Suicidal Children Grow Up: Suicidal Episodes and Effects ofTreatment during Follow-up
CYNTHIA R. PFEFFER, M.D., STEPHEN W. HURT, PH.D., TATSUYUKI KAKUMA, PH.D.,
JOAN R. PESKIN, M.A., CAROL A. SIEFKER, M.A., AND SRINIVAS NAGABHAIRAVA, M.A.
ABSTRACT
Objectives: This paper describes risk for first recurrent suicidal episodes in follow-up of suicidal child psychiatric
inpatients. It identifies relations between suicide attempts in follow-up and psychosocial and psychopharmacological
treatments. Methods: First suicidal episodes involving either suicidal ideation or a suicide attempt in a 6 to 8 year
follow-up period were rated for 69 child psychiatric inpatients and 64 children selected from the community. Psychiatric
treatments were determined from reports from multiple sources. Results: Forty-five percent of 133 subjects reported
a suicidal episode during follow-up. Children who reported suicidal ideation or a suicide attempt were greater than
twice as likely to report a suicidal episode in follow-up than were children from the community. Children treated with
antidepressants in follow-up were more likely to attempt suicide than were those not treated with antidepressants.
Conclusions: Close follow-up of suicidal children is warranted to identify risk and to intervene to prevent suicidal
episodes. Lack of efficacy of naturalistic treatments implies that controlled treatment studies are needed to determine
effective intervention for suicidal children. J. Am. Acad. Child Ado/esc. Psychiatry, 1994, 33, 2:225-230. Key Words:
childhood suicidal episodes, risk, treatment.
Suicide is relatively rare in children. In 1991, the agespecific rate of suicide in 5 to 14 year olds was 0.5per 100,000 population in the United States (NationalCenter for Health Statistics, 1992). However, suicideattempts among children are frequent as suggested byan age-specific rate of 14.4 suicide attempts for everysuicide in 10 to 11 year olds (Andrus et a1., 1991).Despite research identifying morbidity and mortalityresulting from suicidal acts in young children, thereare few studies of the longitudinal course of prepubertalchildren who are at risk for recurrent suicidal acts.
Accepted July 26, 1993.Dr. Pftffir is Professor of Psychiatry, Cornell University Medical College
and Chief Child Psychiatry Inpatient Unit, New York Hospital- WestchesterDivision; Dr. Hurt is Associate Professor of ClinicalPsychology in Psychiatry,Cornell University Medical College; Dr. Kakuma is Biostatistician, New YorkHospital- Westchester Division; Ms. Peskin, Ms. Siefker, and Mr. Nagabhairavaare Research Associates, New York Hospital-Westchester Division.
Thisstudy wassupported by USPHS GrantMH 142120 from the NationalInstitute ofMental Health, 1987-1990. Thispaper waspresented at the 39thAnnual MeetingoftheAmericanAcademy ofChild and Adolescent Psychiatry,October 21, 1992, in Washington, DC.
Correspondence to Dr. Pftffir, TheNew YorkHospital-Westchester Division,Cornell Medical Center, 21 Bloomingdale Road, White Plains, NY 10605.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:2 FEBRUARY 1994
Cohen-Sandler et a1. (1982) reported that 20% of20 prepubertal suicidal psychiatric inpatients thoughtabout or attempted suicide during an 18-month followup period. Kuperman et a1. (1988) used record linkagedata to identify a nine times higher rate of suicide in1,331 former child and adolescent psychiatric inpatientsthan among youth in the general population. At thetime of suicide, the former patients were at least 17years old. Pfeffer and colleagues (1988) reported therewas stability of suicidal ideation among four out of eightprepubertal suicidal children selected in a communitysample. These children had recurrent suicidal ideationwithin a 2-year follow-up period.
