Aims: The aim of this survey was to describe patients in care at a large mental health department in northern Italy who attempted suicide, and the clinical management adopted by their psychiatrists before the event.
Methods: Data collection was based on a questionnaire administered to the reference psychiatrists.
Results: Over a period of 12 months, 166 catchment area residents attempted suicide. Sixty-six (40%) had contacted the mental health department in the previous two years and completed data were obtained on 63. Twenty-nine (46%) suffered from mood, 26 (41%) from personality and 11 (18%) from schizophrenic disorders. Thirty-four attempts occurred within one year of psychiatric ward discharge, mostly in the first quarter.
The reference psychiatrists reported that, at the last evaluation, 38 of 63 patients (60%) presented no change in clinical conditions, and 41 of 63 (68%) were considered at no immediate risk of suicide. Most of the attempted suicides in question (45, 72%) were judged to be unpreventable. In the two logistic regression analyses carried out, no independent variables were able to statistically significantly explain the variance in judged suicidal risk or the preventability of the index attempted suicide.
Conclusions: According to the psychiatrists' descriptions of their last contact with the patients, most attempted suicides have not been preceded by a change in clinical conditions. Moreover, psychiatrists, irrespective of their age and gender, and of patient diagnosis, frequently judged the attempts to have been unpreventable.
PSYCHIATRIC PATIENTS PRESENT high suicide behavior rates. Around 95% of suicides have been reported to suffer from psychiatric illness1 and almost one quarter to have had at least one contact with mental health services in the year before death.2 Many suicide attempters (36%–60%)3–5 also have a history of psychiatric treatment and several contacts with mental health services.
Clinical management and appropriateness of psychiatric consultation after a suicide attempt have been variously evaluated.4,6,7
Studies by Appleby and colleagues2 have provided abundant information on features of clinical care of persons in contact with psychiatric services before completed suicide, but little research has concentrated on these features in patients who attempted suicide.
The aim of this survey was to evaluate routine clinical management of patients by mental health department (MHD) psychiatrists in the period just prior to an attempted suicide (AS). Particular attention was paid to their opinion about suicide risk and the preventability of the AS in their patients, in addition to any suicide prevention strategies adopted during the same period, based on the outcome of their last contact prior to the AS. Another aim was to investigate whether patients' psychiatric diagnosis and length of psychiatric history, or psychiatrists' age and gender correlated with judgment of suicidal risk and of AS preventability. Lastly, consideration was taken of change in reported clinical conditions and changes in therapy made by the psychiatrist during the last appointment with the patient, as developments in the patient's clinical picture may be associated with a change in suicide risk.
This survey was conducted in three stages: identification of people who attempted suicide, identification of those in the sample who had been in contact with the MHD and collection of data through the questionnaire administered to the reference psychiatrists.
The catchment area was the city of Padua, in northeastern Italy, in addition to 19 peripheral municipalities belonging to a single MHD.
The mental health department provides psychiatric care by various multidisciplinary teams, each consisting of a psychiatrist, psychologists, nurses, social workers and educators who operate in several facilities (emergency psychiatric service, psychiatric ward, mental health center-MHC etc.).
In routine clinical management each patient has his/her own reference psychiatrist who arranges a personalized therapeutic program, in cooperation with other staff members.
Data were collected through a questionnaire. Our model was inspired by the National Confidential Inquiry into Suicide and Homicide by People with Mental Illness,2 which we adapted to the object of our survey (i.e. attempted suicide). It was divided into four sections: demographic features; psychiatric and general medical history; details of AS; and details of previous clinical management, focusing particularly on assessment and management of suicide risk. Reference psychiatrists were asked to provide details on their last psychiatric evaluation, i.e. date, place (departmental facility), clinical condition and management; their judgment of suicide risk and of the preventability of the AS; their awareness and adoption of the no-suicide contract and/or other suicide prevention strategies before the attempt. According to the international literature the no-suicide contract consists of a formal agreement by the patient to inform a relative or healthcare provider of any suicidal intent or planning.
Psychiatric diagnoses were not based on structured interviews. Nevertheless, they were longitudinal, i.e. formulated after several assessments and, mostly, in different settings (e.g. in- and out-patient facilities), and based on DSM-IV criteria.
Identification of people who attempted suicide who had been in contact with the MHD
We selected people who had AS over a 12-month period (1 February 2005 to 31 January 2006) and who had been in contact with Padua MHD in the 24 months prior to the AS (index case). For the purposes of this paper we excluded acts occurring after the first attempt in the same 12-month period. The patients were selected from among those who had AS and presented to the emergency room, or were hospitalized in a psychiatric or non-psychiatric ward, or had received care at one of the MHD's community facilities. We were able to identify attempters who had been in contact with Padua MHD in the two years prior to the attempted suicide through records collected in the MHD database since 2002. This enabled us to identify whether the patients who had AS were already in care at the MHD, and consequently their reference psychiatrists and teams.
