Copyright 1994 by the American Psychological Association, Inc. 0022-006X/94/S3.00
Journal of Consulting and Clinical Psychology 1994, Vol. 62, No. 2,297-305
Psychosocial Risk Factors for Future Adolescent Suicide Attempts Peter M. Lewinsohn, Paul Rohde, and John R. Seeley An array of psychosocial risk factors for making a suicide attempt were examined in a representative sample of 1,508 older (14- to 18-year-old) high school students, 26 of whom made a suicide attempt during the year following entry into the study. Strongest predictors of future suicide attempt were history of part attempt, current suicidal ideation and depression, recent attempt by a friend, low self-esteem, and having been born to a teenage mother. The results suggest that adolescents who are depressed and those who attempt suicide share many psychosocial risk factors. The efficacy of two screeners (one consisting of 4 items and the other of 6 variables) is reported. Potential usefulness for research and communitywide prevention is discussed.
retrospective report. Only prospective research, such as the present study, provides information about the extent to which characteristics precede and predict a future suicide attempt. The general model guiding the present study is that the majority of psychosocial risk factors predictive of future suicide attempt overlap with variables associated with future depression (Lewinsohn et al., in press). Consequently, most of the variables included in the present study were selected on the basis of a known or hypothesized relation with depression. It should be noted that the variables in this study may also represent risk factors for other forms of psychopathology. Given the potentially tragic nature of adolescent suicide attempts and the elevated risk of suicide clustering among adolescents (Gould, Wallenstein, Kleinman, O'Carroll, & Mercy, 1990), the identification of adolescents at risk for future attempts before their behavior escalates and becomes more serious would be of obvious value. Whereas the very low base rate of completed suicide makes it extremely difficult for even a very good screening instrument to accurately predict the degree of risk for completed suicide, concentrating on a relatively more frequent phenomenon, such as suicide attempt, may prove more successful. In our recent review of existing instruments used to screen for suicidal ideation and attempts in children and adolescents (Garrison, Lewinsohn, Marsteller, Langhinrichsen, & Lann, 1991), a number of problems were highlighted (e.g., exclusive assessment of suicidal ideation rather than suicidal behavior; inadequate normative information; failure to assess intentionality, lethality, and other aspects of the attempt). Perhaps the most fundamental criticism was that insufficient attention has been paid to issues of validity, particularly predictive validity. To our knowledge, none of the available suicidal screeners has been evaluated prospectively. This article is the third in a series aimed at contributing to the understanding of suicidal behavior among adolescents. In the first article of the series (Andrews & Lewinsohn, 1992), data from the Oregon Adolescent Depression Project (OADP) were presented regarding the lifetime prevalence (7.1%) and 1-year total incidence (1.7%) of suicide attempts. The lifetime prevalence of attempts was higher for female adolescents (10.1%) than for male adolescents (3.8%). Past attempts were also asso-
Recent epidemiological studies suggest that the prevalence of adolescent suicide attempts is surprisingly high, with lifetime rates among high school students ranging from 3.5% to 11% (Andrews & Lewinsohn, 1992; Harkavy-Friedman, Asnis, Boeck, & DiFiore, 1987; Velez & Cohen, 1988). Given the relative frequency of adolescent suicidal behavior, knowledge regarding its etiology and prevention is of great importance. This study identifies the psychosocial risk factors of future suicide attempts in community adolescents and evaluates the relative efficacy of two screeners for future suicide attempts. The distinguishing characteristics of children and adolescents who have a history of suicide attempt have been described in a number of studies (e.g., Asarnow, Carlson, & Guthrie, 1987; Brent, Kolko, Allan, & Brown, 1990; Cohen-Sandier, Berman, & King, 1982; Fowler, Rich, & Young, 1986; Garfinkel, Froese, & Hood, 1982; Lewinsohn, Rohde, & Seeley, 1993; Pfeffer et al., 1991; Shaffer, 1988). In these studies the following have emerged as potential risk factors for suicide attempt: being female and not living with both parents; psychopathology, including a major depressive disorder, particularly when comorbid with other mental disorders; a previous suicide attempt; hopelessness, suicidal ideation, and depression-related cognitions; poor problem-solving abilities and coping skills; impulsivity; recent stressful life events, including suicide attempt by family members or friends and chronic physical illness; family violence and dysfunction; peer difficulties; and lower academic achievement and school problems. Although important, the aforementioned studies have a number of limitations. First, most are based on samples of hospitalized suicide attempters; therefore, the generalizability of findings to community samples cannot be assumed. Second, few studies have evaluated the contribution of variables when examined in combination. Third, previous studies have relied on
Peter M. Lewinsohn, Paul Rohde, and John R. Seeley, Oregon Research Institute, Eugene, Oregon. Preparation of this article was supported in part by National Institute of Mental Health Grant MH40501 to Peter M. Lewinsohn. Correspondence concerning this article should be addressed to Peter M. Lewinsohn, Oregon Research Institute, 1715 Franklin Boulevard, Eugene, Oregon 97403-1983. 297
298
P. LEWINSOHN, P. ROHDE, AND J. SEELEY
dated with single-parent homes, fathers with less education, and past psychopathology; approximately 80% of adolescents who had a past attempt met criteria for a previous psychiatric disorder. In the second article (Lewinsohn, Rohde, & Seeley, 1993), the psychosocial characteristics associated with past suicide attempts were evaluated. Most of the included psychosocial variables were associated with past attempt, although controlling for current depression level eliminated the significance of approximately half of the associations. The present study takes advantage of the prospective nature of the OADP data set and the extensive psychosocial assessment battery. Specifically, we focus on 26 adolescents who attempted suicide between the two assessment points, examining the psychosocial risk factors, singly and in combination, for future suicide attempt and the predictive efficacy of two screeners for future suicide attempts. Predictive efficacy is evaluated vis-a-vis sensitivity (proportion of true cases identified by the screener), specificity (proportion of true noncases identified by the screener), positive predictive value (PPV; proportion of true cases among those identified by the screener as cases), and negative predictive value (NPV; proportion of true noncases among those identified by the screener as noncases).