In a more recent phase of a longirudinal study ofprepubertal and young adolescent suicidal subjects,Pfeffer and colleagues (1993) reported that prepubertaland young adolescent psychiatric inpatient suicide attempters were six times more likely to report a suicideattempt during a 6 to 8 year follow-up than werethose children selected from a community sample.In addition, those child psychiatric inpatients whocontemplated suicide were more than three times atrisk for a future suicide attempt than children selectedfrom the community. Furthermore, half of 20 children
225
PFEFFER ET AL.
who attempted suicide during the follow-up reportedmultiple suicide attempts in the period (Pfeffer et al,1991). Other factors that increased risk for a suicideattempt during the follow-up were the presence of amood disorder, stressful life events, and poor socialadjustment (Pfeffer et al., 1993). The risk was elevatedthree and a half times in the follow-up period by thepresence of a mood disorder and four times by poorsocial adjustment within the year before a suicideattempt.
Other longitudinal studies noted the significance ofprepubertal onset mood disorders as important riskfactors for suicide attempts (Kovacs et al., 1993) andsuicide (Harrington et al., 1990; Rao et al., 1993).Kovacs et al. (1993) reported that a lifetime history ofa mood disorder in prepubertal psychiatric outpatientsenhanced risk by five times for attempting suicide ascompared to having a history of other psychiatricdisorders. Harrington and associates (1990) reportedthat 2.5% of80 children with major depressive disordercommitted suicide within an 18 year follow-up period.Rao and colleagues (1993) described the clinical characteristics of seven suicide victims in a 'sample of 427children and adolescents who were followed up in 10years. This represented a suicide rate of 4.4% over 10years. Seventy-one percent of these suicide victims hada history of major depressive disorder.
This paper will characterize the first recurrent suicidalepisode involving either suicidal ideation or a suicideattempt after an initial assessment of suicidal prepubertal and young adolescent psychiatric inpatients. Thisinformation is important in clarifying clinicians' expectations about the timing and form of a future suicidalstate in a child being evaluated for suicidal behavior.In addition, we will describe the relationship betweentreatment received and the occurrence of suicide attempts in a 6- to 8-year follow-up of child psychiatricpatients. These data have implications for understanding the efficacy of treatment for reducing risk ofrecurrent suicidal episodes in children who were psychiatric inpatients.
METHOD
SampleThe 133 subjects (Pfeffer er al., 1986, 1991, 1992, 1993) are
a subsample of 106 psychiatric inpatients and 101 nonpatientsinitially studied in 1979 to 1982 (Pfeffer et aI., 1986) who providedwritten informed consent to be reinterviewed at a 6- to 8-year
follow-up assessment. There were no significant differences inage, gender, social status (Hollingshead and Redlich, 1958), race/ethniciry, religion, suicidal ideation or attempts, or diagnoses forthese 133 subjects and the 74 subjects who were not reinterviewedat the 6- to 8-year follow-up.
At the initial assessment, the 133 subjects included 53 psychiatricinpatients who reported either suicidal ideation or a suicide attemptwithin 6 months of psychiatric hospitalization; 16 psychiatricinpatients without a history of suicidal ideation or attempts within6 months of hospitalization; and 64 nonpatients who were selectedfrom the general community to match the demographic characteristics of the inpatients. Nine (14.1%) nonpatients reported suicidalideation, and one (1.6%) made a suicide attempt within 6 monthsof the initial assessment.
Characteristics of the subjects have been described previously(Pfeffer 1991, 1992, 1993) and will be described briefly again.Demographic distributions were similar for the 53 suicidal patients,16 nonsuicidal patients, and 64 nonpatient controls. The 133subjects were predominantly white (72.2%), male (72.9%), middlesocial status (52.6%), and Catholic (60.2%). The mean age at theinitial assessment in 1979 to 1982 was 10.5 ± 1.8 years (range4.6 to 14.7 years). All 69 patients and 42 (65.6%) nonpatientshad a psychiatric disorder at the initial assessment. Among thesuicidal and nonsuicidal patients, the most prevalent DSM-IIIdisorders were disruptive and developmental disorders. Mood disorder was significantly more prevalent among the suicidal patients(56.6%) than the nonsuicidal patients (12.5%), Xl (1) = 9.6, P <.003. Among the nonpatients, the most prevalent DSM-IIIdisordersat the initial assessment were disruptive and anxiety disorders.
