British Journal of Psychiatry (1992),161, 749—758

Suicide in Psychiatric

Review Article

Patients:

Risk and Prevention

LOUIS APPLEBY

The preventionof suiciderequiresan understandingof protectiveas well as riskfactors, and the recognitionof high-riskgroups.Factorswhichincreasethe riskinpatientpopulationsinclude previous parasuicide, recent relapse or discharge, features of mental state (depression, psychosis),socialcircumstances(isolation,unemployment),anddemographiccharacteristics (malesex, youngage). Protectivefactorsareunder-researched but are likelyto lieinthe nature of psychiatriccare. Consequently,communitycare may affect suicideby alteringthe level of protectionat criticalperiodsin an episodeof illness.The clinicalpreventionof suicideshould thereforebe a priorityfor communityservices,andthe relationshipbetweensuicideandmental health care shouldbe researchedby a national processof monitoring.

Suicide is the most important consequence of psychiatric disorder, and most major psychiatric disorders carry a high suicide risk. This review of suicide by current and recent psychiatric patients examines both risk factors and protective factors, in particular those which are directly related to the nature and delivery of psychiatric care. The possible effect of community-based treatment on suicide is discussed. Finally, clinical guidelines on preventing suicide in psychiatric patients are presented, drawn from research findings, and future research priorities are proposed.

factors: they are circumstances which, in the presence of considerable risk, act preventively without altering the risk factors themselves. What protective factors are known? Relatively few studies have investigated the inverted question of what stops individuals from killing themselves, counterbalancing an otherwise substantial risk. Linehan et a! (1983) compared psychiatric in-patients who had thought about suicide with others who had actually carried out a parasuicidal act on a ‘¿reasons for living' inventory, composed after surveying a ‘¿street sample' of non-psychiatric subjects. The latter confirmed a common clinical impression that concern for children, religion, and fear of pain were the commonest reasons for not attempting suicide. However, the two psychiatric groups differed only on child concerns. Appleby (1991) obtained a standardised mortality ratio (i.e. the ratio of deaths observed to deaths expected calculated from age-specific death rates) for suicide by women in their first post-natal year and, despite the high rate of psychiatric disorder

Risk factors and protective factors Epidemiological research has identified many factors that increase the risk of suicide in non-psychiatric patients. Although the way in which such factors interact is not clear, they provide the basis of one possible model of suicide prevention, one which is

population-based. Such a strategy aims to reduce the prevalence of a risk factor in the whole population, thus disproportionately reducing the numbers who are at highest risk of its consequences. A population approach to suicide prevention might aim to reduce the rate of depression, for example by a programme of public education like the one developed by the National Institute of Mental Health (Regier et al, 1988). But the effectiveness of

such a programme would depend on how direct the relationship between depression and suicide was. An alternative would be to focus primarily on high-risk subjects. The potential difficulties are (a) that high-risk subjects must be accurately identified, and (1,) that the factors that prevent suicide in these subjects must be understood before they can be used to combat high risk. Thus, protective factors are not simply the mirror image of risk

at this time (Kumar & Robson, 1984; Kendell et a!, 1987), found the rate of suicide to be only one-sixth

of the expected rate. It was concluded that child concerns are a powerful influence on suicide risk even in a high-risk population;

that is, they are an

important protective factor. Psychiatric patients are the group above all others at high risk of suicide. In a study by Allebeck & Ailgulander (1990), a psychiatric diagnosis in in-patient care was the strongest predictor of suicide over 13 years in a large cohort of Swedish conscripts, producing an odds ratio of 11.3. Although many suicides have no such clear history, there is evidence that most, perhaps all, have had some psychiatric disturbance (Robins et al, 1959a; Barraclough et a!, 1974).

749

750

APPLEBY

Several studies have described suicides among current and former psychiatric in-patients, but relatively few have used a case-control design. Fewer still have looked at the nature of psychiatric care, although it may be the most important of all protective factors —¿ firstly, because it acts directly to counter the specific risk of psychiatric illness; secondly, because it is open to preventive intervention; thirdly, because the recent shift in psychiatric care towards community-based treatment could affect the risk of suicide in major mental illness.