Method Subjects and Procedure The sample consisted of 1,508 adolescents who completed a diagnostic interview and questionnaire at two time points approximately 1 year apart; of these, 26 participants (1.7%) reported having made at least one suicide attempt between the two assessments. Participants (ages 14-18) were randomly selected from nine high schools representative of urban and rural districts in western Oregon. A total of 1,710 adolescents completed the initial (T,) interviews between 1987 and 1989 (additional details provided in Lewinsohn, Hops, Roberts, Seeley, & Andrews, 1993). The representativeness of the T, sample was assessed by comparing demographic characteristics of the sample with 1980 census data and with characteristics of adolescents who declined participation. Only minor differences were noted, and the participants may be considered to be representative of high school students in western Oregon. At the second assessment (T2), 1,508 participants (88.2%) returned for a readministration of the interview and questionnaire (mean interval = 13.8 months, SD = 2.3). Biases that may have emerged because of attrition in the T|-T2 panel sample were examined by comparing the adolescents who did not participate at T2 (N = 202) to the 1,508 subjects on demographic characteristics and measures of psychopathology. Few significant differences were present. The two groups did not differ on any measures of depression, and the attrition rate for subjects with a history of suicide attempt at TI (19 of 121 or 15.7%) did not significantly differ from the attrition rate for subjects without a history of suicide attempt (183 of 1,589, or 11.5%), X 2 (1,JV = 1,710)= 1.89,ns. Approximately half of the Ti-T2 panel was female (54%), with an average age of 16.5 (SD = 1.2). The majority (91 %) were Caucasian and most (74%) resided in two-parent homes. Written informed consent was obtained from all adolescents and their legal guardians, and adolescents were paid $25 for their participation at each assessment procedure.
Diagnostic Interview Participants were interviewed at Tj with an adaptation of the Schedule for Affective Disorders and Schizophrenia for School-Age Children that combined features of the Epidemiologic version (K-SADS-E; Or-
vaschel, Puig-Antich, Chambers, Tabrizi, & Johnson, 1982) and the Present Episode version (K-SADS-P). Additional questions were incorporated to provide information on the presence of most Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-IH-R; American Psychiatric Association, 1987) disorders. Interviewers completed a 14-item version of the Hamilton Depression Rating Scale (Hamilton, 1960). As part of the assessment of affective disorders, information was gathered regarding suicidal ideation (thoughts of death or dying, wishes to be dead, thoughts of hurting or killing self). These items were rated for occurrence during the previous 2 weeks and combined into a single measure of K-SADS current suicidal ideation (Andrews & Lewinsohn, 1992). Adolescents at TI were also asked, Have you ever tried to kill yourself or done anything that could have killed you? When a positive response was given, interviewers elicited a description of the behavior. The interviewers asked further questions to rule out purely thrill-seeking behaviors (e.g., skiing very fast on an extremely steep slope). On the basis of this information, interviewers rated each suicide attempt at TI (N = 121) for intentionality (on the 6-point K-SADS scale) and medical lethality (on the 11-point Lethality of Suicide Attempts Rating Scale; Smith, Conroy, & Ehler, 1984). At the T2 interview, subjects were again diagnostically interviewed. The only T2 information used in the present study is whether the subject had attempted suicide between T1 and T2. Using probes similar to those used in the Tj interview, participants were questioned about any suicide attempts since TI . Twenty-six adolescents reported making one or more attempts since the first interview (a description of the attempts is available on request). Of the future attempters, 14 (53.8%) had reported a past suicide attempt at TI , whereas 12 (46.2%) reported this attempt to be their first. The mean suicidal intent score for the attempts was 2.9 (3 = definite but very ambivalent); 26.9% of the attempters had an intent rating of 5 (very serious) or 6 (extreme; careful planning and every expectation of death). The mean medical lethality score was 2 (death is improbable); 11.4% had a score of 5 or greater (5 = death is a fifty-fifty probability). Interview notes for 30 randomly selected nonattempters at T2 were reviewed for the presence of reports of dangerous thrill-seeking behaviors; none were reported. It appears that the K-SADS structure, in which assessment of suicide attempts follows questions regarding suicidal ideation, elicits appropriate responses. Diagnostic interviewers were carefully selected and trained. For reliability purposes, all interviews were audiotaped or videotaped, and 12% were randomly selected and rated by reliability coders. With one exception (anxiety disorders, kappa = .60), kappas for current disorders were consistently greater than .80.
Self-Report Questionnaire Demographics. Adolescents reported gender, age, race (six categories), grade in school, history of repeating a grade, job history, number of siblings (natural and step), birth order, and composition of the household; parents reported maximum level of education (seven categories), age, marital status (five categories), and occupational status (nine categories). Psychosocial constructs. An extensive questionnaire battery of measures was administered, with the intent of assessing all psychosocial variables known or hypothesized to be related to depression (materials available on request). On the basis of extensive pilot studies (Andrews, Lewinsohn, Hops, & Roberts, 1993), most of the measures were shortened (unless noted, the instrument has been abbreviated). Because a large number of variables were administered, measures were reduced to a smaller number of composite scores. Variables were rationally categorized into general clusters, which were submitted to principal-components factor analysis with varimax rotation. Measures in each factor with factor loadings >.40 were standardized and summed