At the 6- to 8-year follow-up, the mean age of the 133 subjectswas 17.0 ± 2.2 years (range 10.9 to 21.3 years). No deaths werereported for the total sample of 207 children (Pfeffer et aI., 1992,1993). Twenty (15.0%) of the 133 subjects (16 [23.2%] patientsand four [6.3%] nonpatients) attempted suicide at least once duringthe follow-up period (Pfeffer et aI., 1991). In general, the patientshad a gradual increase in yearly rates of the first suicide attemptin the follow-up compared with a low rate of suicide attempts inthe period for the nonpatients (Pfeffer er aI., 1991). Fifty-one(96.2%) suicidal patients, 16 (100%) nonsuicidal patients, and 43(67.2%) nonpatients had at least one DSM-III-R psychiatric disorder at some time in the 6- to 8-year follow-up period (Pfeffer etaI., 1991). The most prevalent DSM-III-R psychiatric disorders inthe period for the patients and nonpatienrs were mood, disruptive,and anxiety disorders (Pfeffer et al., 1993).
Assessment Procedures
Research assessments (Pfeffer et aI., 1991, 1992, 1993) involvedseparate semistructured interviews of the subjects and their parentsconducted by trained master's or Ph.D. level psychologists whoachieved high interrater reliability.
The Spectrum of Suicidal Behavior Scale (Pfeffer, 1986) wasused to obtain information about suicidal ideation and acts at theinitial assessment (Pfeffer et aI., 1986). It also was used in thisfollow-up assessment (Pfeffer et al., 1991, 1992, 1993) to obtainretrospective information about the subjects' suicidal ideation andattempts in yearly intervals during the period. Suicidal ideationwas considered present in the follow-up period if it involved apersistent suicidal thought or occurred repeatedly within a yearlyinterval or if it involved a plan for a suicidal act. Final ratings forcurrent and retrospective reports of suicidal ideation or suicideattempts at different yearly intervals during the follow-up wereobtained by using the highest score on the Spectrum of Suicidal
226 j. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:2, FEBRUARY 1994
Behavior Scale for a given yearly interval reported by either thesubjects or parents.
The concept of a suicidal episode in the follow-up period wasoperationalized to be the presence of any suicidal tendency (suicidalideation or a suicide attempt) in a yearly interval of the followup period. The most severe suicidal state (that is, suicide attemptrather than suicidal ideation) of a suicidal episode was used in thedata analyses for a particular suicidal episode. Data analyses in thisreport involve the first suicidal episode in the 6- to 8-year followup for each subject. Interrater agreement measured with K (Cohen,1960) for a first suicidal episode in the follow-up period was highand significant, K = .83, P < .00001, N = 69 interviewer pairs.Parent-child concordance in identifying a first suicidal episode inthe 6- to 8-year follow-up period was adequate and significant,K = .48, P < .0001, N = 124 parent-child pairs.
Information about type, onset, and duration of psychiatrictreatment of the subjects during the follow-up was obtained frommultiple sources including the subjects, the parents, and recordsfrom treatment facilities and schools (Pfeffer et aI., 1992). Allinformation was reviewed separately by two research assistants, anda best-estimate rating was made about the treatment course foreach subject. An annual rate for each treatment received duringthe follow-up was calculated for each subject. For subjects whoreported a first suicide attempt during the follow-up, the annualrate of a specific treatment was calculated by dividing the totalnumber of years for a treatment that was received from the timeof the initial assessment until the first suicide attempt in the followup period by the total number of years of follow-up between theinitial assessment and the first suicide attempt in the follow-up.For subjects without a suicide attempt in the period, the annualrate of a specific treatment was the total number of years for agiven treatment in the follow-up divided by the total number ofyears in the period. In addition, annual rates for all treatmentscombined was calculated in a similar way to identify the totalamount of treatment received.
RESULTS
Suicidal Episodes during Follow-up
Sixty (45.1%) subjects (38[55.1%] patients and22[34.4%] nonpatients) reported at least one suicidalepisode in the follow-up, mean follow-up period =7.16 ± 1.0 years). The first suicidal episode after theinitial assessment of 52 (39.1%) subjects (32[46.4%]patients and 20[31.3%] nonpatients) involved suicidalideation. Eight (6.0%) subjects (6[8.7%] patients and2[3.1 %] nonpatients) reported that their first suicidalepisodes after the initial assessment involved a suicide attempt.