by suicide, as had 2.7% of depressive patients. However, this method has the disadvantage that suicide rates are far from uniform over time. Some studies have concentrated on specific disorders. Guze & Robins (1970) reviewed earlier estimates of the proportion of suicides among the deaths of those with primary affective disorder, quoting a range of l2—60°lo over varying follow-up periods. They concluded that about 15% of patients with primary affective disorder would ultimately kill themselves, but that the highest risk (relative to deaths from all causes) occurred early. A more recent report on in-patient depressives followed up for 14—27years found that 8.5% had died by suicide How high is the risk? (Berglund & Nilsson, 1987). Quarter of a century ago, Temoche et a! (1964) cal Studies based on in-patients are likely to produce culated a mortality ratio for suicide in Massachusetts higher than expected figures for conditions in which inthesixmonthsafterdischarge from in-patientadmission is reserved for the most severe cases. This care, standardised against the general population. was true of Sims & Prior's (1978) study of mortality Over three years, the risk for all patients was in neurotic in-patients followed up for a mean of increased 34-fold. This figure is close to that obtained 11 years: 9.4% had committed suicide, a figure as by King & Barraclough (1990) in Southampton. They high as in major mental illness. The same can be said recorded all deaths from a single catchment area of Berglund's (1984) study of in-patient alcoholics which could have been suicides, and calculated followed up for 10-30 years, in which 6.7% had died the relative risk of mortality among those who by suicide (16% of alldeaths). had recently been psychiatric patients. In an eight The most watertight calculation of the suicide rate year sample totalling 412 deaths, 56% had had a in schizophrenia was that of Allebeck & Wistedt psychiatric referral and 37% had had a psychiatric (1986), who were able to carry out a linkage of the admission at some time. The crude relative risk of Stockholm County in-patient register to the Swedish death among all those with a psychiatric contact in national ‘¿cause of death' register. Of 1190 schizo the previous year was 26. Among those aged 35—44 phrenic patients discharged in a single year and years, the relative risk was 39, while for subjects aged followed up for 10 years, 3.9% were officially over 75 years the corresponding figure was 4. recorded as having died by suicide or an undeter Rorsmann (1973) studied all patients attending mined cause. The standardised mortality ratio was a psychiatric department in Lund, Sweden, during raised in both sexes, at 9.9 in men and 17.5 in 1962, and discovered that 1.4% committed suicide women, suggesting that the male excess among during the next five years. Among men the rate general population suicides may be less evident was 2.2%; among women it was 0.8%. Martin in psychiatric populations, although there is still et a! (1985), studying out-patients only, calculated a larger proportion of males in most psychiatric the likelihood of ‘¿unnatural mortality' —¿ mainly suicide samples. suicide but including homicide - during a 6-12-year follow-up of 500 subjects. The rate they found was over three times the crude expected rate. Predictors Which psychiatric patients commit suicide? of unnatural death were younger age and a diagnosis of alcoholism, drug addiction, antisocial personality, Only those studies which include a comparison or secondary affective disorder. sample of psychiatric patients who do not kill them The overall risk of suicide clearly depends on selves can accurately identify social and clinical length of follow-up. To take account of this, characteristics of patient suicide. Such studies Pokorny (1966) calculated the rate of suicide based have usually looked back over several years, using on person-years at risk to be 165 per 100 000 people whatever clinical data have been routinely collected, per year, 16 times the general population rate. By generally without testing any specific hypothesis. far the highest rate was found in depressed patients, Table I summarises their main findings. whose rate was over 50 times that of the general A history of parasuicide is a risk factor in population. Put another way, l% of all patients seen most studies, occurring in around 50% of cases by the service in the previous 15 years had died (Rorsmann, 1973; Myers & Neal, 1978; Roy,

SUICIDE RISK AND PREVENTION

751

Table 1 Clinical and social characteristics StudySelected associationssampleFlood

diagnosisSize

of psychiatric patients who kill themselves

of suicidePositive

& Seager(1968)All97Psychotic classRobin eta! (1968)API33Depression, (males)Rorsmann (1973)All45Affective (females)Myers & Neal (1978)All100Parasuicide, single)Roy (1982a)All90Schizophrenia, aloneRoy unemploymentWilkinson (1982b)Schizophrenia30Past & Bacon (1984)Schizophrenic parasuicides17Allebeck eta! (1987)Schizophrenia32Alcoholism, aloneDrake (1986)Schizophrenia15HopelessnessCohen & Cotton et a! (1990)Schizophrenia8Past paranoiaFawcett et a! (1990)Affective disorders32Suicide hopelessnessBerglund (1984)Alcoholism88Past

l982a,b).

In

addition,

Flood

&

Seager

(1968)

found that the last admissions of suicides were twice as likely to have followed a parasuicide. In

the study of Rorsmann (1973) in Lund, parasuicide was estimated to increase approximately 10-fold the risk of suicide, although after one parasuicide further episodes did not increase the likelihood of suicide. Among women, there was a tendency for dangerous and violent methods of self-harm to be more common in those who later completed suicide, a conclusion supported by Myers & Neal (1978) after studying all suicides in Shropshire over nine years - in 44% of previous parasuicides, the method used had been violent or particularly dangerous. Uncontrolled studies can offer only descriptive information but can reveal patterns relevant to the preventive role of mental health services. Goh et a!

(1989) analysed 57 suicides by current in-patients and day patients. In only 12% had suicidal ideas been the reason for last admission, although suicidal thoughts had been recorded in 54%; 40% had been described as improving before the suicide took place. Langley

& Bayatti (1984) found that in-patient suicides were most likely to be schizophrenic patients in their 20s ordepressed patients intheir 40s.A study infive New York hospitals related suicide to the same diagnoses and to involuntary status (Gale et a!, 1980). Schizophrenia Some controlled studies have looked specifically at schizophrenia, which Allebeck & Allgulander's

depression,parasuicide,widowhood, low social alcoholism, parasuicide,poor work record disorder (males),parasuicide,divorce (females),living alone unmarried(divorced,widowed, or manic—depressive illness, parasuicide,unmarried (divorced, widowed, or single), living depression,

parasuicide,single (women), living affective or schizoaffective disorder, depressedmood, hopelessness,obsessions,hostility, before 1 year: anxiety, panic, insomnia, anhedonia,loss of concentration, alcohol abuse Suicide after 1 year: parasuicide, depression

(1990) longitudinal study of conscripts found to be the diagnosis with the highest risk. In most, social risk factors have not been found once diagnosis has been controlled and most associations have been with features of mental state (Table 1). In Roy's (1982b) study, the suicides were more likely to have a history of significant depression, and the reason for their most recent admission was more often depression or suicidal ideas. Two subjects had killed themselves in response to hallucinations.