ADOLESCENT SUICIDE ATTEMPT using unit weighting to create composite scores. Any two composite scores found to be strongly correlated (i.e., r >.50) were combined into a single construct (see Lewinsohn et al., in press, for more detail). Using these procedures, most of the psychosocial measures were categorized into 22 constructs (the remaining 27 variables were retained as miscellaneous individual measures and are described later). A brief description of each construct is given below, including the number of items, Cronbach's coefficient alpha (based on scale scores where available rather than individual items), and test-retest (T!-T2) reliability. All variables were scored such that higher values indicated more problematic functioning. Current depression. This construct (67 items, a = .81, r = .40) consisted of the 20-item Center for Epidemiologic Studies-Depression Scale (CES-D; Radloff, 1977), the 21-item Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961), a single (5-point) item assessing depression level during the past week, the interviewerrated Hamilton Depression Rating Scale (Hamilton, 1960), and current depression (major depressive disorder and dysthymia) diagnosis. Other psychopathology: Internalizing behavior problems. We used a construct consisting of 43 items (a = .72, r = .55) assessing the tendency to worry (e.g., Maudsley Obsessional Compulsive Inventory; Hodgson & Rachman, 1977), hypomanic episodes (General Behavior Inventory; Depue et al., 1981), state anxiety (State-Trait Anxiety Inventory; Spielberger, Gorsuch, & Lushene, 1970), quantity and nature of sleep, and hypochondriasis (Pilowsky, 1967). Other psychopalhology: Externalizing behavior problems. This 49item (a - .68, r = .42) construct consisted of the number of current KSADS symptoms for attention deficit-hyperactivity, conduct, and oppositional disorders; an unpublished scale assessing conduct problems; and the number of current DSM-IH-R externalizing disorders. Other psychopathology: Suicidal ideation. This construct (a = .52, r = .39) consisted of a four-item screener and the three K-SADS items assessing current suicidal ideation. Items in the screener (I thought about killing myself, I had thoughts about death, I felt that my family and friends would be better off if I was dead, I felt that I would kill myself if I knew a way) assessed suicidal ideation, on a 4-point scale, for the past week. Stress: Daily hassles. This construct (a = .79, r = .55) consisted of 20 items from the Unpleasant Events Schedule (Lewinsohn, Mermelstein, Alexander, & MacPhillamy, 1985). Stress: Major life events. This construct (a = .78, r = .52) consisted of 14 events from the Schedule of Recent Experience (Holmes & Rahe, 1967) and the Life Events Schedule (Sandier & Block, 1979), rated for occurrence to self, significant others, or both in the past year. Negative cognitions. For this construct, we used 27 items (a = .61, r = .61) from the Frequency of Self-Reinforcement Attitude Questionnaire (Heiby, 1982), the Subjective Probability Questionnaire (Mufioz & Lewinsohn, 1976), the Dysfunctional Attitude Scale (Weissman & Beck, 1978), and items assessing perceived control over one's life (Pearlin & Schooler, 1978). Attributional style. We used the 48-item (a = .63, r = .55) Kastan Attributional Style Questionnaire for Children (Kaslow, Tannenbaum, & Seligman, 1978). Two scores were derived, standardized, and summed: positive events attributed to unstable, external, and specific causes; and negative events attributed to stable, internal, and global causes. Self-consciousness. This construct consisted of nine items (a = .74, r = .54) from the Self-Consciousness Scale (Fenigstein, Scheier, & Buss, 1975). Self-esteem. This construct consisted of nine items (a = .59, r = .62) from the Body Parts Satisfaction Scale (Berscheid, Walster, & Bohrnstedt, 1973), the Physical Appearance Evaluation Subscale (Winstead & Cash, 1984), and the Rosenberg Self-Esteem Scale (Rosenberg, 1965).
299
Social self-competence. This construct consisted of 12 items (a = .81, r = .64) from the Social subscale of the Perceived Competence Scale for Children (Barter, 1982) and adjectives assessing perceived social competence (Lewinsohn, Mischel, Chaplin, & Barton, 1980). Emotional reliance. We used 10 items (a = .83, r = .54) from the Emotional Reliance Scale (Hirschfeld, Klerman, Chodoff, Korchin, & Barrett, 1976) assessing the extent to which individuals desire more support and approval from others and are interpersonally sensitive. Academic aspirations. This nine-item construct (a = .77, r = .74) contained measures of estimated future education, grade average last term, self-perceived adequacy of school performance, perceived ability to complete college, and items adapted from the Importance Placed on Life Goals Scale (Bachman, Johnston, & O'Malley, 1985) assessing the importance of future academic goals. Family aspirations. This five-item construct (a = .61, r = .58) assessed the importance of future goals related to marriage and family (adapted from Bachman et al., 1985). Occupational aspirations. This 3-item construct (a = .63, r = .48) assessed the importance of future income level and steady employment (adapted from Bachman et al., 1985). Coping skills. This construct (a = .76, r= .55) consisted of 17 items from the Self-Control Scale (Rosenbaum, 1980), the Antidepressive Activity Questionnaire (Rippere, 1977), and the Ways of Coping Questionnaire (Folkman & Lazarus, 1980). Family social support. This 24-item construct (a = .77, r = .64) consisted of items from the Appraisal of Parents subscale of the Conflict Behavior Questionnaire (Prinz, Foster, Kent, & O'Leary, 1979), the Parent Attitude Research Instrument (Schaefer, 1965), the Cohesion subscale of the Family Environment Scale (Moos, 1974), the Competence scale of the Youth Self-Report (Achenbach & Edelbrock, 1987), and a paper-and-pencil adaptation of the Arizona Social Support Interview Schedule (ASSIS; Barrera, 1986). Friends' social support. This 15-item construct (a = .72, r = .60) consisted of items from the Social Competence Scale (Harter, 1982), the UCLA Loneliness Scale (Russell, Peplau, & Cutrona, 1980), and the Competence scales of the Youth Self-Report and the ASSIS. Interpersonal: Conflict with parents. This construct (a = .81, r = .51) consisted of the 45-item Issues Checklist (Robin & Weiss, 1980). The occurrence of events during the previous 2 weeks and average intensity were standardized and summed. Interpersonal: Attractiveness. For this construct, we used the 17item (a = .94, r = .22) interviewer-rated Interpersonal Attraction Measure (McCroskey & McCain, 1974). Physical health and illness in the past year. This construct (a = . 51, r = .46) consisted of number of visits to a physician, days spent in bed as a result of illness, and the occurrence of 88 physical symptoms during the previous 12 months. Maturational level. This construct (eight items, a = .64, r = .56 for female adolescents; 11 items, a = .74, r = .74 for male adolescents) contained items adapted from Petersen, Crockett, Richards, and'Boxer (1988) assessing current pubertal status. Items assessed the amount of hair on various body parts and changes in body shape (female adolescents only) and changes in voice (male adolescents only).
