172

0.050 0.07 9

2.781 -2.082

0.006 0.04

0.063 1.028 0.037

0.2 03 0.228 0.235

0.162 0.208 0.242

1.800 2.52 4 2. 131

Age Family/environment Life stress Mothers' psychopathology Final cross-domain analyses Children 's hopelessness scale Conduct problems Children' s depression inventory Study category Life stress Mothers' psychopathology

p-Value

Beta

B-coefficients

Intra-domain analyses Youths' self-report inventories Hopelessness scale Children's depression inventory Piers-Harris self-esteem inventory Demographic/diagnostic Study category Conduct problems

main. For example, all of the variables in the self-perception domain that individually predicted suicidality in the bivariate regressions were entered stepwise into a multiple linear regression. A similar procedure was followed for the significant variables in the demographic/diagnostic and in the family/environment domains . This procedure then produced three independent multiple linear regression equations, each of which assessed predictors of youths ' suicidality in a different domain. Those variables that failed to make a significant contribution when assessed in these intra-domain multiple regressions were excluded from additional analyses. The significant predictors from each of the three intradomain multiple regressions were next entered stepwise into a final cross-domain multiple regression. In this manner, variables that were significant predictors when assessed against variables from the same domain could be assessed for their power in relation to variables from other domains. This procedure then took a series of potential predictors of youths' suicidality and distilled them to determine the most powerful predictors as well as less robust predictors. This model was applied first to the total sample and then to the MDD sample . Results

TOTAL SAMPLE Bivariate Regressions Bivariate regressions assessed the individual predictive power of each variable. Results are compiled in Table 1. As seen, many of the variables demonstrated statistically significant predictive ability with beta-coefficients in the low to moderate range. Youths' Self-Report Inventori es Overall , the youths' self-report inventories appear to be the most powerful individual predictors of suicidality. Greater J .Am.Acad . Child Ado/esc . Psychiatry , 30 :1.Janua ry 1991

0.03 0.04 NS

0.08 0.0 1 0.04 NS NS NS

hopelessness , CHS; depressive feelings, CDI; and lower self-esteem, SCS, were associated with greater suicidality. The CLOC was not helpful in identifying suicidal youths and 'was not included in subsequent analyses . Demographic and Diagnostic Variables The next most powerful set of variables was the demographic/diagnostic group. Inclusion in the MDD study category and conduct problems highly predicted suicidal ity. Age demonstrated a trend . Because previous investigations have found age to be associated with youths' suicidality , age was retained with study category and conduct problem s in the subsequent intra-domain multiple regression . Sex , separation anxiety, and dysthymia were excluded from subsequent analyses . Family and Environmental Variables As a group, the family/environment variables demonstrated the least ability to predict suicidality . A greater lifestress score demonstrated a significant ability to predict suicidality, while mothers ' psychopathology demonstrated near significant predictive power. Greater psychiatric impairment of the youths' mothers was associated with lower suicidality in their children. Because of past studies indicating a role for maternal psychopathology and the trend here in the bivariate analysis, mothers' psychopathology was retained in the intra-domain regression . INTRA-DoMAIN MULTIPLE REGRESSIONS Each of the variables with significant predictive ability was next entered stepwise into its respectiv e intra-domain multiple regression. The results are shown in the first part of Table 2. Of the three self-report inventories, the CHS and the CD! significantly contributed to the prediction of suicidality when entered sequentially into the multiple regression . The SCS

89

MYERS ET AL. TABLE 3. MDD Sample: Bivariate Regressions

Variables Youths' self-report inventories Children's depression inventory Hopelessness scale Piers-Harris self-esteem inventory Children's locus-of-control Demographic/diagnostic Conduct problems Separation anxiety symptoms Prior dysthymia Severity of depression Age Sex Family/environment Mothers' psychopathology Life stress score Mothers' suicidality Fathers' psychopathology Fathers' suicidality

B-coefficients

Beta

R'