Table 1 indicates there is a gradient in rates for thefirst suicidal episodes in the follow-up period acrossthe four subject groups. The highest rates of suicidalepisodes were for the suicidal ideator patients andsuicide attempter patients who have similar rates butwhose rates are significantly higher than that of thenonpatients. Suicidal ideator patients have a 2.67 times
SUICIDAL CHILDREN GROW UP
greater risk for a recurrent suicidal episode than therisk that nonpatients have for a first suicidal episodein the follow-up.
.As shown in Figure 1, hazard rates calculated witha proportional hazard model (Kalbfleish and Prentice,1980) indicate that the rates for a first suicidal episodeduring the follow-up are not constant. The first suicidalepisodes for the suicidal patients who at the initialassessment reported either suicidal ideation or a suicideattempt and for the nonsuicidal patients were mostprevalent in the first 2 years of the follow-up period.In contrast, the nonpatients had a slow but gradualincrease in first suicidal episodes during the follow-up.
In part of Table 1 children are classified into nonsuicidal, suicide ideators, or suicide attempters at theinitial assessment and at the first suicidal episode inthe follow-up period. Among the 69 patients, 27(39.1%) had the same severity of suicidal episodes atboth times whereas for 42 (60.9%) patients the severityof suicidal episodes differed for the initial assessmentand the first suicidal episode in the follow-up. Ofthe 42 patients whose suicidal episodes changed, 33(78.6%) reported improvement in their suicidal severityand 9(21.4%) reported an increase in their suicidalseverity.Analysisof these data involved fitting models ofsymmetty, quasi-symmetry, and marginal homogeneity(Bishop et al., 1975) to examine a trend of suicidalseverity between the initial assessment and the firstsuicidal episode in the follow-up. The goodness-of-fittest based on the symmetty model, G2 (3) = 18.10,P = .0004, and the generalized McNemar test, X2 (1) =13.71, P = .005 (Bowker, 1948) indicate there aresignificant changes in the suicidal severity between theinitial assessment and the first suicidal episode in thefollow-up for individual patients. Furthermore, a goodfit of the quasi-symmetry model.C? (1)= .0039, P =
.950, and a significant result for the marginal homogeneity 0 (2) = 18.099, P = .0001, indicate that amongthe patients there was a significant difference betweenthe group distribution of suicidal episodes at the initialassessment (25 [36.2%] suicide attempter patients,28[40.6%] suicide ideator patients, 16[23.2%] nonsuicidal patients) and the group distribution of the firstsuicidal episodes in the follow-up (6.[8.7%] suicideattempters, 32(46.4%) suicide ideators, and 31 [44.9%]nonsuicidal subjects). This suggests that the patientsas a group report first suicidal episodes in the followup period that are less severe than those at the initial assessment.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:2, FEBRUARY 1994 227
PFEFFER ET AL.
TABLE 1Numbers and Rates of First Suicidal Episodes in Follow-up Period
Initial Subject Group
Number of Subjects in Rate ofFollow-up Who Are Suicidal Episode
Nonsuicidal Ideators Attempters (per 100 per Person Year)
a 10 suicidal nonpatients were combined with 54 nonsuicidal non patients because there was no difference in rates of first suicidalepisodes after the initial assessment for these groups (Wald test = 0.232, P < .816).
A series of Cox proportional hazard regression analyses (Kalbfleish and Prentice, 1980) were used to determine the associations between treatment services orpsychopharmacological treatments and the time to theoccurrence of first suicide attempt in the follow-upperiod. In these analyses, the four initial subject groups(suicide attempter patients, suicide ideator patients,nonsuicidal patients, and non patients) were stratifiedto control for nonproportional hazard rates of thesegroups of subjects. Twenty suicide attempts occurredin the 6- to 8-year follow-up (described in a previouspaper, Pfeffer et al., 1991). When the annual treatmentrate for the combined treatment services involvingpsychiatric hospitalization, residential treatment, dayhospitalization, and outpatient treatment in the followup period was evaluated as a covariate, the resultssuggested there was no significant relation between
Relation between Treatment and Suicidal Acts
suicidal episodes in the follow-up period, 42 (65.6%)nonsuicidal nonpatients, 20 (31.3%) suicidal ideatornonpatients, and 2 (3.1%) suicide attempter nonpatients. This was indicated by a good fit of the quasisymmetty model (0[1]=2.699, p = .10) and a trendfor a significant result for the marginal homogeneitymodel G (2) = 5.76, P = .056. The results of theseanalyses suggest that severity of suicidal episodes initially and in the follow-up period for specific nonpatients and for the nonpatients as a group are likely tobe different. It should be noted these analyses areexploratory as there are relatively few nonpatients whoreported suicidal ideation or attempts.