Similarly, Cohen et a! (1990), whose small sample was entirely male, found a link with certain symptoms (Table 1) and with a secondary diagnosis of affective disorder on entry to the study or shortly before death. Drake et a! (1984), having related increased risk to suicidal ideas and threats, low mood, hopelessness, awareness of the effect of the illness, and fear of further mental disintegration, found in a later study (Drake& Cotton, 1986)of the same subjects that there was no association with depression, once multivariate analysis had adjusted for hopelessness. This backed the conclusion of Beck et a! (1985), based on their sample of 11 suicidal in-patients, that hopelessness was the best long-term predictor of suicide. Wilkinson & Bacon (1984) examined a more restricted experimental sample composed of schizo phrenic patients who committed suicide after an episode of parasuicide. Although almost half the suicides were depressed at the time of the last admission, the difference from controls did not reach statistical significance. Parasuicide could not be

752

APPLEBY

compared, but the authors did not believe that the risk of suicide in schizophrenic first-time parasuicides (8% during 0—14years' follow-up) was higher than the risk in all parasuicides. In contrast, Allebeck et a! (1987) ascribed a numerical value to the increased risk arising from some clinical and social features, calculating a relative risk of 5 for previous parasuicide. In men, alcohol abuse was associated with a relative risk of 2.5, while in women, being unmarried (relative risk =9) or living alone (relative risk =2) were more common among the suicides. Unlike the others, this study did not find an association with depression, and it concluded that many schizophrenic suicides were impulsive rather than mood-related, and consequently difficult to predict and prevent. Put together, these studies suggest that the schizophrenic patients at most risk of suicide are young men with a short illness, a history of parasuicide, features of affective disorder, and hopelessness. Suicidal ideas will often have been noted during the last admission. They may be unemployed, and possibly unmarried and alone. Yet there are difficulties in using this information clinically. The results do not present a consistent pattern, perhaps because of low numbers in some reports; and the correlates of risk are common enough in psychiatric populations to be of limited predictive

value.

Affective disorders Fawcett et a! (1987, 1990) carried out a controlled follow-up study of suicide in major affective disorder, including schizoaffective disorder. Over 10 years, 3.6% of their subjects committed suicide, and differences were identified between the 410/o who killed themselves in the first year of follow-up and those whose risk had been longer term (Table 1). Alcoholism Which alcoholics kill themselves is a question complicated by concurrent depression. In Berglund's (1984) long-term follow-up of 1312 admissions of alcoholics, suicides were more likely to have been depressed at the time of initial admission. However, perhaps because the eventual suicides were early in their psychiatric history, there was no increase in parasuicide at that time, and no association with neuropsychiatric complications such as delirium tremens and cognitive impairment. Looking at parasuicide and alcohol in a different way, Beck & Steer (1989) used multiple regression statistics to identify alcoholism as a principal predictor of suicide inparasuicides.

Murphy et a! (1979) reported that almost one-third of alcoholic suicides had experienced the loss of a close relationship, either by separation (usually) or bereavement, in the previous six weeks, and that such events were skewed towards the date of death. Rich et a! (1988) took this further, finding thatsubstance abusers(including alcoholics) who committed suicide, both with and without affective disorder, were more likely to have experienced a recent interpersonal lossor conflict thansuicides with affective disorder only. When do psychiatric patients kill themselves? This amounts to two questions. Firstly, when in the course of a patient's history is he/she most vulnerable? Secondly, when during an episode of illness is suicide most likely?

Course Flood & Seager (1968) thought there was no characteristic time for suicide during an illness but that there was a tendency for it to occur during the first four years. Myers & Neal (1978), although they did not study duration of illness per se, found the mean age of their sample to be 54 (they did not quote a range),

suggesting

that either the age of

onset was late or the duration was long. Having matched for age and time of presentation, Roy (1982b) presented sex differences within the schizophrenic experimental group, finding that men became ill younger, presented earlier, and killed themselves not only younger but after a shorter duration of illness —¿ 4.8 years in men compared with 9.8 years in women. The eight suicides (all male) in the schizophrenic cohort of Cohen et a! (1990) were even earlier in the course of their illness, the mean duration being only 2.9 years, while 49°lo of the subject sample of Goh et a! (1989) had been ill for less than a year. In contrast, the schizophrenic sample of Wilkinson & Bacon (1984) had been diagnosed later than in these studies and both sexes had been ill for an average of 10 years, which was not different from controls, although the suicides had had a mean of seven admissions during this time, twice as many as controls. Episode Barraclough, quoted by Crammer (1984), reported 73 in-patient suicides in which the timing was known: 18% had died within a week of admission and 34% within a month, while 23% had been in hospital for over a year. Similar results were quoted