T-Test

p-Value

0.049 0.084 -0.021

0.304 0.313 -0.215

0.092 0.098 0.047

8.749 7.839 -3.477

0.004 0.007 0.07 NS

1.071 -0.638

0.240 -0.204

0.058 0.042

5.878 -4.157

0.02 0.04 NS NS NS NS

-0.210 0.153

-0.168 0.159

-0.028 0.025

2.820 2.514

0.1 0.1 NS NS NS

failed to further contribute to the prediction of suicidality. The B-coefficients and the statistical significance of the CHS and CDI from the bivariate regressions were reduced considerably in the multiple regression due to the strong predictive ability of each variable and their high correlation with each other (r = 0.621). When study category (MDD versus control status), conduct problems, and age were entered stepwise into the regression equation, both study category and conduct problems continued to significantly predict suicidality. There was no appreciable loss of the predictive ability demonstrated in the bivariate and multiple regressions. Age failed to contribute and was excluded from subsequent analyses. Of the family/environment variables, both lifestress and mothers' psychopathology continued to significantly predict suicidality. The contributions of each of these variables to the multiple regression were similar to their individual predictive power in the bivariate regressions. FINAL CRoss-DOMAIN REGRESSION

Six variables significantly predicted suicidality when assessed in relation to variables tapping similar domains: CDI, CHS, belonging to the MDD study category, conduct problems, lifestress score, and mothers' psychopathology score. These variables were entered stepwise into a final crossdomain regression. Results are shown in the latter part of Table 2. Conduct problems emerged as the strongest predictor of suicidality followed by the cm. The CHS demonstrated a trend toward significant prediction. The CHS was entered first into this final multiple regression, since it had the highest correlation coefficient with suicidality. Next, the conduct problems variable was entered. The CHS was very highly significant when the variable of conduct problems was the only other one included (beta = 0.348; t(2,100) = 3.796, P < 0.0003). However, the addition of the cm reduced the contribution of the CHS. Study category, life 90

stress, and mothers' psychopathology no longer predicted suicidality. MDD Sample The procedures outlined above for the total sample were repeated for the MDD sample. Severity of depression replaced study category. Bivariate Regressions Results of the bivariate regressions for each variable are shown in Table 3. As seen, all of the beta-coefficients were again in the low to moderate range. Youths' Self-Report Inventories As a group, the self-report inventories were again the best individual predictors of suicidality. The cm and the CHS very strongly predicted suicidality. However, the SCS demonstrated only a trend toward significant prediction. Because of its near significance and the predictive ability of the SCS in the total sample, this variable was retained in the subsequent multiple regression. The CLOC was again not helpful in determining suicidality. Note that the self-report inventories were not as strong individual predictors for the MDD sample as they were for the total sample. Demographic and Diagnostic Variables Conduct problems again highly predicted suicidality. Separation anxiety symptoms were inversely correlated with suicidality, and increasing separation anxiety therefore predicted lower suicidality in MDD youths. Age no longer demonstrated a trend. Note that the severity of depression, as measured by the Kiddie-SADS, was not a significant predictor. Thus, suicidality did not appear to be a function of how depressed the youth was. A prior diagnosis of dysthymia also was not correlated with suicidal behaviors, suggesting that suicidality did not reflect a chronically depressed state. l.Am.Acad. Child Adolesc.Psychiatry, 30:1, lanuary 1991

RISKS FOR SUICIDALITY IN MAJOR DEPRESSION TABLE

4. MDD Sample: Multiple Regressions for Intra- and Cross-Domain Analyses

Variable

B-coefficients

Intra-domain analyses Youth's self-report inventories Children's depression inventory Children's hopelessness scale Demographic/diagnostic Conduct problems Separation anxiety symptoms Family/environment Mothers' psychopathology Life stress score Final cross-domain analyses Children's depression inventory Conduct problems Separation anxiety symptoms