We were able to cross-check data by contacting each MHD psychiatrist on a monthly basis to enquire whether any of their patients had AS in the intervening period. This enabled us to collect a complete, reliable sample.
The definition of AS was based on WHO/EURO criteria, i.e.: ‘An act with a nonfatal outcome in which an individual deliberately initiates a nonhabitual behavior that, without intervention by others, will cause self-harm, or deliberately ingests a substance in excess of the prescribed or generally recognized therapeutic dosage, and which is aimed at realizing changes which he/she desired via the actual or expected physical consequences’.8
We excluded all behaviors involving the deliberate infliction of direct physical harm to one's own body without any intent to die as a consequence of the behaviour.9
When we had identified the people who had attempted suicide and their reference psychiatrists, the providers were requested to fill out the questionnaire. After giving a thorough description of the study, the participating psychiatrists signed a written informed consent form. The participating psychiatrists obtained signed consent to take part in the study from the people who had attempted suicide.
Institutional review board approval was obtained.
During the 12-month period we identified 226 persons who had attempted suicide, 166 (74.4%) of whom were living in the MHD catchment area. We were unable to identify place of residence for three people. Sixty-six of the sample of 166 persons living in the MHD catchment area (40%) were in contact with the MHD. We collected 63 completed questionnaires filled out by 28 psychiatrists; we excluded three patients because their psychiatrists declined to take part in the study. The mean time between the index attempt and data collection was 7.18 weeks (SD 3.35).
Characteristics of suicide attempters in care at MHD
The sample consisted of 63 adult Caucasians (mean age 42.8 years, SD 13.8; range 20–84); 41 (65%) were female, with a female-to-male ratio of 1.8:1. Forty-six (73%) had never been married or were divorced, separated or widowed. As regards social status, 20 (32%) were currently unemployed and 19 (31%) were economically inactive (i.e. retired people, students, housewives).
Among the sixty-three persons forming the studied population, the commonest AS method was drug poisoning (n = 37, 61%), among both men (n = 13, 59%) and women (n = 24, 58.5%). Eight attempters used a second method, mostly alcohol (n = 6). This population did not statistically significantly differ in mean age (42.8 years, SD 13.8, range 20–84 vs 39.9 years, SD 15.4, range 18–82), gender ratio (female 65.1% vs 61%), and AS methods (61% vs 68.8% drug poisoning) from the 100 cases living in the catchment area but not in contact with the MHD.
Characteristics of MHD psychiatrists
Twenty-eight psychiatrists took part in the study, 15 of whom were female (54%); 15 psychiatrists were aged between 27 and 44 years and 13 were aged 45 or over. The sample consisted of 21 (75%) consultant psychiatrists and seven (25%) resident psychiatrists. These latter had a tutor who supervised them in management of their patients.
The database was developed and kept up-to-date using spss software (version 12.0) (SPSS Inc., Chicago, IL, USA). Descriptive and multivariate analysis was performed using the same software.
A bivariate analysis was performed to study the possible relations between couples of categorical variables. Since data were independent, the χ2-test was used to compare frequencies in m*n tables (without the Yates correction) and Fisher's exact test was selected when expected frequencies were less than five. For continuous variables, a means comparison was carried out using a t-test. A P-value < 0.05 was considered significant.
Multivariate analysis was performed using two stepwise logistic regression analyses (significance level for entering = 0.10 and significance level for removing = 0.05) to identify those predictors (among psychiatrists' age and gender, patients' diagnosis and length of psychiatric history) that had a statistically significant correlation with judgment of immediate, short-time and long-time suicidal risk (presence vs absence of risk) in the first model, with preventability (preventable vs unpreventable) in the second.
We omitted from the analysis any cases with missing data on one or more variables.
The primary psychiatric diagnoses were mood (24 major depressive disorders, 38%; five bipolar disorders, 8%), personality (n = 14, 22%, 10 of whom had borderline personality disorder), schizophrenic disorders (n = 11, 17%), alcohol/drug dependence (n = 3, 5%), eating disorder (n = 1, 2%) and other disorders (n = 5, 8%).
Twenty-two patients (35%) presented a comorbid disorder (12 personality disorders, five cases of substance/alcohol abuse, five anxiety disorders not otherwise specified).