SUICIDE RISK AND PREVENTION

753

carewithsuicides afterthis by Galeeta!(1980), who foundin-patient suicide days of,in-patient almost, but not quite, in-patient versus out totakeplaceeither intheacutephaseofpsychotictime —¿ patient suicides. The in-patients were more likely to relapse and therefore in the first week of admission, or after a month, at a time of clinical improvement have been diagnosed as depressed, schizophrenic or paranoid, and their sex distribution was equal while but often when there were problems in planning in the out-patients men were twice as common. discharge. In summary, there is no single point during the Roy's (l982a) figures for time after discharge showed the same pattern, 18% dying within one week course of a chronic psychiatric illness when suicide of entering out-patient care, 44% within one month, risk is at its peak. The first few years seem to and 89% within one year. Of his sample, 90% died represent a high-risk time, particularly for young within a month of their last out-patient appointment, schizophrenic males, but patients are also at risk 17% within one day. Flood & Seager (1968) found some years after the onset of their disorder. In the something similar, if less dramatic. More than half course of an episode, the maximum risk occurs at of their suicides were committed within three months two points: (a) at the beginning of the acute phase; of in-patient discharge, despite out-patient follow-up (b) after in-patient discharge (presumably during and in many cases. The last admission had often been after clinical resolution), especially in the first three brief, less than two weeks in one-fifth of the sample, months. As one of the few consistent findings, these and had often ended with self-discharge. Myers & crucial periods highlight the potential impact of Neal (1973) supported these findings, 36% of their community care on suicide, since it is at these times sample having killed themselves within a week of that patients are increasingly likely to be managed seeing a doctor, 630/o within one month. Either without in-patient supervision. Two additional points on timing concern diurnal because of the illness or because of the nature of care, or both, there are clearly dangerous periods in variation and clustering. Salmons (1984), whose the course of psychiatric disorders, both at the subjects were all in-patients, most suffering from beginning and at the end of an episode of illness, depressive psychosis, found that early morning a pointconfirmed statistically byCopaseta!(1971). was a likely time for suicide. As for clustering, For schizophrenia, Roy (1982b) reported com there are persuasive but sporadic reports of imitation parable figures, the time since last discharge being suicides or episodesof self-harm (Crawford& six months for suicides and 29 months for controls. Willis, 1966;Zemishlanyet a!,1987),but itis Again the distribution of out-patient suicides was rare. When Modestin & Wurmle (1989) examined all skewed towards the point of discharge, 25% dying in-patient suicides in two Swiss hospitals, they found within two weeks and 50% within three months. The no evidence of temporal clustering. durations presented by Wilkinson & Bacon (1984) were much longer, 32 months in men and 10.5 How do psychiatric patients kill themselves? months in women, although the range was wide and the shortest period was only a few days. In the men Identification of frequency differences in suicide but not in the women, the post-discharge period was method between psychiatric patients and the general longer than in the control subjects. However, recent population would require an age- and sex-controlled contact with day-patient or out-patient care was the comparison study, as suicide method is influenced rule —¿ all the suicides in this study had been seen in by both these factors (Office of Population Censuses one of these settings no more than a month earlier. and Surveys, 1990). Although no such study has The Monroe County study (Kraft & Babigian, been carried out, there is a general consensus in 1976) not only found that half of those suicides the research that psychiatric patients use more violent witha psychiatric history had had contactwith means. Two methodological approaches have come psychiatric services within the previous month but to this conclusion —¿ firstly, the detailing of methods that this was particularly true of those diagnosed used by psychiatric suicides, and secondly, the study with depressive neurosis. In contrast, Jensen (1966) of the psychiatric background of those who commit suicide violently. found those neurotic patients who had been in patients andhadlater killed themselves didso“¿some The figures found by studies of method have time after discharge―,unlike psychotic patients, who been remarkably close, at 70—90%.Violent methods were most likely to die in the early acute phase of were used by 79% of the 57 suicides among in an episode or shortly after discharge. patients and day patients described by Goh et a! Some support for this difference between diagnoses (1989), 88% of the in-patient suicides in the Exe came from Sundqvist-Stensmann (1987), who Vale Hospital study (Langley & Bayatti, 1984), compared suicides currently in, or within seven 70% of suicides by current and recent in-patients

754 in Bristol

APPLEBY (Morgan

& Priest,

1991),

and 75%

of

the men and 55% of the women in Roy's (1982a) sample of ‘¿all diagnosis'

these drugs, being well established, may be more readily prescribed for the most depressed patients.

suicides in and out of

hospital. It is possible, however, that non-violent Which aspects of care can influence suicide? attemptsare more readily detected and treated when they occur in close proximity to hospital care, Location of care biasing the figures on fatal outcome towards methods which cause sudden death. Rorsmann's (1973) study It has already been detailed how suicide is associated contained evidence against this in that the previous with certain clinical features and particular periods parasuicides offemalesuicides werealsomorelikely of a patient's history, implying that suitably directed tobe violent. clinical management can reduce risk, and that the Similar results have been reported for schizo nature and location of care are important influences phrenia. In Roy's (1982b) schizophrenic subgroup, on outcome in potentially suicidal patients. This 80% died violently, and most of those whose is, quite literally, a vital issue at a time of not only method involved jumping or shooting were men. community-based care but under-resourced care. The Allebeck & Wistedt (1986) described violent means central question for research into suicide among in 72% of their schizophrenic suicide sample. No sex psychiatric patients is therefore: does community differences were found, but when specific methods care as it currently operates influence the suicide were listed, men were more likely to die by hanging rate? or jumping, women by drowning and poisoning. In The studies discussed so far did not set out the sample of Wilkinson & Bacon (1984), the suicide to answer this point, but Walk (1967) addressed methods were mixed, 580/odying by violent means, the question in a different way, comparing rates especially jumping and drowning, the remainder of suicide in Chichester before and after the being self-poisonings. Could this reflect the fact that introduction of a community-based mental health the sample was older than in some studies? If so, service. There was no change, but it was thought the value of a comparison study that would control that the new service might have protected elderly for age is underlined. people. There are methodological reasons for not Several studies have reported on people who jump generalising these conclusions: the study was geo from high places. Sims & O'Brien (1979) studied 22 graphically and numerically limited; its figures psychiatric patients who did so, although six of these included all suicides rather than the high-risk group jumped from heights of less than six feet. All but it was interested in, psychiatric patients; and it did not relate suicide by individuals to the nature of one were psychotic, five had jumped previously, and all but two survived, both deaths occurring in care they were receiving at the time. Williams et a! (1986) have also pointed out how inaccuracies can patients whose hallucinations told them to commit suicide. Salmons (1984) confined her study to arise in ‘¿before and after' comparisons when rates completed suicides in hospital and found that 13 of rather than trends are compared. The same authors later contrasted the suicide rate 14 cases were by jumping. Pounder (1985) collected 29 suicides by jumping in the Adelaide area. The trends in Italy before and after 1978, the year in majority were young and, although only 14% were which law 180 was passed, requiring the closure of mental hospitals. They discovered that the rate for hospital in-patients, there was a large preponderance the whole country had steadily increased since 1976 of schizophrenia compared with general population suicides, a finding endorsed by Cantor et a! (1989) but that the increase was not uniform, being found in their study of suicides and parasuicides from two in 10 of 19 regions, particularly in the north of the country. When they correlated regional changes with bridges in Brisbane. provision of services per million population, there In non-psychotic patients such as those diagnosed was no relationship to the number of mental hospital as alcohol or drug dependent or sociopathic, beds or the available community services, although Ovenstone & Kreitman (1974) suggested that over dose was the likely method of suicide, as in the there was a negative correlation with numbers of general population. In depression, an extrinsic psychiatric beds in general hospitals. Whether the location of care is directly associated factor may influence suicide —¿ the inherent toxicity of antidepressants. Farmer & Pinder (1989) have with suicide in high-risk groups has rarely been proposed this explanation for the finding that tested. Cohen et a! (1990) contrasted two forms amitriptyline and dothiepin are associated with the of community care and found no difference in the rate of suicide. Published series of suicides highest rates of fatal overdose (over 50 per million among psychiatric patients which are not restricted prescriptions), although it could be argued that