T-Test

p-Value

Beta

R2

0.057

0.371

0.138

3.318

1.002 -0.579

0.225 -0.185

0.058 0.092

2.290 -1.885

0.02 0.06

-0.022 0.158

-0.173 0.164

0.028 0.055

-1.747 1.658

0.08 0.1

0.044 0.914

0.273 0.215

0.092 0.138

2.681 2.108

Famity and Environment Variables

None of these variables was individually significant in predicting suicidality. Of particular interest is the failure of ~ife stress to predict suicidality in the depressed sample when It was such a strong predictor in the total sample. The maternal psychopathology correlation was again inverse. Because lifestress and mothers' psychopathology showed trends toward prediction, they were retained for the intradomain multiple regression. INTRA-DoMAIN MULTIPLE REGRESSION Results of the three sets of multiple regressions for the MDD sample are shown in the first part of Table 4. Of the ~elf-repo~ inve?tories, only the cm accounted for significant vanance III the intra-domain multiple regression. In the demographic/diagnostic regression, both conduct and separation anxiety problems contributed to the prediction of suicidality comparably to their predictive power in the bivariate regressions. When the family/environment variables were entered into their multiple regression equation, both variables still only demonstrated trends toward significant prediction. They were not included in additional analyses. FINAL CRoss-DoMAIN MULTIPLE REGRESSION Only three variables remained to be entered into the final regression for the MDD sample: CDI score, conduct problems, and separation anxiety symptoms. Results are shown in the latter part of Table 4. . Both the CDI and conduct problems significantly predieted suicidality in this final assessment. The cm surpas~ed conduct problems in predictive power. Separation anxiety no longer contributed when considered across domains. Discussion This is the first investigation to assess risks for suicidality specifically to MDD youths, and thereby to control for a diagnosis of MDD in suicide assessment. Total Sample

Comorbid conduct problems and the CDI emerged as the J.Am. Acad. Child Adolesc. Psychiatry, 30:1, January 1991

0.002 NS

0.009 0.04 NS

most powerful predictors of suicidality in the total sample. Conduct problems appeared to outweigh MDD as a risk factor, since conduct problems, but not MDD, predicted suicidality in the final cross-domain regression. However, since the large majority of youths with conduct problems also had MDD, these results might indicate that conduct problems increase the risk beyond that contributed by MDD, which was as strong a predictor as conduct problems in the intra-domain multiple regression. Other investigations have also found that suicidal youths, especially attempters and completers, have conduct problems as evidenced in impulsivity and aggression (Cohen-Sandler et aI., 1982; Pfeffer et aI., 1982, 1983a, 1983b, 1984; Apter et aI., 1988), trouble with the law (Shaffer, 1974; Miller et aI., 1982), and substance abuse (Shaffer, 1974; Shafii et aI., 1985; Brent et aI., 1988; Levy and Deykin, 1989). These data complement the current findings to suggest that conduct problems comprise a major risk of suicidality. However, the relative risks of MDD and conduct problems are not readily ascertained from these studies. Other studies have found that suicidal youths with conduct problems may endorse lower self-reported depression (Carlson and Cantwell, 1982; Marriage et aI., 1986). However, no indications were found from the cm or the Kiddie-SADS severity scale that more suicidal youths were less severely depressed. This may represent the fact that the youths in this study expressed their suicidality predominantly in ideation, planning, or mild attempts, rather than severe attempts. The lack of predictive ability for MDD in the final crossdomain regression contrasts with the continued predictive power of the CDI. This discrepancy did not reflect a significant correlation between severity of MDD and the CDI for the present sample, despite the association of MDD and the cm noted in previous studies. The CDI may measure a dimension distinct from MDD. The CDI measures depressive feelings and thinking while the youth is distressed, and these symptoms may be present in nondepressed as well as depressed youths (Carlson and Cantwell, 1982). These thoughts and feelings may be more important to the development of suicidality than the syndrome of MDD. The lack of predictive ability of MDD severity and of prior dysthymia 91

MYERS ET AL.