Miscellaneous Measures Additional measures were not included in the constructs because either they did not fit rationally into a general cluster or they did not load sufficiently on a factor (i.e., ^ .40). Two items from the BDI were selected to assess hopelessness (r = .23) and suicidal ideation (r = .27). Adolescents reported whether a family member (r = . 16) or close friend (r = . 18) had tried to commit suicide within the past year, whether either of their parents had died before the adolescent was 12 years of age, the number of times during their lifetime they had moved to a new place to
300
P. LEWINSOHN, P. ROHDE, AND J. SEELEY
live (r = .83), and whether they had moved within the past year (r = . 18). Six items from the Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960) assessed the tendency to present oneself in a socially desirable manner (a = .53, r = .54). Fifteen items from the Hypomanic Personality Scale (Eckblad & Chapman, 1986) identified persons with hypomanic personality style (a = .68, r = .55). Ten items from the Vocabulary subtest of the Shipley Institute of Living Scale (Shipley, 1940) measured verbal ability (a = .72, r = .71). Adolescents reported satisfaction with grade average (r = .24), perceived parental satisfaction with grade average (r = .30), frequency of school absence (r = .34) and tardiness (r = .39) in the past 6 weeks, and frequency of failure to complete homework (r = .46). Additional health measures included self-rated health (r = .49); the Quetelet Obesity Index (weight/height2) as a measure of adiposity (body fat percentage; r = .75); frequency of exercise (r = .47); adequacy of appetite (r = .43); energy level relative to others in age group (r = .51); use of medications (r = .29); overnight stays in a hospital (r = .05); whether an injury or illness in the past year had caused difficulties with feeding, dressing (r = .24), climbing stairs, or getting outdoors (r = .27), an inability to work or participate in school (r = .23), or a reduction or change in activity level (r = .21); and the lifetime occurrence of 88 physical symptoms.
Statistical Analyses Demographic characteristics were examined using chi-square analysis and analysis of variance, with alpha set at p < .05 (two-tailed). Psychosocial risk factors for future suicide attempt were examined using logistic regression, which follows the same general principles used in linear regression. The logistic model is considered more appropriate than the linear model when the outcome variable is dichotomous (Hosmer & Lemeshow, 1990). The associations between risk factors and future suicide attempt are presented by means of the odds ratio, which compares the odds of making a suicide attempt for those with the risk factor to the odds of making an attempt for those without the risk factor. An odds ratio of 1.0 indicates no association between the risk factor and future suicide attempt. An odds ratio of 2.0 for a dichotomous variable indicates that suicide attempts occur twice as often among those who have the risk factor than those who do not. For continuous variables, the odds ratio indicates the increased likelihood of suicide attempt given an increase of one unit in the independent variable. The predictive significance of each of the psychosocial variables and the interaction of variables with gender were examined first. Given the number of psychosocial variables, the alpha level of each comparison was set at p < .005. Because specific hypotheses stipulated a predicted direction of results, one-tailed significance tests were computed. Given that many of the individual variables were correlated with current depression level (that was the reason for their inclusion into the study) and were intercorrelated with each other, two additional steps were taken. First, the ability of significant individual risk factors to predict future suicide attempt controlling for current depression level was examined. Second, predictors that remained significant after controlling for depression were considered for inclusion in multivariate analyses. The significant individual risk factors (controlling for current depression) were examined in multiple logistic regression analyses to determine which variables made a unique (i.e., independent of the other variables) contribution to the prediction of future suicide attempt. The model goodness-of-fit statistics had acceptable values for all of the logistic regression analyses reported herein (i.e., the -2 log likelihood chisquare p values exceeded . 10). The screening properties of both a brief suicide screening instrument and the multiple logistic regression-derived risk factors were then examined. Overall classification ability and comparison of the two screeners were examined by the \buden Index (Youden, 1950), a measure (ranging from 0 to 1, with standard error) of the proportions correctly classified in case and control groups.
Results Demographic Characteristics Associated With Future Suicide Attempt Two of the demographic variables significantly predicted future attempt: Compared with the control group, the future attempters had younger mothers (39.0 years of age vs. 41.2), t( 1,427) = 2.08, and had parents with less education, t( 1,374) = 2.06. To clarify the magnitude of the effect of mother's age, mother's age was dichotomized into teenage mother (i.e., mother was less than 20 years of age at the time of the subject's birth) versus nonteenage mother. Twenty-seven percent of the attempters had a teenage mother compared with 9% of the nonattempters, x20, N = 1,300) = 7.90, p < .01. When teenage mothers were not included in the analysis, mother's age no longer significantly predicted suicide attempts. There was a trend for female adolescents to be overrepresented in the future attempters group (69% vs. 53%); x 2 (l, N = 1,508) = 2.56, p = .11. We examined whether the expected preponderance of female future attempters would be accentuated among those with a history of suicide attempt. Female adolescents with a past attempt were no more likely than male adolescents with a past attempt to make an additional attempt during the course of the study; rates for these female adolescents were actually slightly lower than rates for the male adolescents (13% and 17%, respectively), x 2 (l, # = 102) = .23, ns. Differences on the remaining variables (age, race, grade in school, repeating a grade, adolescent's job history, father's age, number of siblings, number in household, birth order, and either parent's occupational status) did not attain the .05 level of significance.
Psychosocial Variables Associated With Future Suicide Attempt Separate logistic regression analyses were conducted for each of the psychosocial variables examining the main effect of the variable and then the interaction of the variable with gender in association with future suicide attempt. None of the interactions with gender were significant. The odds ratios and 99.5% confidence intervals for significant variables appear in the first column of Table 1 (the second and third columns of Table 1 contain data from this sample that have been presented elsewhere; these two columns will be addressed in the discussion section). Measures of past psychopathology were also examined but were nonsignificant. Future suicide attempts were significantly associated with approximately half (22 of 50) of the psychosocial variables. Examination of the magnitude of the odds ratios indicates that the strongest predictor of future attempt was the occurrence of a past attempt. Adolescents with a history of suicide attempt were almost 18 times more likely than adolescents with no past attempt to make a suicide attempt during the Ti-T2 period.