In most cases (n = 54, 86%) the time between onset of the psychiatric disorder and the index attempt was more than one year. Sixteen percent had a history of episodes of outwardly directed aggressiveness. Forty-seven (75%) patients had at least one voluntary admission to the psychiatric ward before the attempted suicide. The mean number of admissions was 7.6 (SD 11.4). Nine (14.3%) patients had also undergone compulsory hospitalization at least once.
Most patients with a history of hospitalization had AS within one year of discharge from the psychiatric ward (n = 34, 72%), particularly in the first three months (n = 25, 58%). Forty-one patients (65%) had a history of previous AS, and 20 (32%) of self-injurious behaviors.
All patients included in the sample were in care at the MHC; five of them were also in treatment at the day hospital and two were living in a residential facility. The clinical care features of the patients are reported in Table 1.
Table 1. Clinical care characteristics† of people who attempted suicide within two years of contact with the mental health department
|Therapeutic tools|| |
| Pharmacotherapy‡ (m.d. = 3)|| |
| Benzodiazepines||45 (75)|
| Antipsychotics||42 (70)|
| Antidepressants||38 (63)|
| Anticonvulsants||20 (33)|
| Lithium||3 (5)|
| Other||3 (5)|
| Psychotherapy||17 (27)|
| Compliance with appointments (m.d. = 2)|| |
| Good (up to 2 missed contacts in the last year)||40 (66)|
| Moderate (has missed less than half the appointments)||8 (13)|
| Bad (missed more than half the appointments)||13 (21)|
| Compliance with Pharmacotherapy (m.d. = 2)|| |
| Good (up to 2 missed administrations per month)||33 (54)|
| Moderate (up to 1 missed administration per week)||11 (18)|
| Bad (more than 1 missed administration per week)||17 (28)|
| Compliance with therapeutic program (m.d. = 2)|| |
| Good||27 (44)|
| Moderate||15 (25)|
| Bad||19 (31)|
|Attempted suicide preventability (m.d. = 3)|| |
| AS judged preventable||17 (28)|
| AS judged unpreventable||43 (72)|
|Features of last psychiatric contact|| |
| Time of last contact (m.d. = 2)|| |
| <24 h||3 (5)|
| 1–7 days||16 (26)|
| 1–4 weeks||23 (38)|
| 5–13 weeks||14 (23)|
| >13 weeks||5 (8)|
| Pharmacological therapy|| |
| No prescription||2 (3)|
| New prescription||8 (13)|
| Continuation||53 (84)|
| Psychiatric conditions at last contact|| |
| Unchanged||38 (60)|
| Change in the symptoms||1 (2)|
| Relapse||2 (3)|
| In remission||3 (5)|
| Better||13 (21)|
| Worse||6 (9)|
The most commonly used drugs were benzodiazepines (75%), followed by anti-psychotics (70%), and antidepressants (63%). Fifty-two patients (87%) were taking a combination of two (n = 22, 37%), three (n = 22, 37%) or four (n = 8, 13%) psychotropic drugs. Patients showed poor adherence with psychopharmacological treatment (one or more missed drug administrations per week) in 46% of cases.
The psychiatrists saw the patients once per month (n = 29, 46%) or once per week (n = 22, 35%). However, around one third of patients (34%) were poorly compliant with appointments. Ten patients (16%) were judged to have poor compliance with all aspects: psychopharmacological treatment, MHD appointments and their general psychiatric program.
Nineteen (31%) attempters had made their last contact with the psychiatrist in the week prior to the suicide attempt, and 42 patients (68.8%) had contacted the psychiatric service in the month beforehand. Seven patients (11%) had missed their last arranged appointment, three of whom had been contacted by telephone. In eight cases the last contact before the index attempt was requested by the patient him/herself.
During the last contact, the psychiatrist had not identified any change in conditions in 38 attempters (60%); in 13 cases (21%) there had been an improvement in clinical conditions and two patients were recovering from the symptoms. Worsening or a relapse in symptoms was reported in only eight attempters (13%).
Ten suicidal patients (16%) were using the ‘no-suicide contract’ with their reference psychiatrists.
According to the psychiatrists only 17 AS (28%, three missing data) could have been prevented. As preventive measures they suggested: contact with the patients' families (n = 4), co-operation with other facilities (n = 3), changes in pharmacotherapy (n = 2), use of the ‘no-suicide contract’ (n = 2), more frequent nurse contacts and/or telephone calls (n = 2), hospitalization (n = 1), prevention of manic episodes (n = 1), encouragement to accept aging (n = 1), anxiety containment (n = 1).