SUICIDE RISK AND PREVENTION to one treatment setting usually find that out-patients predominate, but without a suitable comparative design, this means little to the estimation of risk. Reports of the duration of out-patient care before suicide have repeatedly confirmed that risk is associated with recent change from in-patient to out-patient status. This can be explained in one (or more) of three ways: (a) coincidence —¿ transitions in care may occur at points in an illness when risk is high; (b) loss of prevention —¿ an abrupt lessening of supervision may push the balance of risk and protection towards risk; (c) exacerbation —¿ major changes in care may place stresses on patients and thus increase risk. If transitions to less supervised forms of care do increase risk, the policy of short admissions followed by community-based care may lead to a failure to protect patients at times when their risk is faffing only gradually, if it is faffing at all. In support of this disturbing possibility, Perris et a! (1980) observed

that both the number

of

suicides in Sweden and the proportion occurring in the first month after discharge were increasing, and suggested that early discharge could be responsible. Flood & Seager (1968) agreed that the recent admissions of suicides were more likely to have been brief, but Roy (1982b) found no difference in duration of admission, or in the use of social case work or day care. Identification Because many of the factors correlating with suicide are also associated with psychiatric illness in general, it is a difficult task to predict the few patients who will kill themselves. Pokorny (1983) carried out a prospective study of 4800 consecutive admissions, following them up for a mean of five years. The low suicide rate relative to the frequency of suicide risk factors meant that there were many false positives but, more surprisingly, there were also many false negatives. This was true whether the prediction was attempted using single risk items, factor scores, or discriminant function. Given the low specificity and sensitivity of risk assessment, it is not surprising that awareness of imminent suicide in clinical practice is often poor. Of the in-patient and day-patient suicides identified by Goh et a! (1989), suicidal thoughts had been noted in 58% but in only 28% was the suicide risk itself specifically

referred

to in the medical

notes, while the equivalent figure for the nursing notes was 21%. No more than 11% were on special nursing observations.

755

Intent Both Robins et a! (l959a,b) and Barraclough eta!(1974) established that suicides have usually informed someone directly of their intent and it is also clear that indirect communication is common, for example through previous parasuicide or complaints of hopelessness. Whether either leads to successful intervention is unknown but in the (uncontrolled) study by Goh et a! (1989), 5% of patients who killed themselves while in-patients or day patients had told a staff member of their intention to do so in the two weeks before they died. Ovenstone & Kreitman (1974) put forward the view that communication of intent, among other things, differentiated two syndromes of suicide. In one, exemplified by the bereaved spouse, previously ‘¿stable personalities' reacted to acute stress by committing suicide using a highly dangerous method, often self-injury, and without informing anyone. In the other, people diagnosed as sociopathic or dependent on drugs or alcohol, killed themselves, often by overdose, while others were in the vicinity and after clearly stating their intent. Despite the latter pattern, Lewis & Appleby (1988), in a study of psychiatrists' attitudes, found that after a diagnosis of personality disorder, a patient's suicidal urges were more likely to be regarded as under control, and that this correlated with the patient being seen as manipulative, attention-seeking, annoying and undeserving of National Health Service time. This agreed with the study of in-patient suicides by Morgan & Priest (1984), who found that staff were often critical of future suicides for being over dependent, provocative and unreasonable, and for deliberately assuming disabilities. Flood & Seager (1968) put the point more strongly. In their study, the feature which most strikingly distinguished suicides was “¿disturbed relationships with hospital staff, resulting in premature discharge―. In addition, only 6% of the suicides had been judged to have a normal personality, compared to over 20% in the two comparison groups, a possible reflection of poor relationships between staff and patients. Treatment The details of care were not an area of comparison in the Myers & Neal (1978) study, although there was descriptive evidence of untreated depression despite recent contact with psychiatric services. Of those suicides with a history of depression and whose treatment at the time of death was known, 70% were receiving no more than a quarter of the recommended dose of antidepressant and

APPLEBY

756

most of these were receiving none at all (although many were taking anxiolytics). In the Lund sample (Rorsmann, 1983), suicides by men were more likely in clinic attenders who subsequently received no treatment. Studies of schizophrenic

suicide (Roy,

positioned on the ground floor. However, of the 73 in-patient suicides discussed by Crammer (1984), 57% occurred away from the hospital (more than half of these were on leave) and only l80/ooccurred on a psychiatric ward.

1982b;

Wilkinson & Bacon, 1984) found no association with the taking of drug treatment or its nature, but two uncontrolled studies of affective disorder have come to ambiguous conclusions. Schou & Weeke (1988) estimated that 70% of manic—depressivesuicides were receiving adequate prophylaxis before death, while Modestin (1985) found that almost half of a series of clinically depressed suicides were not being treated. Rutz eta! (1989) took a different approach, observing the suicide rate in Gotland before and after an education programme for general practitioners on the diagnosis and treatment of depression, and comparing it with the suicide rate in Sweden as a whole. Their finding was a sharp decline which was not evident nationally. The patient's attitude to treatment and staff may also be important. Drake et a! (1984) found that suicide was more likely in schizophrenic patients who felt inadequate in relation to their treatment goals. Virkkunen (1976) similarly found that schizophrenic suicides, more than suicides with other diagnoses, were negative towards their treatment and carers in the months before death.