o
0.40
0.35
0.30
H 0.25AZ 0.20AR
0.150
0.10
0.05
0.000
Similar analyses (Bishop et al., 1975; Bowker, 1948)were used to compare the initial and the follow-upsuicidal episodes for the 64 nonpatients. The severityof suicidal episodes for the non patients was similar atthe initial assessment and in the follow-up period for35 (54.7%) nonpatients who were not suicidal andfor 2(3.1 %) who reported suicidal ideas. Furthermore,17 (26.6%) non patients had an increase, and 8(12.5%)had a decrease in their severity of suicidal episodes inthe follow-up period. A goodness-of-fit test based ona symmetty model (G2[3] =8.46, p = .04) and thegeneralized McNemar test (X: [1]=4.48, p = .04) suggested that the severity of suicidal episodes for eachnon patient initially and in the follow-up period was notsimilar. The group distribution of severity of suicidalepisodes for the nonpatients initially (54 [84.4%] nonsuicidal nonpatients, nine [14.1%] suicidal ideator nonpatients, one [1.5%] suicide attempter nonpatient) wasdifferent than the group distribution of severity of first
, 1 2 3 4 5 6 7 8 9 10YEARS FROM TIME1 TO SUICIDAL EPISODE
Fig. 1 Hazard for first suicidal episode for subject groups at time!.
228 ]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:2, FEBRUARY 1994
annual rate of combined treatment services and theoccurrence of a first suicide attempt in the period, X2
(1) = 1.71, P < .191. There was no significant relationbetween the time to occurrence of a first suicide attemptin the follow-up and any specific treatment servicesuch as psychiatric hospitalization (X2[l] = 1.47 P <
.225), residential treatment (X2[1] = 0.40, P < .529),day hospitalization (X2[1] = .32, P < .571), or outpatienttreatment (X2[l]=1.28, p < .258). There was a significant positive association between the time to the occurrence of a first suicide attempt during the followup and the rate of combined psychopharmacologicaltreatments, X2 (1) = 6.55, P < .011, relative risk(RR) 3.87, 95% confidence interval (CI) 1.37, 10.89.Specifically, this finding was indicative of a significantpositive association between the time to occurrence ofa suicide attempt in the follow-up period and theannual rate of antidepressant medication treatment, X2
(1) = l1.58,p< .0007, RR 17.03, 95% CI 3.33, 87.10.
DISCUSSION
This report of predominantly prepubertal psychiatricinpatients and a comparison group of non patientssuggests that patients with a history of suicidal ideationor a suicide attempt are at significant risk for a recurrentepisode of suicidal ideation or a suicide attempt, especially within 2 years of psychiatric hospitalization. Thisfinding concurs with those of a study of adolescentpsychiatric inpatient suicide attempters that highlightedthe serious risk for a suicide attempt shortly afterpsychiatric hospitalization. These adolescents reporteda significant number of recurrent suicide attemptswithin 6 months after hospital discharge (Brent etaI., 1993).
The present study suggested that the severity of thefirst suicidal episode in the follow-up period is notsimilar to that of the suicidal episode at the initialassessment. Although more than 50% of the suicideideator patients at the initial assessment were likely toreport a recurrent episode of suicidal ideation, only12% of the suicide attempter patients at the initialassessment reported a recurrent suicide attempt at thefirst suicidal episode in the follow-up. In general,the first recurrent suicidal episode for the group ofpsychiatric inpatients was less severe than were thesuicidal episodes at the initial assessment.