in the intra-domain regression for the MDD sample additionally suggests that the cm is tapping a different aspect of these youths' psychopathology. The cm has not been previously well explored as a tool for discriminating suicidal youths . Carlson and Cantwell (1982) found that the cm identified youths with greater feelings of suicidality, including suicide attempts , as well as youths with depressive disorders . Depression correlated better with suicidal ideation than with suicide attempts. Kazdin and colleagues (\ 983), however, did not find a positive correlation between suicidality and the cm in hospitalized preadolescents. The Beck Depression Inventory (Beck et aI., 1961), the parent instrument of the cm, has also been reliably used to identify depressed or suicidal adolescents in some studies (Strober et aI., 1981; Teri, 1982), but it failed to discriminate adolescent minority females who had made suicide attempts in another study (Rotheram-Borus and Trautman, 1988). Additional consideration of the cm as a simple screening tool for youthful suicidality appears warranted. Interestingly, the CDI emerged as a stronger predictor of suicidality than did the CHS. The lower predicitve ability of hopelessness in the youths in this study contrasts with its strong predictive ability in suicidal adults (Minkoff et aI., 1973; Beck et aI. , 1975; Bedrosian and Beck, 1979; Wetzel et aI., 1980; Dyer and Kreitman, 1984; Beck et aI., 1985; Fawcett et aI., 1987) and inpatient preadolescents (Kazdin et aI., 1983) but agrees with other findings in children and adolescents (Carlson and Cantwell, 1982; Asarnow et aI., 1987). One caveat applies to interpreting the self-report inventories. These questionnaires were not necessarily completed at the time that the youth was actively distressed or suicidal. For example, some youths had recovered from their MDD episode or other acute problems by the time they were assessed. Others were still distressed but felt better during the day of assessment. If all subjects were actively suicidal at the time of completing the questionnaire , hopelessness may have been a more powerful predictor. McCauley and colleaues (\ 988) previously reported normalization of the selfperception instruments when depression remits. Life stress and maternal psychopathology predicted suicidality in the bivariate and intra-domain multiple regressions, but they were not as predictive as conduct problems, the CDI, or even the CHS, since they were not significant predictors in the cross-domain regression. The suicidal youths in the current study appear similar to those of Cohen-Sandler and colleagues (1982) who experienced increasing life stresses as they matured . They also appear to come from more chaotic homes, a factor probably also reflected in the high predictive ability for conduct problems. These results are consistent with the increased life stress associated with suicidality in adults (Paykel et aI., 1975; Roy, 1982; Rich et aI., 1986). Maternal psychopathology negatively predicted suicidality in the intra-domain analyses . Several investigations have found increased occurrences of familial depression and suicidality (Pfeffer et aI., 1984, 1986; Myers et aI., 1985; Weiner and Pfeffer, 1986) similar to those in adult studies

92

(Johnson and Hunt, 1979; Murphy and Wetzel , 1982; Roy, 1983; Egeland and Sussex, 1985; Wender et aI., 1986), but others have not (Pfeffer et aI. , 1982). A more interesting finding is the negative predictive ability of maternal psychopathology. Since the beta-coefficient and variance are low, and this negative relationship has not been previously reported, it must be considered tentative. Two interpretations are suggested. Youths' suicidality may represent a dimension distinct from familial affectivity, as also suggested by Puig-Antich and colleagues (\989). Alternatively , severe mood disturbance in mothers may elicit a caretaking role from the child (Beardslee and Podorefsky, 1988) with decreased thoughts of self-destruction. The reason for the lack of predictive power for age is unclear. Previous investigations have shown an increasing risk with increasing age. The discrepancy may reflect the high degree of psychopathology for the current sample at all ages. In the total sample, then, conduct problems placed a youth at greatest risk of suicidality, compounding the risk associated with MDD. The highly significant predictive ability of conduct problems appears to emerge against a profile of depressive thinking , Cfrl; feelings of hopelessness, CHS; and increased life stress and family chaos, but possibly lower depressive psychopathology of mothers. MDD Sample

