Study on suicide in depressed inpatients

Study on suicide in depressed inpatients

Journal ofAffectiveDisorders, 15 (1988) 157-162 Elsevier 157 JAD 00562 Study on suicide in depressed inpatients Jifi M o d e s t i n a n d W a l t ...

386KB Sizes 0 Downloads 99 Views

Journal ofAffectiveDisorders, 15 (1988) 157-162 Elsevier

157

JAD 00562

Study on suicide in depressed inpatients Jifi M o d e s t i n a n d W a l t e r K o p p Psychiatric University Clinic, Bolligenstrasse 111, 3072 Berne, Switzerland (Received 18 August 1987) (Revision received 7 March 1988) (Accepted 14 March 1988)

Summary Seventy-five depressed clinical suicides were compared with 50 depressed non-suicide clinical controls, using the method of stepwise logistic regression analysis. A set of the six best discriminating variables was identified, which comprised male sex, suicidal behavior at index admission and during index hospitalization, number of previous psychiatric hospitalizations, broken home and social exits. These factors should be taken into consideration when estimating suicidal danger in all types of depressed inpatients.

Key words: Suicide; Depression; Discriminating variables

Introduction Some 30-70% of suicides in the general population (Dorpat and Ripley, 1960; Barraclough et al., 1974; Robins, 1981) and a substantial proportion of psychiatric suicides (Roy, 1982) suffer from depression. The ultimate risk of suicide for patients with primary affective disorder amounts to 15% (Guze and Robins, 1970), and as many as 37% of patients suffering from affective disorder who committed suicide did so during their hospital care (Weeke, 1979). Patients with affective disorders are overrepresented among hospital suicides (Farberow et al., 1966; Gale et al., 1980), the suicide risk being greater for affective disorders

Address for correspondence: Dr. J. Modestin, Psychiatric University Clinic, Bolligenstrasse 111, 3072 Berne, Switzerland.

than for other diagnostic groups at each point in time of a patient's hospital stay (Copas and Robin, 1982). Moreover, an increase in the suicide rate of psychiatric inpatients has been observed (Modestin, 1982), particularly affecting those suffering from affective disorders (Barner-Rasmussen et al., 1986). However, there are only a few controlled studies devoted to the completed suicide of depressed patients in general and of depressed inpatients in particular. Compared with depressed controls, depressed suicides - in individual samples differently defined as to their patient status - were found to be more severely depressed (Barraclough et al., 1974) and to suffer more from a delusional type of depression (Roose et al., 1983). Significantly more depressed suicides were of the male sex, unmarried, living alone and unemployed. They were more suicidal in the last admission, and in

0165-0327/88/$03.50 © 1988 Elsevier Science Publishers B.V, (Biomedical Division)

158 their past history they had attempted suicide up to ten times more frequently (Barraclough et al., 1974; Barraclough and Pallis, 1975; Roy, 1983, 1984). We report the results of a study of completed suicide in depressed patients, which comprised a sufficiently large number of suicides, included a control group, and was performed exclusively on inpatients.

Methods

A total of 149 inpatients who had committed suicide were identified by examining the police files (the most complete registration of suicides in Switzerland). Their suicide had occurred while they were on the hospital rolls of two Swiss psychiatric institutions in the years 1960-1981, irrespective of whether the act was committed in the hospital itself, during a short-term permitted leave or an escape. An equal number of 149 clinical non-suicides served as control group. Even the dependency of suicide frequency upon such variables as sex and age has not been demonstrated reliably in the literature using adequately large samples of exclusively inpatient suicides (Beisser and Blanchette, 1961; Kern, 1973; Gale et al., 1980). The control patients were therefore matched with the suicides for date of index admission and inpatient status but otherwise were chosen by a random procedure. Compared with a random sample of 1100 admissions that occurred in the same period (Koch, 1986), our control group proved to be representative of the entire admitted inpatient population. Clinical charts of all these inpatients were thoroughly scrutinized by one of us (J.M.) with regard to the presence of all phenomena listed in the Research Diagnostic Criteria ( R D C ) for a selected group of functional disorders (Spitzer et al., 1980) and all patients were given an R D C diagnosis (Modestin, 1984). A total of 75 (50.3%) suicides and 50 (34.6%) controls suffered from a depression (X 2 = 7.94; P < 0.01). The depressed suicides and the depressed controls were compared with each other for a larger set of the potential risk variables using the statistical method of stepwise logistic regression analysis.