SUICIDAL CHILDREN GROW UP
Approximately 48% of the psychiatric patients reported a decrease in severity of first suicidal episodesduring the follow-up. This finding concurs with a 3year longitudinal study of 76 depressed child andadolescent psychiatric outpatients in which the severityof the first recurrent suicidal episode was less than theindex episode (Myers et aI., 1991). Myers and colleagues (1991) also reported on the episodic nature ofsuicidal ideation and attempts during follow-up. Theynoted that suicidal states "develop, resolve, and thendevelop again" (page 808).
The findings described in this paper must be integrated along with data regarding suicide attempts inthis 6- to 8-year follow-up period (Pfeffer et aI., 1991,1993). Suicide attempter patients were six times morelikely and suicide ideator patients were 3.6 times morelikely than non patients to report at least one suicideattempt during the 6- to 8-year follow-up period (Pfeffer et aI., 1993). Although the first suicidal episode inthe period for the patients was likely to be less severethan was the suicidal episode at the initial assessment,as a longer time elapsed in the follow-up, suicidalpatients reported suicide attempts during subsequentsuicidal episodes in the 6- to 8-year follow-up period.Furthermore, 50% of the 20 subjects who attemptedsuicide in the follow-up period made multiple suicideattempts in the 6- to 8-year follow-up (Pfeffer et al.,1991). These data emphasize that young children whoreport suicidal ideation or who attempt suicide arelikely to have multiple recurrences of suicidal ideas oracts. No deaths were reported among the 207 subjectswho were initially evaluated in 1979 to 1982 (Pfefferet aI., 1993). This is most likely related to the factthat the subjects have not traversed the high-risk periodfor suicidal death, which occurs in adolescents andyoung adults who are 15 to 24 years old (NationalCenter for Health Statistics, 1992). The mean age ofthe subjects at the 6- to 8-year follow-up assessmentwas 17 years, and the subjects also had a wide agespan that ranged from 10.9 to 21.3 years. Thus, theywere just entering the high-risk period for suicideamong youth.
This study suggested that treatment did not diminishrisk for a first suicide attempt in the follow-up period.In fact, 55% of the 20 subjects who reported a firstsuicide attempt in the follow-up period were in treatment at the time of this suicide attempt (Pfeffer et aI.,1992). These findings may be related to the naturalistic
j. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 33:2, FEBRUARY 1994 229
PFEFFER ET A L.
design of this study in which subjects were not assignedto treatment by the investigators. After discharge fromthe hospital at the initial assessment, the subjects received treatment from clinicians practicing in the commun ity. The treatments received in the follow-upperiod were general intervent ions offered in inpatient ,residential, day hospital, and outpatient facilities. Suchtreatments may not have been aimed specifically toreduce suicidal ideation or suicidal acts. The findingsregarding treatment in this study concur with thosereported by Brent and colleagues (1993) for 134 adolescent psychiatric inpatients. In that study, 13 (9.7%)adolescents attempted suicide within 6 months afterhospital discharge despite their receiving treatmentduring the follow-up period. The 13 adolescents whoattempted suicide in that 6-month follow-up weremore likely than were the nonattempters to be rehospitalized in the follow-up period. Similar to the studyof Brent and colleagues (1993), the findings of thepresent study suggest that subjects who attemptedsuicide in the follow-up period received longer treatment with antidepressants in the follow-up than didsubjects who did not attempt suicide within the period.Specifically, the findings indicated that subjects treatedfor longer times with ant idepressant medication, compared with those with shorter periods on ant idepressantmedication were 17 times more likely to attempt suicidein the follow-up period. This finding appears counterintuitive. However, it is most likely that treatment withantidepressant medications was employed for subjectswho exhibited severe depressive symptoms. It may beinferred that this lends support to findings of thisstudy suggesting that a mood disorder is a strong riskfactor for a suicide attempt (Pfeffer et aI., 1993).
Clinical and Research Implications
The findings of this study that prepubertal andyoung adolescent psychiatric patients who report suicidal ideation or suicide attempts are at significant riskfor recurrent suicidal ideation and acts implies thatclose monitoring, especially within the first 2 yearsafter hospital discharge, is warranted to ident ify riskfor suicidal episodes and to offer interventions toprotect against recurrent suicidal episodes as thesesuicidal children grow up.