Associations With Future Suicide Attempt Controlling for Depression Because the independent variables were initially selected on the bases of a known or hypothesized association with depres-
301
ADOLESCENT SUICIDE ATTEMPT Table 1 Relation ofPsychosocial Variables and Future Suicide Attempt, Future Depressive Episode, and Past Suicide Attempt Future depression8
Future attempt
Past attemptb
Variable
OR
99.5% CI
OR
99% CI
OR
99% CI
Current Depression" Externalizing Problem Behaviors" Internalizing Problem Behaviors" Suicidal Ideation" Daily Hassles" Major Life Events" Negative Cognitions" Attributional Style" Self-Consciousness" Self-Esteem" Social Self-Competence" Emotional Reliance" Academic Aspirations" Family Aspirations" Occupational Aspirations" Coping Skills" Family Support" Friends Support" Conflict with Parents" Interpersonal Attractiveness" Health/Illness" Maturation Level" Past suicide attempt BDI hopelessness" BDI suicidal ideation" Attempt by family Attempt by friend Early death of parent Moved in lifetime" Moved in last year Social desirability" Hypomanic measure" Vocabulary" Satisfaction with grades Parental satisfaction with grades Missed school days" Late for school" Failure to do homework" Self-fated health" Obesity" Frequency of exercise" Problems with appetite" Energy level" Use of medications Hospital stays" Difficulties with: Feeding/dressing Climbing stairs Work or school Activities Lifetime physical symptoms"
3.0 2.3
2.0-4.3 .4-3.6* .8-8.8 .6-2.7* .6-1.7 .0-1.2 2.4-9.5* 1.6-5.7* 1.0-1.3 2.0-8.1* 1.5-5.4 1.0-1.2 — — — 1.1-1.3* 2.0-9.2* 1.3-5.4 —
1.8 1.6 2.8 1.5 1.0 1.1 1.6 1.4 1.1 1.5 — 1.1 — — — 1.1 1.6 —
1.4-2.4 1.2-2.2 1.9-4.1 1.2-1.9 1.0-1.1 1.0-1.1 1.2-2.3 1.1-1.9 1.0-1.1 1.1-2.1 — 1.0-1.1 — — — 1.0-1.1 1.1-2.2 — 1.1-1.7
— — — 6.0-52.0* .6-5.1 3.6-13.2* — .2-10.2* — — — — — — — — — — — — — — 1.5-4.8* — — —
—
2.4 2.3 3.6 1.8 1.0 1.1 3.2 2.1 1.1 2.5 1.8 1.1 2.0 — — 1.1 2.6 2.1 1.4 1.6 1.9 — NA 2.0 3.9 —
1.9-3.0 1.7-3.0 2.4-5.3 1.5-2.2 1.0-1.1 1.0-1.1 2.2-4.5 1.5-2.8 1.0-1.1 1.8-3.5 1.3-2.5 1.0-1.2 1.5-2.8 — — 1.1-1.2 1.8-3.8 1.5-3.1 1.1-1.7 1.2-2.1 1.4-2.5 — NA 1.4-2.8 2.7-5.6 — 1.1-3.6 —
4.0
2.1 1.6 1.1 4.8 3.1 1.1 4.0 2.8 1.1 — — — 1.2 4.3 2.7 — — — — 17.6 2.8 6.9 — 3.5 — — — — — — — — — — — — — — 2.7 — — — — 8.2 — 3.2 —
— 1.6-44.0 — 1.1-9.5 —
1.4
1.5 — 6.1 — 2.1 — — — — — — — — 1.7 — — —
—
1.2-1.9 — 3.4-11.2 — 1.5-3.0 — — — — — — — — 1.1-2.7 — —
2.0 — 1.4 —
1.1-1.7
—
1.2 1.1 — 2.6 2.6 1.5
1.0-1.4 1.0-1.2 — 1.6-4.4 1.6-4.4 1.2-1.8
1.4
1.3 1.8 — — 1.8 2.2 — —
1.0-1.5 1.1-3.1 — — 1.2-2.7 1.5-3.2 — —
1.9 — 1.5 1.6 1.9 2.0 3.8
1.2-1.7 1.3-2.7 — 1.1-2.0 1.1-2.2 1.3-2.8 1.1-3.7 1.7-8.5
— 5.2 — 2.5 1.1
— 1.4-19.2 — 1.2-5.4 1.0-1.1
— 2.2 2.0 1.0
— 1.3-3.6 1.1-3.5 1.0-1.1
Note. Dashes indicate that odds ratios are nonsignificant. OR = odds ratio, adjusted for gender. CI = confidence interval; NA = not applicable. * Results presented in Lewinsohn, Roberts, Seeley, Rohde, Gotlib, and Hops (in press). b Results presented in Lewinsohn, Rohde, and Seeley (1993). " Continuous variable. * Significant at p < .05 after controlling for concurrent depression.
302
P. LEWINSOHN, P. ROHDE, AND J. SEELEY
sion, the extent to which significant psychosocial variables continued to predict future attempt was examined in logistic regressions controlling for the effects of current depression level. Controlling for current level of depression eliminated half of the previously significant variables. Future suicide attempts continued to be significantly (p < .05) associated with externalizing problem behaviors (odds ratio = 1.6), suicidal behavior (1.5), pessimism (2.3), attributions (1.7), self-esteem (2.3), coping skills (1.1), social support from family (2.6), past suicide attempt (8.1), BDI suicidal ideation (4.4), suicide attempt by a friend (2.4), and problems with appetite (1.5).
Multivariate Associations With Future Suicide Attempt To examine the extent to which significant individual variables made a unique contribution to the prediction of future suicide attempt in this sample, we entered two blocks of measures in a multiple logistic regression analysis. The first block consisted of current depression level and the significant demographic differences, including gender, which had approached significance. The second block, which was examined with backward stepwise variable deselection, consisted of the psychosocial variables that had a significant association with future suicide attempt after controlling for current depression. Criterion for removal from the solution was set at p > .05 (likelihoodratio test). In addition to current depression (odds ratio = 1.8; 95% confidence interval = 1.2-2.6) and younger mother's age (1.1; 1.01.2), four psychosocial variables were retained in the final solution: past suicide attempt (5.8; 2.5-13.6), recent suicide attempt by a friend (3.2; 1.4-7.3), suicidal ideation as per the BDI item (2.6; 1.5-4.5), and self-esteem (1.8; 1.0-3.1). Similar to multiple linear regression, multiple logistic regression combines the predictor variables (within the logistic function) by multiplying each variable by its beta weight and summing these terms plus a constant. Using an approximation ofR2 proposed for use with logistic regression (Hosmer & Lemeshow, 1990; p. 148), the six-factor solution accounted for 39.8% of the variance of future suicide attempts. To examine the possibility that the results were in part attributable to a general tendency of adolescent suicide attempters to answer questionnaire items in a less socially desirable direction, we recomputed the multiple logistic regression analysis, forcing in the social desirability measure as the first block, followed by current depression level and the demographic variables as the second block and the psychosocial variables as the third block. Social desirability did not make a significant contribution and did not change the contribution of the other variables.