The following demographic and psychosocial variables were investigated using the method of a retrospective analysis of the hospital clinical records: sex, age, marital status, existence of children, foreign-born status, religion, place of residence, educational and professional level, social class of patient and his family of origin, intergenerational social mobility, broken home before 16 years of age, early social adjustment before 25 years of age (judged as disturbed in case of delinquency a n d / o r vocational difficulties a n d / o r disability in regard to reliable, lasting relationships a n d / o r difficulties in separating from the family of origin), living and vocational situation at the time of the index admission, chronic disability, placement under tutelage and life events in the last year. The psychiatric and medical variables examined included the following: diagnosis, age on occurrence of first symptoms, duration of psychiatric illness, number and duration of psychiatric hospitalizations, length of last (index) hospitalization, psychiatric disorders and alcoholism in first-degree relatives and suicide in the family, previous suicide attempts and their seriousness, reason for index admission, suicidal behavior immediately before and during the index hospitalization, aggressive behavior in the year preceding suicide and physical problems at index hospitalization. Unless defined differently, the situation of the patient at the time of the suicide or discharge served as the point of reference in the evaluation of the individual variables. The median of the length of the index hospitalization was 9 weeks in the suicide and 5 weeks in the control group. Neither group differed significantly with regard to the length of the index hospital stay (median test). All variables investigated were either primarily clearly determinable (such as sex and age) or an attempt was made to define and operationalize them as exactly as possible (e.g., the patient was judged chronically disabled if he had spent more than half of the previous 5 years in some institution a n d / o r was unable to work). Social class was determined using the classification by Moore and Kleining (1960). Suicide attempts were evaluated using the classification by Motto (1965). Life events were investigated and classified according to Paykel et al. (1971, 1975). In many cases our

159

clinical charts included psychiatric expert opinions and clinical records from other psychiatric institutions. Occasionally, the family of the patient was contacted to obtain the missing data. Thus, full information was available in the vast majority of cases (90.4%), the missing data concerning mainly two variables (social class of the family of origin of the patient and his intergenerational mobility). Stepwise logistic regression analysis was used in order to sort out variables with optimal discrimination. It is a statistical multivariate method suited to the analysis of the mutual relations between categorical variables when the dependent variable is a dichotomous or ordinal-scaled one and the independent/explaining variables are categorical a n d / o r numerical. The purpose of the logistic regression analysis is to bring out the essential part of the information obtained from the numerous cross-tabulations of two variables (SAS, 1986; Walker and Duncan, 1967). The logistic regression analysis does not take into account the interaction of more than two variables; nevertheless, the connection between a factor and the response may be different at the different levels of another factor (intervening third variable). Therefore, using the log-linear model in analogy to a variance analytical procedure, the mutual interactions of the most significant variables yielded by the logistic regression analysis and appearing in the model were investigated. No such interactions have been found whereas the significances of the main effects were confirmed. Results

Table 1 presents the distribution of the suicides and controls by R D C diagnoses. There is a significant difference between the groups regarding diagnostic distribution, patients with bipolar affective and schizoaffective disorders being overrepresented and those with minor depressive disorder underrepresented in the suicide group. The stepwise logistic regression analysis yielded a model of the six (out of the total of 57) best discriminating variables, the criterion for the inclusion in the model of each variable being P < 0.05. These variables are presented in Table 2, along with their frequencies in both groups and