Conclusions:prevention in clinical practice and research

Certain features of the mental state, social back ground, and course of the illness appear to be associated with increased risk. Many of these are risk factors for suicide by the non-psychiatric population also but some are different (e.g. age) and in some (e.g. marital status) the evidence is less definite in patient groups. A problem with most risk factors is that they are relatively common in all psychiatric patient groups. They correlate well with suicide but predict it badly. Nevertheless, taken together, they offer the opportunity to identify high-risk groups within the already high-risk population of psychiatric patients. More refined prediction of risk must wait for research into the interaction of the key factors. Few studies have looked at protective factors, but the presence of dependants, specified by Flood & Seager (1968), implies that some of these too are common to psychiatric subjects and the general population. However, the principal protective factors must lie in the nature of psychiatric care. The move to the community challenges efforts at Ward milieu suicide prevention because it affects patient care at two Crammer (1984) included plans for rehabilitation points in an episode of illness, which appear to be or discharge in his list of explanations for ‘¿epidemics' points of relatively high suicide risk: the initial acute of suicide in UK hospitals, suggesting that disruption phase, when a patient may be treated at home, and the of a patient's routine could place him or her at period of recovery when he/she may be discharged risk. Other explanations listed included poor staff earlier than was once conventional practice, perhaps communication and poor observation on wards, at a time of incomplete recovery. This is not to whether because of physical design or staff numbers, oppose community care but it does strongly suggest training, or morale. In one hospital outbreak, that community-based care plans should emphasise blame was placed on poor staff relationships —¿ recognition and early preventive intervention in high arguing among senior staff and faulty supervision risk patients. The basis of that recognition is given in of juniors leading to poor morale and poor care Table 2. The general clinical priorities for preventive (Anonymous, 1977). A less convincing relationship aspects of care are listed as a mnemonic below. of a similar kind was found by Coser (1976) studying The following points should receive close attention: a US hospital epidemic. Alternatively, Kahne (1968) concluded that it was the relative size and S —¿ Sources of staff hostility towards patients turnover of staff and patient populations, but U—Understaffing and poor design of treatment not the actual staff numbers, that was related to settings suicide in hospitals.

Sims & O'Brien (1979), in their study of fatal and non-fatal self-harm by jumping, found that most subjects jumped from a hospital building and recommended that wards for disturbed patients should be staffed with highly trained nurses and

I —¿ Identification

of high-risk individuals

periods C I D E

—¿ Changes in location of care —¿ Isolation —¿ Depressed mood —¿ Expressions of intent.

and

SUICIDE RISK AND PREVENTION

757

Table 2 Featureswhichmightincreasethe riskof suicide Feature

Factors

Mental state

Psychosis Depressedmood Hopelessness Suicidal ideas Suicidal content to psychotic phenomena Communicationof intent Previousparasuicide History < 4 years Severaladmissions Long history with recent change Living alone Single, divorced or widowed Unemployed Male Young Acute relapse Recentdischarge In-patient or recent out-patient Recenttransition in care Staff hostility to patient High staff or patient turnover Low morale Insufficient observationfacilities Inadequatestaff expertise

Past history

Social/demographic features

Current episode/treatment Ward and staff

One priority for suicide research in the age of community treatment is to study in detail the relation ship between suicide and psychiatric and other services, particularly its location, nature and transitions. This can only be attempted using a substantial sample of psychiatric suicides and a case-control design. The numbers required would be more easily obtained through a region-wide or multicentre study. Alternatively, a practice could be borrowed from obstetrics. The Report on Confidential Enquiries into Maternal Deaths (Department of Health, 1990) records

all deaths

by women

within

a year of

childbirth, dividing them into direct, indirect and fortuitous according to their causal link with child bearing. The details of each case are recorded so that the number of deaths, their cause, and their remediable features can be monitored and used to improve prevention. Such a system of data collection would be invaluable if it were able to detect sudden deaths within a year of psychiatric contact where the cause was suicide or undetermined'. Morgan & Priest (1991) have suggested a similar system by which health districts could monitor their sudden deaths locally. Either model would be able to monitor a potentially preventable public health problem

affecting a priority area of health care at a time of fundamental change in policy and practice. 1. Since submission of this article, a Confidential Enquiry into Homicides and Suicides by Mentally Ill People has been established with the support of the Department of Health and the Royal College of Psychiatrists.

References AuaaEcIc, P. & WISTEDT,B. (1986) Mortality in schizophrenia. A ten-year follow-up based on the Stockholm County in-patient register. Archives of General Psychiatry, 43, 650—653. V@iu..&, A., KRISTJANSSON, B., et a! (1987) Risk factors for suicide among patients with schizophrenia. Acta Psychiatrica Scandinavica,76, 414—419. & ALLOULANDER, C. (1990) Suicide among young men: psychiatric illness, deviant behaviour and substance abuse. Ada PsychiatricaScandinavica,81, 565—570. ANONYMOUS (1977)

A suicide

epidemic

in a psychiatric

hospital.

Diseases of the Nervous System, 38, 327-33 1. Apputay,

L. (1991)

Suicide

during

pregnancy

and

in the

first

postnatal year. British Medical Journal, 302, 137-140. BARRACLOUGH,

B. M.,

BUNCH,

J.,

NELSON,

B.,

et al (1974)

A

hundred cases of suicide. British Journal of PsychIatry, 125, 355—373. Bacic, A. T., Smaa, R. A., KOVACS, M., et al(1985) Hopelessness and eventual

suicide: a 10-year prospective

study of patients

hospitalised with suicidal ideation. American Journal of Psychiatry,142, 559—563. & (1989) Clinical predictors of eventual suicide: a five- to ten-year prospective study of suicide attempters. Journalof Affective Disorders,17, 203—209. BEROLUND, M. (1984)

Suicide

in alcoholism.