The cm and comorbid conduct problems remained the most powerful predictors of suicidality in the MDD youths. However, the CDI surpassed conduct problems in predictive ability. The increased power of the CDI does not reflect the severity or chronicity of depression. The CHS failed to indicate any predictive power in either the cross-domain regression or the intra-domain regression. In the MDD sample, more so than in the total sample, the CDI may tap a dimension of distress distinct from hopelessness, or from self-esteem, SCS. It also appears to predict risk beyond that conferred by a diagnosis of MDD . In contrast to the contributing role of conduct problems, comorbid separation anxiety symptoms may protect MDD youths against suicidal thoughts and behaviors . Separation anxiety was not significant in the final regression ; however, it negatively predicted suicidality in the intra-domain regression. The reason for its protective effect is unclear and has not been previously reported. Separation anxiety does not appear to protect by decreasing morbidity (Mitchell et aI. , 1988). Fears of separation and death may preclude the development of death impulses. The negative results are relevant. Neither maternal psychopathology nor life stress emerged as predictors in the MDD sample, even in the bivariate regressions. These results along with the decreased predictive power of conduct problems in the MDD sample suggest a decreased role for impulsivity and family chaos in the suicidality of MDD compared to heterogeneous youths. These results are consistent with those of Cohen-Sandler and colleagues (1982). They found increasing life stress through the life-span for diagnostically heterogeneous suicidal youths compared to nonsuicidal depressed youths. Thus, life stress appears to l .Am .A cad. Child Adolesc .Psychiatry.Bti.l .January 1991

RISKS FOR SUICIDALITY IN MAJOR DEPRESSION

play a role in suicidality beyond the contribution it makes to depression, but its effect is less marked with more mood disordered youths. Implications of Findings

These findings suggests that MDD youths with suicidality might comprise a distinct group of MDD youths. Theyappear to differ from a more heterogeneous group regarding the effects of life stress and maternal psychopathology, but they share a common proclivity for impulsivity that may distinguish them from other MDD youths. Puig-Antich and colleagues (1989) also concluded that nonsuicidal, nonconduct disordered MDD children comprise a homogeneous group of familial onset mood disorder, while those MDD children with suicidality and/or conduct disorder comprise a nonfamilial form of MDD marked by impulsivity and familial dysfunction. Future investigations might focus on suicidality in conduct disordered compared to MDD youths with determination of their relative risks for suicidality. The identification of distinct subgroups of MDD youths based on impulsivity and suicidality may also help to understand the syndrome of MDD in children and teens. References Apter, A., Bleich, A., Plutchik, R., Mendelsohn, S. & Tyano, S. (1988), Suicidal behavior, depression, and conduct disorder in hospitalized adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 27:696-699. Asarnow, J. R., Carlson, G. A. & Guthrie, D. (1987), Coping strategies, self-perceptions, hopelessness, and perceived family environments in depressed and suicidal children. J Consult. Clin. Psychol.,55:361-366. Beardslee, W. R. & Podorefsky, D. (1988), Resilient adolescents whose parents have serious affective and other psychiatric disorders: importance of self-understanding and relationships. Am. J. Psychiatry, 145:63-69. Beck, A. T., Ward, C. H., Mendelsohn, M., Mock, 1. E. & Erbaugh, J. (1961), An inventory for measuring depression. Arch. Gen. Psychiatry, 4:561-571. - - Weissman, A., Lester, D. & Trexler, L. (1974), The measurement of pessimism: the hopelessness scale. J. Consult. Clin. Psychol.,42:861-865. - - Kovacs, M. & Weissman, M. M. (1975), Hopelessness and suicidal behavior: an overview. JAMA 234:1146-1149. - - Steer, R. A., Kovacs, M. & Garrison, B. (1985), Hopelessness and eventual suicide: a lO-year prospective study of patients hospitalized with suicidal ideation. Am. J. Psychiatry, 142:559-563. Bedrosian, R. C. & Beck, A. T. (1979), Cognitive aspects of suicidal behavior. Suicide Life Threat. Behav., 2:96-97. Brent, D. A., Perper, 1. A., Goldstein, C. E. et al. (1988), Risk factors for adolescent suicide: a comparison of adolescent suicide victims with suicidal inpatients. Arch. Gen. Psychiatry, 45:581588. Carlson, G. A. & Cantwell, D. P. (1982), Suicidal behavior and depression in children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 21:361-368. - - Asarnow, J. R. & Orbach, I. (1987), Developmental aspects of suicidal behavior in children: I. J. Am. Acad. Child Adolesc. Psychiatry, 26:186-192. Chambers, W. 1., Puig-Antich, J., Hirsch, M., Paex, P., Ambrosini, P. J., Tabrizi, M. A. & Davies, M. (1985), The assessment of affective disorders in children and adolescents by semi-structured interview. Arch. Gen. Psychiatry, 42:697-702. Cohen-Sandler, R., Berman, A. L. & King, R. A. (1982), Life stress and symptomatology: determinations of suicidal behavior in children. J. Am. Acad. Child Adolesc. Psychiatry, 21:178-186.