Screening for Future Suicide Attempt The screening efficacy of the brief (four-item) suicide screener, at five cutpoints, is shown in Table 2. As can be seen, sensitivity and specificity were both greater than 80% at a cutpoint greater than or equal to 6, although PPV was only 7%. Applied to the study sample, this translates into 21 true positives (i.e., identified future attempters), 269 false positives, 5 false negatives (i.e., future attempters not identified by the screener), and 1,211 true negatives. A cutpoint of 6 was also
the point that maximized the \buden Index (.63). By raising or lowering the cutpoint, one could maximize the screening property of most interest (e.g., a cutpoint of 7 would increase the specificity [88%] and PPV [9%] at the expense of sensitivity [65%]). For all calculations, NPV was greater than or equal to 98%, which is attributable to the fact that the vast majority of adolescents identified as noncases by the screener did not make a suicide attempt during the 1-year follow-up period. Using the predicted probability retained from the multiple logistic regression solution, the screening ability of the six TI risk factors was calculated for various predicted probabilities of future attempt; results appear in the middle portion of Table 2. Sensitivity (81%), specificity (92%), and PPV (17%) were maximized at the .04 predicted probability cutpoint (Youden Index = .75). The potential value of the six risk factors as a screener for future suicide attempt can also be illustrated by treating each risk factor as a dichotomous variable (continuous variables were considered to be present if the subject scored one or more standard deviations above the mean). The distribution of risk factors was highly skewed: 49% of the sample had zero risk factors, 91% had no more than two risk factors, and only 8 subjects (0.6%) had five or six of the factors. Screening properties based on the number of factors are shown in the lower portion of Table 2. As can be seen, the screening ability of a cutpoint of three risk factors approaches the level of classification obtained in the multiple logistic regression solution, and for those with four or more risk factors, the PPV is 20%. The mean number of risk factors for female and male attempters did not significantly differ (3.1 and 3.4, respectively), t(24) - —.64, ns, although female nonattempters (1.0) had more risk factors than male nonattempters (0.6), t( 1,395) = 6.44, p < .001. Rates of increase in the probability of future suicide attempt as a function of the number of risk factors were comparable for female and male adolescents.
Discussion Two demographic variables predicted future adolescent suicide attempts: having been born to a teenage mother and less parental education. Although there was a trend for more female adolescents to make a future suicide attempt than male adolescents (2.2% vs. 1.1 %, respectively), gender differences in the predictive ability of the psychosocial variables were nonsignificant. At the univariate level, many of the psychosocial variables were significantly associated with future suicide attempt. Adolescents who made a future suicide attempt were more likely to have a history of attempt, psychopathology (depression, externalizing problems, internalizing problems), depression-related cognitions, current suicidal ideation, low self-esteem, low perceived social support from family members and friends, exposure to suicidal behavior by peers, poor self-rated health, and greater perceived functional impairment as a result of illness or injury. In multivariate analyses, six variables were found to make significant unique contributions to the prediction of future suicide attempt: past suicide attempts, recent suicide attempt by a friend, suicidal ideation as per one item of the BDI, current depression, lowered self-esteem, and having a younger mother.
303
ADOLESCENT SUICIDE ATTEMPT Table 2 Performance of Two Screenersfor Future Suicide Attempts Sensitivity
Specificity
PPV
NPV
gg 81 65 58 50
66 82 88 92 94
4 7 9 11 12
100 100 99 99 99
92 g5 85 81 77 58 42 27
78 gg 91 92 93 97 99 100
g 12 15 17 19 26 39 70
100 100 100 100 100 99 99 9g
100 92 73 35 15
50 80 92 9g 100
4 8 14 20 50
100 100 100 99 98
Screener Four-item screener Cutpoint 5 6 7 8 9 Six-factor screener Multiple logistic regression solution predicted probability .01 .02 .03 .04 .05 .10 .20 .50 Presence of dichotomous variables 1 2 3 4 5
Note. Sensitivity = proportion of attempters identified by screener; Specificity = proportion of nonattempters identified by screener; PPV = proportion of attempters among those identified by screener as attempters; NPV = proportion of nonattempters among those identified by screener as nonattempters. Examination of the two screening instruments suggests that reasonably high levels of sensitivity and specificity for future adolescent suicidal behavior can be obtained, although the ratio of false positives to true positives will be high (i.e., low PPV). It is possible that some of the false-positive adolescents would have attempted suicide if they had been assessed over a longer followup period. As expected, the strongest predictor of future suicide attempt was past attempt. However, it is important to note that five additional measures in the multiple logistic regression analyses were predictive even after controlling for the influence of past attempt. Consistent with previous research with adolescents and adults, psychopathology appeared to represent an important potentiator or precondition for suicidal behavior, and depression was the most influential form of pathology. The significance of reduced self-esteem in suicidal behavior has been addressed in an extensive literature regarding the association of self-derogation, self-hatred, and suicidal behavior (e.g., Ryan et al., 1987), although it has never been evaluated prospectively in adolescents. We are not aware of any other study reporting an association between adolescent suicide attempt and having been born to a teenage mother. If replicated, this variable represents an important new risk factor and adds to the growing literature pointing to the vulnerability of children born to teenage mothers (e.g., Hechtman, 1989). Some measures were not associated with future suicide attempt, even at the univariate level (e.g., early death of a parent). Contrary to expectation, the influence of suicide attempts by family members (which included parents, stepparents, siblings,
or other relatives) was nonsignificant; 3.8% of the adolescent attempters had a family member attempt suicide, compared with 4.1% of the adolescent nonattempters. Our general model of suicidal behavior stipulated that variables predictive of depression in adolescents would also predict future suicide attempts. The second column of Table 1 presents the T! variables that significantly (p < .01) predicted a future (i.e., T!-TZ) depressive episode in this sample, as originally reported in Lewinsohn et al. (in press). As can be seen, future depression and future suicidal behavior are indeed associated with many of the same psychosocial risk factors. Specifically, of the 22 primary psychosocial constructs in the study, 13 were associated with the future occurrence of both depression and suicide attempt. Of the miscellaneous variables, past suicide attempt, BDI suicidal ideation, and three health problems predicted both events. Also shown in Table 1 (third column) are the variables found to be significantly (p < .01) associated with a past suicide attempt in the study sample (Lewinsohn et al., 1993). Almost all of the variables predictive of suicide attempt and depression were also significantly associated with past suicide attempt. One perspective that may be useful is to think of variables only associated with future attempts as triggers for suicidal behavior. Only one variable in the present study (i.e., illness- or injuryrelated difficulties climbing stairs and getting outdoors) fit that pattern. Conversely, variables only associated with past attempts can be conceptualized as consequences. According to this perspective, the fact that indices of problematic school performance (e.g., academic aspirations, self-satisfaction and per-