A prospective

study

of 88 suicides: I. The multidimensional diagnosis at first admission. Archives of General Psychiatry, 41, 888—891. & NILSSON,K. (1987) Mortality in severe depression. A prospective study including 103 suicides. Ada Psychiatrica Scandinavica,76, 372—380. CANTOR, C. H.,

HILL, M. A. & McLAcrn@,

B. K. (1989)

Suicide

and related behaviour from river bridges. A clinical perspective. British Journal of Psychiatry, 155, 829—835. COHEN, L. 3., TEST, M. A. & BROwN, R. L. (1990)

Suicide and

schizophrenia: data from a prospective community treatment study. American Journal of Psychiatry, 147, 602—607. CopAs,

J. B.,

FREEMAN-BROWNE, D.

L. & ROBIN, A. A. (1971)

Danger periods for suicide in patients under treatment. Psychological Medicine, 1, 400—404. Coma, R. L. (1976) Suicide and the relational system: a case study

in a mental hospital. Journal of Health and Social Behaviour, 17, 318—327. [email protected], J. L. (1984) The special characteristics

of suicide in

hospital in-patients. British Journal of Psychiatry, 145, 460-476. CRAwFORD,

J.

P.

& WILLIS,

J.

H.

(1966)

Double

suicide

in

psychiatric hospital patients. British Journal of Psychiat,y, 112, 1231—1235. DEPARTMENT OFHa@m (1990) Report on Confidential Enquiries

intoMaternalDeathsin Englandand Wales.London:HMSO. DRAKE,

R.

B.,

GATES,

C.,

Col-rON,

P.

0.

et a! (1984)

Suicide

amongschizophrenics.Whois at risk?Journal of Nervousand Mental Disease, 172, 613—617.

& CorroN,P. 0. (1986)Depression, hopelessness and suicidein chronicschizophrenia.BritishJournal of Psychiatry, 148, 554—559. FARMER, R. D. T. & Pnmaa, R. M. (1989) Why do fatal overdose rates vary between antidepressants? Acta Psychiatrica

Scandinavica,80 (suppl. 354), 25—35. FAWCE1-r,

J.,

SCHEI'TNER,

W.,

CLAIUC, 0.,

ci a! (1987)

Clinical

predictors of suicide in patients with major affective disorders: a controlled prospectivestudy. American Journal of Psychiatry, 144, 35—40.

Fooo, L., ci a! (1990) Time-related predictors of suicide in major affective disorder. American Journal of Psychiatry, 147, 1189—1194. FLOOD,

R.

A.

& SEA0ER,

C.

P.

(1968)

A retrospective

examination

of psychiatric case records of patients who subsequently committed suicide. British Journal of Psychiatry, 114,443-450.

APPLEBY

758 GALE, S. W.,

MESNIKOFF, A.,

FINE, 3., et a! (1980)

A study

of

suicide in state mental hospitals in New York City. Psychiatric

Quarterly, 52, 201—213. suicides: are there preventable factors? Profile of the psychiatric hospital suicide. British Journal of Psychiatry, 154, 247-249. S. B. & ROBINS, E. (1970)

Suicide

and

primary

affective

disorder. British Journal of Psychiatry, 117, 437-438. JENsEN,

L.

(1966)

Nosocomial

suicides

and

suicides

among

discharged patients. Acta Psychiatrica Scandinavica, 42(suppl. 191), 149—170. KAHNE, M. J. (1968) Suicides in mental hospitals: effects

of personnel

and patient

turnover.

a study of the

Journal

of Health

and

Social Behaviour, 9, 255—266. KENDELL,

R.

B.,

CHALMERS,

3.

C.

&

PLATZ,

C.

(1987)

Epidemiology of puerperal psychoses. British Journal Psychiatry, 150, 662—673. KING, E. & BARRACLOUGH, B. (1990) illness. A study of a single catchment

of

Violent death and mental area over 8 years. British

Journal of Psychiatry, 156, 714-720. KR@r, D. P. & BABIGIAN, H. M. (1976) Suicide by persons with

and without psychiatric contacts. Archives of General Psychiatry, 33, 209—215. KUMAR,

R.

& ROBSON,

K.

M.

(1984)

A prospective

study

of

emotional disorders in childbearing women. British Journal of Psychiatry, 144, 35—47. LANGLEY,

0.

E.

& BAYArFI,

N.

(1984)

Suicides

in

Exe

Vale

Hospital 1972-1981. BritishJournalof Psychiatry, 145,463-467. LEWIS, 0.

& APPLEBY, L. (1988)

Personality

disorder:

000DSTEIN,

J. L.,

NIELSEN, S. L.,

et a! (1983)

Reasons for staying alive when you are thinking of killing yourself:

the Reason

For Living

inventory.

Journal

S. B., et a! (1985)

therapy

Mortality

& PRIEST, P. (1984)

in

psychiatric

patients.

Report

of

D. J. (1985)

Suicide

by leaping

from

multistorey

car

parks. Medicine, Science and the Law, 25, 179—188. REGIER, D. A., HIRSCHFIELD,

R. M.,

GOODWIN, F. K., ci a! (1988)

The NIMH Depression Awareness, Recognition, and Treatment Program: structure, aims, and scientific basis. American Journal of Psychiatry, 145, 1351—1357. RICH, C. L.,

FOWLER, R. C.,

FoGARi-y,

L. A.,

et a! (1988)

San

Diego suicide study. III. Relationship between diagnoses and stressors. Archives of General Psychiatry, 45, 589—592. ROBIN A. A.,

BROOKE, E. M. & FREEMAN-BROWNE, D. L. (1968)

Some aspects of suicide in psychiatric patients in Southend. British Journal of Psychiatry, 114, 739-747. ROBINS, B., MURPHY, 0. B., WILKINSON, R. H., ci al(l959a)

Some

clinical considerations in the prevention of suicide based on a study of 134 successful suicides. American Journal of Public Health, 49, 888—899. —¿,

GASSNER,

S.

&

KAYES,

J.

(l959b)

The

communication

of

suicidal intent: a study of 134 consecutive cases of successful (completed) suicide. American Journal of Psychiatry, 115, 724—733. RORSMANN, B. (1973)

Suicide

in psychiatric

patients:

a comparative

study. Social Psychiatry, 8, 55-66. Roy, A. (1982a) Risk factors for suicide in psychiatric

patients.