J.Am.Acad. Child Adolesc.Psychiatry, 30:1, January 1991

Corder, B. F., Shorr, W. & Corder, R. F. (1974), A study of social and psychological characteristics of adolescent suicide attempters in an urban disadvantaged area. Adolescence, 9:1-6. Dyer, J. A. T. & Kreitman, N. (1984), Hopelessness, depression, and suicidal intent in parasuicide. Br. J. Psychiatry, 144:127-133. Egeland, J. A. & Sussex, J. N. (1985), Suicide and family loading for affective disorder. JAMA 254:915-918. Endicott, J. & Spitzer, R. L. (1978), A diagnostic interview: the Schedule for affective disorders and schizophrenia. Arch. Gen. Psychiatry, 35:837-844. Fawcett, J., Scheftner, W., Clark, D. et al. (1987), Clinical predictors of suicide in patients with major affective disorders: a controlled prospective study. Am. J. Psychiatry, 144:35-40. Garber, J. (1981, September), Conceptual issues in the diagnosis of childhood depression. Paper presented at the American Psychological Association Annual Convention. Los Angeles. Garfinkel, B. D., Froese, A. & Hood, J. (1982), Suicide attempts in children and adolescents. Am. J. Psychiatry, 139:1257-1261. Guze, S. B. & Robins, E. (1970), Suicide and primary affective disorder. Br. J. Psychiatry, 117:437-438. Hoberman, H. M. & Garfinkel, B. D. (1988), Completed suicide in children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 27:689-695. Johnson, G. F. & Hunt, G. (1979), Suicidal behavior in bipolar manicdepressive patients and their families. Compr. Psychiatry, 20:159164. Kazdin, A. E., French, N. H., Onis, A. S., Esveldt-Dawson, K. & Sherick, R. B. (1983), Hopelessness, depression and suicidal intent among psychiatrically disturbed inpatient children. J. Consult. Clin. Psychol., 51:504-510. - - Rodgers, A & Colbus, D. (1986), The Hopelessness Scale for Children: psychometric characteristics and concurrent validity. J. Consult. Clin. Psychol., 54:241-245. Kovacs, M. (1980/1981), Rating scales to assess depression in school age children. Acta. Paedopsychiatrica., 46:305-315. - - Beck, A. T. (1977), An empirical-clinical approach toward a definition of childhood depression. In: Depression in Childhood, eds. J. G. Schulterbrandt & A. Raskin. New York: Raven Press, pp. 1-25. Levy, 1. C. & Deykin, E. Y. (1989), Suicidality, depression, and substance abuse in adolescence. Am. J. Psychiatry, 146:1462-1467. Marriage, K., Fine, S., Moretti, M. & Haley, G. (1986), Relationship between depression and conduct disorder in children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 25:687-691. McCauley, E., Mitchell, J., Burke, P. & Moss, S. (1988), Cognitive attributes of depression in children and adolescents. J. Consult. Clin. Psychol., 56:903-908. Miller, M., Chiles, J. & Barnes, R. (1982), Suicide attempters within a delinquent population. J. Consult., Clin. Psychol., 50:491-498. Minkoff, K., Bergman, E., Beck, A. T. & Beck R. (1973), Hopelessness, depression, and attempted suicide. Am. J. Psychiatry, 130:455-487. Mitchell, J., McCauley, E., Burke, P.M. et al. (1988), Phenomenology of depression in children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 27:12-20. - - McCauley, E., Burke, P., Calderon, R. & Schloredt, K. (1989), Psychopathology in parents of depressed children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry, 28:352-357. Murphy, G. E. & Wetzel, R. D. (1982), Family history of suicidal behavior among suicide attempters. J. Nerv. Ment. Dis., 170:8690. Myers, K. M., Burke, P. & McCauley, E. (1985), Suicidal behavior by hospitalized preadolescent children on a psychiatric unit. J. Am. Acad. Child Adolesc. Psychiatry, 24:474-480. Nowicki, S. & Strickland, B. R. (1973), A locus of control scale for children. J. Consult. Clin. Psychol., 40: 148-154. O'Carrol, P. (1987), Epidemiology of adolescent suicide. In: Strategiesfor Studying Suicide and Suicidal Behavior, eds. I. Lann & E. Moskcowita. Washington, D.C.: National Institutes of Mental Health. Orbach, I., Gross, Y. & Glaubman, H. (1981), Some common characteristics of latency-age suicidal children: a tentative model based on case study analyses. Suicide Life-Threat. Behav., 11:180-190.