TABLE 1 THE DISTRIBUTION OF THE S U I C I D E A N D C O N T R O L GROUPS BY RDC D I A G N O S I S Precentages are given in parentheses. Depressed clinical suicides

Depressed clinical controls

(n = 75)

(n = 50)

Major depressive disorder Minor depressive disorder Intermittent depressive disorder Bipolar disorder with mania/hypomania Schizoaffective disorder, depressed type

34 (45) 11 (15)

22 (44) 20 (40)

6 (8)

4 (8)

15 (20)

3 (6)

9 (12)

1 (2)

Significance

X2 =15,61; d f = 4 ; P = 0.004

the corresponding X 2 values, P values and r values, the latter representing the empirical measure of the discriminatory power of the variable. With the help of the model the patients of both groups could be correctly identified in 76% of cases (sensitivity 79%, specificity 72%) using the significance level of P < 0.05. The residual X 2 for all variables not included in the model (which should not be significant) was not significant (X 2 = 42.70; dr= 51; P = 0.79).

TABLE 2 VARIABLES Y I E L D E D BY STEPWISE LOGISTIC REGRESSION ANALYSIS AS BEST D I S C R I M I N A T I N G BETWEEN THE CLINICAL DEPRESSED SUICIDES A N D CLINICAL DEPRESSED CONTROLS Percentages are given in parentheses. Variable

Suicides

Controls X2

P

r

(n=75) (n=50) Suicidal behavior during the index hospitalization Sex (men) Number of previous hospitalizations (mean+SD) Suicide attempt at index admission Broken home Social exits

42 (56) 52 (69)

6.0+6.6 21 (28) 25 (35) 27 (36)

10 (20) 17 (34)

13.40 < 0.001 0.26 16.40 < 0.001 0.29

2.7_+2.2 10.09 8 (16) 9 (18) 12 (24)

4.52 5.43 3.88

0.002 0.22 0.034 0.12 0.020 0.14 0.049 0.11

160 TABLE 3 FREQUENCY OF SUICIDAL BEHAVIOR IN T H E SUICIDE AND CONTROL GROUPS AT INDEX ADMISSION AND DURING INDEX HOSPITALIZATION P e r c e n t a g e s a r e given in p a r e n t h e s e s . Suicidal b e h a v ior at i n d e x adrnission

Suicidal b e h a v ior d u r i n g i n d e x hospitalization

Significanoe

Depressed clinical suicides ( n = 75) Depressed clinical controls ( n ~ 50)

47 (63)

42 (56)

n.s.

28 (56)

10 (20)

X 2 = 12.17 P < 0.001

Significance

n.s.

X z ~ 14.56 P < 0.001

Suicidal behavior (suicidal thoughts and threats, suicide attempts) during the index hospitalization was the first variable sorted out in the course of the analysis. Table 3 shows the frequency of suicidal behavior in both groups of patients at the time of their index admission and during their index hospital stay. Discussion

The retrospective evaluation of clinical records is always fraught with inherent problems of lack of data a n d / o r their questionable quality. On the other hand, clinical notes can give information of sufficient reliability and validity if they are properly recorded (Csernansky et al., 1983). This was generally the case with our charts and we consequently looked for missing data. The allocation of patients to suicide and control groups occurred independently of the recording of the data and we have no reason for considering the quality of the data systematically worse in one group than in another. The evaluation of all case records occurred in a random order by the same investigator who, however, was not evaluating the charts 'blind'. As ' h a r d ' data were examined for the most part we do not believe his possible bias would have influenced the results in a significant way.