304
P. LEWINSOHN, P. ROHDE, AND J. SEELEY
ceived parental satisfaction with grades, school absenteeism) were only associated with past attempts suggests that impaired academic functioning follows and may be a consequence of suicidal behavior. A similar pattern is suggested for several selfreport measures of physical illness (i.e., following an attempt, adolescents are more likely to rate themselves as being in poor health and as taking medications). Several limitations of the study should be noted. First, many statistical comparisons were conducted, which increase the risk of Type I error; interpretation of positive results should be considered tentative pending cross-validation. A second limitation of the study is the exclusive reliance on adolescent self-report. Consequently, our measure of suicide attempts is potentially vulnerable to over- and underreporting biases. The use of multiple informants and, in cases of disagreement, best estimate diagnoses would have been more rigorous in providing independent corroboration of the actual occurrence of the attempt. In the strictest sense, this study only deals with reports of psychosocial functioning predicting future reports of suicide attempts. An additional limitation is that the number of adolescents who made an attempt between the two assessment points was relatively small, which made it difficult to examine gender differences. We were also unable to determine whether the predictors for repeat attempters were different from predictors for first-time attempters. Because a few of the psychosocial measures had only fair to moderate levels of internal consistency (i.e., a = .50-.70), alternative labels for some constructs are possible. Last, it is important to keep in mind that this article focuses on suicide attempts; the risk factors for suicide completions may be different. Despite such limitations, the study did have some positive points. First, the prospective design provided an opportunity to determine the ability of psychosocial risk factors to predict future suicide attempts. To our knowledge, no other prospective study of adolescent suicide attempts has been published to date. Second, our data set provided a unique opportunity to compare the predictors of suicide attempt with those of future depression. Third, the study made use of a representative community sample, which should enhance the generalizability of the findings. Finally, more so than in previous research, a wide array of putative risk factors was examined both singly and in combination.
References Achenbach, T. M., & Edelbrock, C. S. (1987). Manual for the Youth SelfReport and Profile. Burlington: University of Vermont, Department of Psychiatry. American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author. Andrews, J. A., & Lewinsohn, P. M. (1992). Suicidal attempts among older adolescents: Prevalence and co-occurrence with psychiatric disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 31, 655-662. Andrews, J. A., Lewinsohn, P. M., Hops, H., & Roberts, R. E. (1993). Psychometric properties of scales for the measurement of psychosocial variables associated with depression in adolescence. PsychologicalReports, 73, 1019-1046. Asarnow, J. R., Carlson, G. A., & Guthrie, D. (1987). Coping strategies,
self-perceptions, hopelessness, and perceived family environments in depressed and suicidal children. Journal of Consulting and Clinical Psychology, 55, 361-366. Bachman, J. G., Johnston, J., & O'Malley, P. M. (1985, May). Some recent trends in the aspirations, concerns, and behaviors of American young people. Paper presented at the Annual Conference of the American Association for Public Opinion Research, McAfee, NJ. Barrera, M., Jr. (1986). Distinctions between social support concepts, measures, and models. American Journal of Community Psychology, 14, 413-445. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-57'1. Berscheid, E., Walster, E., & Bohrnstedt, G. (1973). The happy American body: A survey report. Psychology Today, 7, 119-131. Brent, D. A., Kolko, D. J., Allan, M. J., & Brown, R. V. (1990). Suicidality in affectively disordered adolescent inpatients. Journal of the American Academy of Child and Adolescent Psychiatry, 29, 586-593. Cohen-Sandier, R., Berman, A. L., & King, R. A. (1982). Life stress and symptomatology: Determinants of suicidal behavior in children. Journal of the American Academy of Child Psychiatry, 21, 178-186. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354. Depue, R. A., Slater, J. F., Wolfsetter-Kausch, H., Klein, D., Goplerud, E., & Farr, D. (1981). A behavioral paradigm for identifying persons at risk for bipolar depressive disorder: A conceptual framework and five validation studies. Journal of Abnormal Psychology, 90, 381-437. Eckblad, M., & Chapman, L. J. (1986). Development of a scale for hypomanic personality and validation. Journal of Abnormal Psychology, 95,214-222. Fenigstein, A., Scheier, M. F., & Buss, A. H. (1975). Public and private self-consciousness: Assessment and theory. Journal of Consulting and Clinical Psychology, 43, 522-527. Folkman, S., & Lazarus, R. S. (1980). An analysis of coping in a middleaged community sample. Journal of Health and Social Behavior, 21, 219-239. Fowler, R. C., Rich, C. L., & Young, D. (1986). San Diego suicide study. II. Substance abuse in young cases. Archives of General Psychiatry, 43, 962-965. Garfinkel, B. D., Froese, A., & Hood, J. (1982). Suicide attempts in children and adolescents. American Journal of Psychiatry, 139, 1257-1261. Garrison, C. Z., Lewinsohn, P. M., Marsteller, F., Langhinrichsen, J., & Lann, I. (1991). The assessment of suicidal behavior in adolescents. Suicide and Life Threatening Behavior, 21, 217-230. Gould, M. S., Wallenstein, S., Kleinman, M. H., O'Carroll, P., & Mercy, J. (1990). Suicide clusters: An examination of age-specific effects. American Journal of Public Health, 80,211-212. Hamilton, M. (1960). A rating scale for depression. Journal of Neurology and Neurosurgery, 23, 56-61. Harkavy-Friedman, J., Asnis, G., Boeck, M., & DiFiore, J. (1987). Prevalence of specific suicidal behaviors in a high school sample. American Journal of Psychiatry, 144, 1203-1206. Harter, S. (1982). The Perceived Competence Scale for Children. Child Development, 53, 87-97. Hechtman, L. (1989). Teenage mothers and their children: Risks and problems: A review. Canadian Journal of Psychiatry, 34, 569-575. Heiby, E. M. (1982). A Self-Reinforcement Questionnaire. Behavior Research Therapy, 20, 397-401. Hirschfeld, R. M. A., Klerman, G. L., Chodoff, P., Korchin, S., & Barrett, J. (1976). Dependency, self-esteem, and clinical depression. Journal of the American Academy of Psychoanalysis, 4, 373-388.