Research that uses controlled treatment trials isneeded to identify efficacious treatments to preventrecurrent suicidal acts among youth. These interventions should target specific risk factors for suicidalbehavior such as adverse life events, poor social adjustment, and psychopathologies, especially mood andsubstance abuse disorders (Pfeffer et al., 1991, 1993).
REFERENCESAndrus JK, Fleming DW , Heum ann MA, Wassell JT, H opkins DD,
Gordon J (1991), Surveillance of attempted suicide among adolescentsin Oregon, 1988 . Am] Public Health 81:1067-1069
Bishop YMM, Feinberg SE, H olland PW (1975), Discrete M ultivariateAnalysis: Theory and Practice, Cam bridge, MA: MIT Press
Bowker AH (1948) . A test for symme try in contingency tables. Journalof the A merican Statistical Association 43:572-574
Brent DA, Kolko DJ, Wa rtella ME er aI. (1993), Adolescent psychia tricinpa tients ' risk of suicide attempt at 6-month follow-up . JAm AcadChild Adolesc Psychiatry 32 :95-105
Co hen J (1960), A coefficient of agreement for nominal scales. EducationalPsychology M easures 20:346-376
Cohea-Sandler R, Berma n AL, King RA (1982), A follow-up study ofhospitalized suicidal chi ldren . JAm Acad Child Psychiatry 2 1:398-403
H arrington R, Rudge H , Rutter M , Pickles A, Hill J (1990), Adult outcomesof childhood and adolescent dep ression. Arch Gen Psychiatry47 :465- 673
H ollingshead AB, Redlich F (195 8), Social Class and M ental Illness, NewYork: John Wi ley and Sons
Kalbfleisch JD , Prent ice RL (19 80), The Statistical Analysis ofFailure Tim eData. New York: Wiley and Sons
Kovacs M , Go ldston D, Ga tsonis C (1993), Suicidal behaviors and child hood-onset depressive disorders: a long itudi nal investigat ion . J AmAcad Child Adolesc Psychiatry 32:8- 20
Kuperman S, Black DW , Burns T L (1988) , Excess suicide among formerlyhospitalized child psychiatry patients. J Clin Psychiatry 49:88-92
Myers K, McCauley E, Calderon R, T reder R (1991), T he 3-year longirudinal course of suicidaliry and predictive factors for subsequent suicidaliryin youths with major depressive disorder. J Am Acad Child AdolescPsychiatry30:804-810
N ational Center for He alth Statistics (1992), Annual summary of births,marriages, divorces, and deaths: Un ited States, 199 1. M onthly VitalStatistics Report, Vol 40(13). Hyattsville, MD: Pub lic H ealth Service
Pfeffer CR (1986), The Suicidal Child. New York: Gu ilford Press IncPfeffer C R, Klerman GL, H urt SW , Kakuma T, Peskin JR, Siefker CA
(1993) , Suicidal ch'ildren grow up: rates and psychosocial risk factorsfor suicide att empts during follow-up. J Am Acad Child AdolescPsychiatry32 :106-113
Pfeffer C R, Klerman GL, H urt SW, Lesser M, Peskin JR, Siefker CA(1991), Suicidal children grow up: demographic and clin ical risk factorsfor adolescent suicide attempts . J Am Acad Child Adolesc Psychiatry30 :609--616
Pfeffer CR, Lipkins R, Plurchik R, Mi zruchi M (1988 ), No rmal childrenat risk for suicidal behavior : a two year follow-up study. JAm AcadChild Adolesc Psychiatry 27:34-41
Pfeffer CR, Peskin JR, Siefker CA (1992) , Suicidal children grow up:psychiatric treatment during follow-up period. J Am Acad of ChildAdolesc Psychiatry3 1:679- 685
Pfeffer C R, Plurchik R, Mizruchi MS, Lipkin s R (1986), Suicidal behaviorin child psychiatri c inpatients and outpat ients and in non parients , AmJ Psychiatry 143:733- 738
Rao U, Weissman MM, Martin JA, Hammond RW (1993) , Childhooddepression and risk of suicide: a preliminary report of a lon gitud inalstudy. J Am Acad Child Adolesc Psychiatry 32:21-27
230 ] , AM . ACA D . CHILD AD O LES C. PSYCHIATRY, 33:2 , FEB RUA RY 1994