93

MYERS ET AL.

- - Rosenheim, E. & Hary, E. (1987), Some aspects of cognitive functioning in suicidal children. J. Am . Acad. Child Adolesc. Psychiatry, 26:181-185. Pallis, D. J., Barraclough , B. M., Levy, A. B., Jenkins, J. S. & Sainsbury, P. (1982), Estimating suicide risk among attempted suicides. Br. J. Psychiatry, 141:37-44 . Paykel, E. S., Prusoff, B. A. & Myers, J. K. (1975), Suicide attempts and recent life events. Arch. Gen. Psychiatry, 32:327-333. Pfeffer, C. R. (1987), The Suicidal Child. New York: Guilford Press. - - Conte, H. R., Plutchik, R. & Jerret , 1. (1979), Suicidal behavior in latency-age children: an empirical study. J . Am. Acad. Child Adolesc. Psychiatry , 18:679-692. - - - - - - - - (1980), Suicidal behavior in latency-age children: an empirical study: an outpatient population. J . Am . Acad. Child Adolesc. Psychiatry, 19:703-710 . - - Solomon, G., Plutchik, R., Mizruchi, M. S. & Weiner, A. (1982), Suicidal behavior in latency-age psychiatric inpatients: a replication and cross validation. J. Am. Acad. Child Adolesc. Psychiatry, 21:564-569. - - Plutchik, R. & Mizruchi , M. S. (l983a) , Predictors of assaultiveness in latency-age children. Am. J . Psychiatry, 140:31-3 5. - - - - - - (1983b), Suicidal and assaultive behavior in children: classification measurement and interrelations. Am. J . Psychiatry, 140: 154-157. - - Zuckerman, S., Plutchik, R. & Mizruchi, M. S. (1984), Suicidal behavior in normal school children: a comparison with child psychiatric inpatients. J . Am. Acad. ChildAdolesc. Psychiatry, 23:4 16423. - - Plutchik, R., Mizruchi, M. S. & Lipkins, R. (1986), Suicidal behavior in child psychiatric inpatients and outpatients and in nonpatients. Am. J. Psychiatry, 143:733-738. Piers, E. V. (1969), Manual for the Piers-Harris Children's Selfconcept Scale. Nashville, Tennessee: Counselor Recordings and Tests. Puig-Antich, J., Goetz, D., Davies, M. et al. (1989), A controlled family history study of prepubertal major depressive disorder. Arch. Gen. Psychiatry , 46:406-4 18. Rich, C. L. , Young, D. & Fowler, R. C. (1986), The San Diego suicide study: 1. Young versus old subjects. Arch. Gen. Psychiatry, 43:577- 582. Robbins, D. R. & Alessi , N. E. (1985), Depressive symptoms and suicidal behaviors in adolescents. Am. J . Psychiatry, 142:588-5 92. Rotheram-Borus, M. J . & Trautman, P. D. (1988), Hopelessness, depression, and suicidal intent among adolescent suicide attempters. J. Am . Acad. Child Adolesc. Psychiatry, 27:700-704. Roy, A. (1982), Risk factors for suicide in psychiatric patients. Arch.