Archives of General Psychiatry. 39, 1089-1095. —¿

(l982b)

Suicide

in

chronic

schizophrenia.

British

Journal

of Psychiatry, 141, 613—617. of suicide on Gotland after systematic postgraduate education of general practitioners. Acta Psychiatrica Scandinavica, 80, 151—154.

SCHOU, M. & WunKE,

in depressed clinical

suicides. Acta Psychiatrica Scandinavica, 71, 111-116. & WURMLE,0. (1989) Role of modelling in in-patient suicide: a lack of supporting evidence. British Journal of Psychiatry,155, 511—514. MORGAN, H. 0.

suicide

British Journal

of Psychiatry, 145, 469—472.

in a follow-up of 500 psychiatric outpatients II. Cause-specific mortality. Archives of General Psychiatry, 42, 58—66. MODESTIN, J. (1985) Antidepressive

of

SALMONS,P. H. (1984) Suicide in high buildings.

of Consulting

and Clinical Psychology, 51, 276—286. MARTIN, R. L., CLONINGER, R., Guza,

Prediction

RUTZ, W., VON KNORRING,L. & WOLINDER,J. (1989) Frequency

the patients

psychiatrists dislike. British Journal of Psychiatry, 153,44—49. LINEHAN, M. M.,

(1983)

a prospectivestudy. Archivesof GeneralPsychiatry,40,249-257. POUNDER,

GOH, S. B., SAU@4oNS, P. H. & WHrrriNcrroN, R. M. (1989) Hospital

GUZE,

—¿

Assessment

of suicide

risk in

A.

(1988)

Did

manic

depressive

patients

who committed suicide receive prophylactic or continuation treatment at the time? British Journal of Psychiatry, 153, 324—327. SIMS, A. & PRIoR,

P. (1978)

The

pattern

of mortality

in severe

neuroses. British Journal of Psychiatry, 133. 299—305. —¿

&

O'BIuItN,

K.

(1979)

Autokabalesis:

an

account

of

mentally

ill people who jump from buildings. Medicine, Science and the Law, 19, 195—198.

psychiatric in-patients. British Journal of Psychiatry, 145,467-469.

SUNDQvIST-STENSMAN,

& —¿ (1991) Suicide and other unexpected deaths among psychiatric in-patients. The Bristol Confidential Inquiry. British Journal of Psychiatry, 158, 368-374.

with psychiatriccare:an analysisof 57 cases in a Swedishcounty. Acta Psychiatrica Scandinavica,76, 15—20. TEMoCHE,A., PUGH,T. F. & MCMAHON,B. (1964) Suicide rates among current and former mental institution patients. Journal of Nervous and Mental Disease, 138, 124-131.

MURPHY, 0.

E., ARMSTRONG, 3. W.,

HERMELE, S. L., ci al (1979)

Suicide and alcoholism. Interpersonal loss confirmed as a predictor. Archives of General Psychiatry, 36, 65-69. Myms, D. H. & NEAK,C. D. (1978) Suicide in psychiatric patients. British Journal of Psychiatry, 133, 38-44. Ornca OF PoPuI@noN CENSUSES @rsm SURVEYS(1990) Mortality

Statistics: Cause. London: HMSO. OvEN5T0NE,

I. M. K. & KREm.w@i,

N. (1974)

Two

syndromes

of

suicide. British Journal of Psychiatry, 124, 336—345. FERRIS, C.,

Bsmcow,

J. & JACOBSON, L. (1980)

Some

remarks

on

the incidence of successful suicide in psychiatric care. Social Psychiatry, 15, 161—166. P0K0RNY, A. D. (1964) Suicide rates in various

psychiatric

dis

orders. Journal of Nervous and Mental Disease, 139, 499-506. —¿

(1966)A

follow-up

study

of

618

suicidal

patients.

Journal of Psychiatry, 122, 1109-1116.

American

U. B. (1987)

VuuucuNEN, M. (1976) Attitude

Suicides

in close

to psychiatric

connection

treatment

before

suicide in schizophrenia and paranoid psychoses. British Journal of Psychiatry, 128, 47—49. WA.uc,D. (1967)Suicideand community care. British Journal of Psychiatry,113, 1381—1391. WILKINSON,

0.

&

BACON,

N.

A.

(1984)

A

clinical

and

epidemi

ological survey of parasuicide and suicide in Edinburgh schizophrenics. Psychological Medicine, 14, 899—912. WILLIAMS,

P.,

DE SAIvIA,

D. & TANSELLA,

M.

(1986)

Suicide,

psychiatric reform, and the provision of psychiatric services in Italy. Social Psychiatry, 21, 89-95. Zausm@y, Z., WEINBERGER, A., BEN-BASSAT, M.. et a! (1987) An epidemic of suicide attempts by burning in a psychiatric hospital. British Journal of Psychiatry, 150, 704-706.

Louis Appleby, BSc,MRCP,MRCPsych, Senior Lecturer, Department of Psychiatry, University Hospital of

South Manchester, Manchester M20 8LR

Suicide in psychiatric patients: risk and prevention. L Appleby BJP 1992, 161:749-758. Access the most recent version at DOI: 10.1192/bjp.161.6.749

References Reprints/ permissions You can respond to this article at Downloaded from

This article cites 0 articles, 0 of which you can access for free at: http://bjp.rcpsych.org/content/161/6/749#BIBL To obtain reprints or permission to reproduce material from this paper, please write to [email protected] /letters/submit/bjprcpsych;161/6/749 http://bjp.rcpsych.org/ on March 20, 2017 Published by The Royal College of Psychiatrists

To subscribe to The British Journal of Psychiatry go to: http://bjp.rcpsych.org/site/subscriptions/