In our investigation, the significant overrepresentation of depressed patients among clinical suicides was confirmed. Also, a significant overrepresentation of the endogenous/psychotic (bipolar and schizoaffective) type of depression among the depressed inpatients suicides was found, even though the P value of 0.004 will be exaggerated due to the low expected frequencies in 20% of the cells of Table 1. The greatest suicide risk has been claimed for reactive depressive psychosis (Qdegard, 1967), depressive neurosis (Sletten et al., 1972) and psychogenic/reactive depression (Ciompi, 1976). This is at variance with our finding. However, it is not surprising that a substantial proportion of those who commit suicide in spite of the help being offered to them should suffer from a psychotic type of depression. Although of statistical significance in the univariate evaluation, the diagnosis of a depressive subtype does not appear as a variable in the model resulting from multivariate analysis. On the one hand, the diagnosis of depressive subtype does not belong to the best discriminating variables: on the other, those variables included in the model discriminate between the suicides and controls disregarding the type of depressive illness the patients suffered from. In our investigation recent suicidal behavior, male sex, frequent previous psychiatric hospitalizations, situation of early broken home, and social exits proved to be the best discriminating variables between depressed clinical suicides and controls, allowing the correct classification of 76% of the combined sample. On the one hand, since we used 125 subjects and 57 predictor variables in the logistic regression analysis, the degree of separation of the cases which was achieved might easily be exaggerated; the reliability of the method decreases in dependence on the number of variables investigated relative to the sample size. On the other hand, applying the log-linear model to the analysis of the data, the discriminatory importance of the three most distinguishing variables mentioned above could be confirmed. The suicidal risk is especially great in patients with a diagnosis of depression and a history of suicide attempts (Avery and Winokur, 1978). Depressed suicides studied by Roy (1983) and Vogel and Wolfersdorf (1985) were more frequently ad-

161

mitted in their last period of illness because of suicidal behavior. In our study there were no differences between clinical suicides and controls in the frequency of suicidal manifestations (including suicidal thoughts, threats and attempts) at index admission. However, while the frequency of suicidal manifestations remained high in the course of the index hospitalization in the group of suicides, it diminished significantly in the group of depressed controls, so that the difference between both groups became highly significant in the course of hospital treatment. Thus, in the majority of depressed inpatients the suicide did not occur unexpectedly, but was preceded by continuous suicidal manifestations. Therefore, it was probably not the recognition of suicidal danger, but rather the execution of effective therapeutic measures, which failed. An overrepresentation of the male sex was found among depressed suicides in the general population (Barraclough and Pailis, 1975), in a population of depressed hospitalized and discharged psychiatric patients (Dingman and McGlashan, 1986) and it was confirmed for the depressed suicides recruited exclusively from psychiatric inpatients by the present study. Significantly more psychiatric admissions were noted in depressed suicides by Roy (1983) and in hospital suicides by Schlosser and Strehle-Jung (1982). In our study the average number of previous hospitalizations was twice as high in clinical suicides as in clinical controls, whereas no differences were found in the duration of illness. In view of this finding the course of the illness of our suicides is assumed to have been more deleterious. Suicide occurred significantly more often in those discharged inpatients who were depressed and experienced adverse life events than in those who rated positively on only one of these variables (Pokorny and Kaplan, 1976). The significance of a broken home for completed suicide was pointed out by Roy (1984) for manic-depressive but not neurotic depressed suicides (Roy, 1983). A correlation between recent disruption of close interpersonal relations and early important interpersonal losses has been demonstrated in suicide attempters (Greer et al., 1966; Levi et al., 1966) but not confirmed in completed suicide of mental hospital patients (Farberow and Reynolds, 1971). Both