94

Gen. Psychiatry, 39:1089- 1095. - - (1983), Family history of suicide. Arch. Gen. Psychiatry, 40:971974. Shaffer, D. (1974), Suicide in childhood and early adolescence. J. Child Psyhol. Psychiatry, 15:275-291. - - ( 1982), Diagnostic considerations in suicidal behavior in children and adolescents. J . Am. Acad. Child Adolesc . Psychiatry, 2 1:4 14416. - - Fisher , P. (1981), The epidemiology of suicide in children and young adolescents. J . Am . Acad. Child Adolesc. Psychiatry, 20:545565. - - Bacon, K., Fisher , P. & Garland , A. (1987), Review of Youth Suicide Prevention Programs. New York: New York State's Task Force. Shafii , M., Carrigan, S., Whittinghill , J. R. & Derrick , A. (1985), Psychological autopsy of completed suicide in children and adolescents. Am. J. Psychiatry, 142:1061-1064. Spitzer, R. L., Endicott, J. & Robins, E. (1978), Research diagnostic criteria: rationale and reliability. Arch. Gen. Psychiatry , 35:778782. Strober, M., Green , J. & Carlson, G. (1981), Utility of the Beck Depression Inventory with psychiatrically hospitalized adolescents. J. Consult. Clin. Psychol., 49:482-483. Teri, L. (1982), The use of the Beck Depression Inventory with adolescents. J . Abnorm. Child Psychol.• 10:277- 284. Tishler, C. L. , McKenry, P. C. & Morgan, K. C. (198 1), Adolescent suicide attempts: some significant factors. Suicide Life Threat. Behav. 11:86-92. Trautman, P. & Shaffer, D. (1984), Treatment of child and adolescent suicide attempters. In: Suicide in the Young , eds. H. Sudak, A. Ford & N. Rushforth. Boston: Wiley, pp. 307-324. Velez, C. & Cohen, P. (1988), Suicidal behavior and ideation in a community sample of children: maternal and youth reports. J . Am . Acad. Child Adolesc. Psychiatry, 27:349-356. Weiner, A. S. & Pfeffer C. R. (1986), Suicidal status, depression, and intellectual functioning in preadolescent psychiatric inpatients. Compr. Psychiatry. 27:372- 380. Wender, P. H., Kety, S. S. , Rosenthal , D. et al. (1986), Psychiatric disorders in the biological and adoptive families of adoptive individuals with affective disorders. Arch. Gen. Psychiatry, 43:923929. . Wetzel, R. D., Margulies, T . & Davis, R. (1980) , Hopelessness, depression, and suicide intent. J . Clin. Psychol .. 41: 159-160. Williams, C. & Lyons, C. (1976), Family interactions and adolescent suicidal behavior: a preliminary investigation. Aust. N. Z. J . Psychiatry, 10:243-25 1.

J .Am.Acad . Child Adolesc . Psychiatry , 30 :1. January 1991

Risks for suicidality in major depressive disorder.

This investigation developed a hierarchical multiple regression model to assess the potential risk factors for suicidality in youths 7 to 17 years old...
8MB Sizes 0 Downloads 0 Views