ADOLESCENT SUICIDE ATTEMPT Hodgson, R. J., & Rachman, S. (1977). Obsessional-compulsive complaints. Behavior Research Therapy, 15, 389-395. Holmes, T. H., & Rahe, R. H. (1967). Schedule of recent experiences. Seattle: University of Washington School of Medicine. Hosmer, D. W., & Lemeshow, S. (1990). Applied logistic regression. New York: Wiley. Kaslow, N., Tannenbaum, R., & Seligman, M. (1978). TheKASTAN:A children's attributional style questionnaire. Unpublished manuscript. Lewinsohn, P. M., Hops, H., Roberts, R. E., Seeley, J. R., & Andrews, J. A. (1993). Adolescent psychopathology: I. Prevalence and incidence of depression and other DSM-HI-R disorders in high school students. Journal of Abnormal Psychology, 101, 133-144. Lewinsohn, P. M., Mermelstein, R. M., Alexander, C, & MacPhillamy, D. (1985). The Unpleasant Events Schedule: A scale for the measurement of aversive events. Journal of Clinical Psychology, 41,483-498. Lewinsohn, P. M., Mischel, W, Chaplin, W, & Barton, R. (1980). Social competence and depression: The role of illusory self-perceptions. Journal of Abnormal Psychology, 89, 203-212. Lewinsohn, P. M., Roberts, R. E., Seeley, J. R., Rohde, P., Gotlib, I. H., & Hops, H. (in press). Adolescent psychopathology: II. Psychosocial risk factors for depression. Journal of Abnormal Psychology. Lewinsohn, P. M., Rohde, P., & Seeley, J. R. (1993). Psychosocial characteristics of adolescents with a history of suicide attempt. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 6068. McCroskey, J. C., & McCain, T. A. (1974). The measurement of interpersonal attraction. Speech Monographs, 41,261 -266. Moos, R. H. (1974). Family environment scale and preliminary manual. Palo Alto, CA: Consulting Psychologists Press. Munoz, R. F, & Lewinsohn, P. M. (1976). The Subjective Probability Questionnaire. Unpublished manuscript. Orvaschel, H., Puig-Antich, J., Chambers, W, Tabrizi, M. A., & Johnson, R. (1982). Retrospective assessment of prepubertal major depression with the Kiddie-SADS-E. Journal of the American Academy of Child Psychiatry, 21, 392-397. Pearlin, L. I., & Schooler, C. (1978). The structure of coping. Journal of Health and Social Behavior, 19, 2-21. Petersen, A. C., Crockett, L., Richards, M., & Boxer, A. (1988). A selfreport measure of pubertal status: Reliability, validity, and initial norms. Journal of Youth and Adolescence, 17, 117-133. Pfeffer, C. R., Klerman, G. L., Hurt, S. W, Lesser, M., Peskin, J. R., & Siefker, C. A. (1991). Suicidal children grow up: Demographic and clinical risk factors for adolescent suicide attempts. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 609-616. Pilowsky, I. (1967). Dimensions of hypochondriasis. British Journal of Psychiatry, 113, 89-93. Prinz, R. J., Foster, S., Kent, R. N., & O'Leary, K. D. (1979). Multivariate assessment of conflict in distressed and nondistressed motheradolescent dyads. Journal of Applied Behavior Analysis, 12,691-700. Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale
305
for research in the general population. Applied Psychological Measurement, 1, 385-401. Rippere, V. (1977). Some cognitive dimensions of antidepressive behavior. Behavior Research Therapy, 15, 57-63. Robin, A. L., & Weiss, J. G. (1980). Criterion-related validity of behavioral and self-report measures of problem-solving communication skills in distressed and nondistressed parent-adolescent dyads. Behavioral Assessment, 2, 339-352. Rosenbaum, M. (1980). A schedule for assessing self-control behaviors: Preliminary findings. Behavior Therapy, 11, 109-121. Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Russell, D., Peplau, A., & Cutrona, C. (1980). The revised UCLA Loneliness Scale: Concurrent and discriminant validity evidence. Journal of Personality and Social Psychology, 39, 472-480. Ryan, N. D., Puig-Antich, J., Ambrosini, P., Rabinovich, H., Robinson, D., Nelson, B., lyengar, S., & Twomey, J. (1987). The clinical picture of major depression in children and adolescents. Archives of General Psychiatry, 44, 854-861. Sandier, I. N., & Block, M. (1979). Life stress and maladaption of children. American Journal of Community Psychology, 7, 425-439. Schaefer, E. S. (1965). Children's reports of parental behavior: An inventory. Child Development, 36, 413-424. Shaffer, D. (1988). The epidemiology of teen suicide: An examination of risk factors. Journal of Clinical Psychiatry, 49, 36-41. Shipley, W. C. (1940). A self-administering scale for measuring intellectual impairment and deterioration. Journal of Psychology, 9, 371377. Smith, K., Conroy, R. W, & Ehler, B. D. (1984). Lethality of Suicide Attempt Rating Scale. Suicide and Life Threatening Behavior, 14, 215-242. Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). Manual for the State- Trait A nxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Velez, C. N., & Cohen, P. (1988). Suicidal behavior and ideation in a community sample of children: Maternal and youth reports. Journal of the American Academy of Child and Adolescent Psychiatry, 27, 349-356. Weissman, A. N., & Beck, A. T. (1978, November). Development and validation of the Dysfunctional Attitude Scale. Paper presented at the annual meeting of the Association for Advancement of Behavior Therapy, Chicago. Winstead, B. A., & Cash, T. F. (1984). Reliability and validity of the Body Self-Relationship Questionnaire. Unpublished manuscript. Youden, W. J. (1950). Index for rating diagnostic tests. Cancer, 3, 3235.
Received June 8, 1992 Revision received May 13, 1993 Accepted May 26, 1993 •