variables, broken home before 16 years of age and social exits in the year preceding suicide, appeared in our model, even though they discriminated less well between the groups (lower r). Our results indicate that both factors may have been of importance regarding depressed clinical suicides. In correspondence with our results, previous psychiatric hospital admissions, suicidal impulses and special stress were found to be predictors most significantly indicating suicidal outcome within 2 years in a population of patients hospitalized due to a depressive or suicidal state (Motto et al., 1985). In contrast to these authors, it is not our intent to present a new suicide prediction scale. We agree with Pokorny (1983) that all such attempts will miss many cases and identify too many false-positive cases to be workable. This should not create a sense of nihilism in the clinician, but rather reaffirm the importance of the individualized clinical assessment of every patient (Goldney et al., 1985). Nevetheless, there are some areas which always have to be addressed, and the results presented here may help in defining them in the population of depressed inpatients, keeping in mind that the absence of one or another of the risk indicators does not necessarily diminish or exclude suicide risk in the individual patient (Motto, 1985). These areas encompass factors closely related to the depressive illness (suicidal manifestations) and its course (previous hospitalizations) but probably also factors of a psychological nature (broken home and social exits) independent of the affective disorder. Thus, suicide in a depressed inpatient appears to be a complex phenomenon defying a simple linear explanation. References Avery, D. and Winokur, G. (1978) Suicide, attempted suicide, and relapse rate in depression. Arch. Gen. Psychiatry 35, 749 753. Barner-Rasmussen, P., Dupont, A. and Bille, H. (1986) Suicide in psychiatric patients in Denmark, 1971 1981. Demographic and diagnostic description. Acta Psychiatr. Scand. 73, 441-448. Barraclough, B.M. and Pallis, D.J. (1975) Depression followed by suicide: a comparison of depressed suicides with living depressives. Psychol. Med. (London) 5, 55-61. Barraclough, B., Bunch, J., Nelson, B. and Sainsbury, P. (1974) A hundred cases of suicide: clinical aspects. Br. J. Psychiatry 125, 355-373.

162 Beisser, A.R. and Blanchette, J.E. (1961) A study of suicides in a mental hospital. Dis. Nerv. Syst. 22, 365 369. Ciompi, L. (1976) Late suicide in former mental patients. Psychiatr. Clin. 9, 59 63. Copas, J.B. and Robin, A., (1982) Suicide in psychiatric in-patients. Br. J. Psychiatry 141, 503-511. Csernansky, J.G., Yesavage, J.A., Maloney, W. and Kaplan, J. (1983) The treatment response scale: a retrospective method of assessing response to neuroleptics. Am. J. Psychiatry 140, 1210-1213. Dingman, C.W. and McGlashan, T.H. (1986) Discriminating characteristics of suicides. Acta Psychiatr. Stand. 74, 91 97. Dorpat, T.L. and Ripley, H.S. (1960) A study of suicide in the Seattle area. Compr. Psychiatry 1, 349-359. Farberow, N.L. and Reynolds, D.K. (1971) Dyadic crisis suicides in mental hospital patients. J. Abnorm. Psychol. 78, 77 85. Farberow, N.L., Shneidman, E.S. and Neuringer, C. (1966) Case history and hospitalization factors in suicides of neuropsychiatric hospital patients. J. Nerv. MenL Dis. I42, 32-44. Gale, S.W., Mesnikoff, A., Fine, J. and Talbot, J.A. (1980) A study of suicide in state mental hospitals in New York City. Psychiatr. Q. 52, 201-213. Goldney, R.D., Positano, S., Spence, N.D. and Rosenman, S.J. (1985) Suicide in association with psychiatric hospitalization. Aust. N . Z . J . Psychiatry 19, 177 183. Greer, S., G u n n , J.C. and Koller, K.M. (1966) Aetiological factors in attempted suicide. Br. Med. J. 2, 1352 1355. Guze, S.B. and Robins, E. (1970) Suicide and primary affective disorders. Br. J. Psychiatry 117, 437 438. Kern, R. (1973) Die 100 Suizidf:alle der Psychiatrischen Universit~.tsklinik Ztirich von 1900 bis 1972. Dissertation, Medizinische FakuhRt der Universit~it Ziirich. Koch, U. (1986) Suizid in der psychiatrischen Klinik: Line vergleichende Studie anhand sozio-dem,ographischer und diagnostischer Kriterien. Dissertation, Medizinische Fakult~t der Universit5t Bern. Levi, L.D., Fales, C.H., Stein, M. and Sharp, V.H. (1066) Separation and attempted suicide. Arch. Gen. Psychiatry 15, 158-164. Modestin, J. (1982) Suizid in der psychiatrischen Institution. Nervenarzt 53, 254-261. Modestin, J. (1984) Clinical diagnostic practice reviewed. A comparison of clinical and R D C diagnoses. Psychopathology 17, 80 89. Moore, H. and Kleining, G. (1960) Das soziale Selbstbild der Gesellschaftsschichten in Deutschland. K{51ner Ztschr. Soziol. Sozialpsychiat. 12, 86-119. Motto, J.A. (1965) Suicide attempts. Arch. Gen. Psychiatry 13, 516-520. Motto, J.A. (1985) Paradoxes of suicide risk assessment. Hillside J. Clin. Psychiatry 7, 109-119.

Motto, J.A., Heilbron, D.C. and Juster, R.P. (1985) Development of a clinical instrument to estimate suicide risk. Am. J. Psychiatry 142, 6g0-686. Odegard, O. (1967) Mortality in Norwegian psychiatric hospitals 1950-1962. Acta Genet. Star. Med. 17, 137-153. Paykel, E.S., Prusoff, B.A. and Uhlenhuth, E.G. (1971) Scaling of life events. Arch. Gen. Psychiatry 25, 340 347. Paykel, E.S., Prusoff, B.A. and Myers, J.K. (1975) Suicide attempts and recent life events. Arch. Gen. Psychiatry 32, 327 333. Pokorny, A.D. (1983) Prediction of suicide in psychiatric patients. Arch. Gen. Psychiatry 40, 249-257. Pokorny, A.D. and Kaplan, H.B. (1976) Suicide following psychiatric hospitalization. J. Nerv. Ment. Dis. 162, 119 125. Robins, E. (1981) The Final Months - Study of the Lives of 134 Persons who Committed Suicide. Oxford University Press, New York-Oxford. Roose, S.P., Glassman, A.H., Walsh, B.T., Woodring, S. and Vital-Herne, J. (1983) Depression, delusions, and suicide. Am. J. Psychiatry 140, 1159-1162. Roy, A. (1982) Risk factors for suicide in psychiatric patients. Arch. Gen. Psychiatry 39, 1089-1095. Roy, A. (1983) Suicide in depressives. Compr. Psychiatry 24, 487-491. Roy, A. (1984) Suicide in recurrent affective disorder patients. Can. J. Psychiatry 29, 319-322. SAS (Statistical Analysis System) (1986) SUGI Supplemental Library User's Guide, Version 5 Edn. SAS, Cary, NC, pp. 269-293. Schlosser, J. and Strehle-Jung, G. (1982) Suizide wfihrend psychiatrischer Klinikbehandlung. Psychiatr. Prax. 9, 20-26. Sletten, I.W., Brown, M.L., Evenson, R.C. and Altman, H. (1972) Suicide in mental hospital patients. Dis. Nerv. Syst. 33, 328-334. Spitzer, R., Endicott, J. and Robins, E. (1980) Research Diagnostic Criteria for a selected group of functional disorders, 3rd edn. New York State Psychiatric Institute, New York, NY. Vogel, R. and Wolfersdorf, M. (1985) M~Sglichkeiten zur Vorhersage suizidaler Verhaltensweisen bei stationfir behandelten depressiven Patienten. In: M. Wolfersdorf, R. Wohlt and G. Hole (Eds.), Depressionsstationen. Roderer, Regensburg, pp. 176.- 189. Walker, S.H. and Duncan, D.B. (1967) Estimation of the probability of an event as a function of several independent variables. Biometrika 54, 167-179. Weeke. A. (1979) Causes of death in manic-depressives. In: M. Schou and E. Str,Smgren (Eds.), Origin, Prevention and Treatment of Affective Disorders. Academic Press, New York, NY, pp. 289 299.