Most attempters were not considered to be at immediate risk of suicide (i.e. within 48 h) at their last contact (n = 41 out of 60, 69%, three missing data), or in the short term (within days or weeks; n = 30 out of 59, 51%, four missing data), or long term (n = 26 out of 60, 44%, three missing data) (Fig. 1).
Level of suicide risk judged by the psychiatrist during the last contact was tested through using the χ2-test. There were no statistically significant correlations with patients' psychiatric diagnosis, length of psychiatric history, psychiatrists' gender or age.
There was instead a statistically significant difference (χ2 = 9.356; d.f. = 2; P = 0.009) between therapists' age (but not gender) and their assessment of patients' clinical conditions at the last contact. In particular, the therapists aged less than 45 years, compared with those aged over 44 years, more frequently assessed patients as having worsened (n = 12 vs n = 1; 33% vs 3%) versus being unvaried (n = 18 vs n = 21; 50% vs 70%) or having improved (n = 6 vs n = 8; 17% vs 27%). Nevertheless, there was a clearly significant gender but not age difference in the psychiatrists' therapeutic choices. Eight female psychiatrists (23%), but no male psychiatrists, had changed the drug regimen at the last contact (χ2 = 8.670; d.f. = 3; P = 0.034).
In the two logistic regression analyses no independent variables were able to statistically significantly explain the variance in judged suicidal risk or in the preventability of the index attempted suicide.
Considering the small number of persons in our sample, we tried aggregating some variable modalities. No statistically significant covariate was found either using diagnosis as a categorical variable (with one reference option), or creating as many ‘dummy’ variables as the diagnostic modalities. We also tried dividing the variable ‘age of psychiatrist’ into groups >44 years and <45 years, without obtaining a P-value < 0.05.
To the best of our knowledge, this is the first survey of a population from a specific catchment area to consider the routine approach adopted by psychiatrists at a mental health department towards their attempted suicide patients.
Suominen et al. reported that many suicide attempters had been in contact with general health care services during the 12 months before and after the index attempt, and that half of them had had contact in the 30 days before the event.10 According to our data, 40% of attempters were in care at the MHD during the previous two years; most of them contacted the service during the month prior to the event, and these contacts had mostly been arranged. Only very few patients missed their last appointment. The demographic and clinical characteristics of our sample and circumstances of the people who attempt suicide essentially concur with the literature.11–15
It is well known that many completed suicides occur within three months of discharge from a psychiatric ward.16–18 Similarly, our data suggest that the risk of AS is higher shortly after hospitalization, as 72% of cases occurred in the first year and 58% in the first three months post discharge.
Evaluation of suicide risk
Our psychiatrists judged most patients to be at no or low immediate suicide behavior risk at the time of their last appointment, irrespective of diagnosis and length of psychiatric history, or of the gender and age of the attending psychiatrist. While we cannot rule out that clinicians did apply sound clinical judgment and that the suicide attempts were really unavoidable, there may be other explanations for this bias towards the unpredictability of suicidal behavior. Firstly, estimates of suicidal risk lack sensitivity and specificity, suggesting the need to improve risk evaluation strategies and/or find more specific parameters to assess risk, such as the distinction between warning signs and risk factors proposed by Rudd and colleagues.19 According to these authors, a suicide warning sign is the earliest detectable, most subjective indicator of heightened risk for suicide in the near-term (minutes, hours or days), such as hopelessness, rage, anger, revenge seeking, but also talking or writing about death, dying or suicide etc.
On the other hand, risk factors have a more objective quality and their presence elevates the long-term probabilistic risk for a suicidal crisis.
Secondly, judging suicide risk to be high in the absence of preventive measures would imply inappropriate treatment. The psychiatrists themselves may not have been willing to admit such error. Moreover, MHD psychiatrists frequently have to deal with very difficult patients, such as those suffering from borderline personality disorder. These patients chronically think about suicide and threaten to carry it out. However, they tend not to attempt or complete suicide when their suicide threats are most alarming, but later in the course of their illness.20 Nevertheless, most psychiatrists, irrespective of gender, judged their patients to be in reasonably good (mental) conditions, with improvement of, or no change in, symptoms at the last contact before the attempted suicide. This has generally been shown to be the case for suicide,21–23 too. One explanation for this scenario, observed in both attempters and suicides after documented clinical improvement (or stabilization), is that attention towards patients may be reduced as clinical conditions improve (as may the heightened sense of alert accompanying the previous crisis situation). Moreover, conditions may only partially improve, leading to increased suicidal risk as behavioral disinhibition rises in the presence of suicidal ideation.24,25 Hence the importance of not relaxing supervision and clinical care just when patients seem to be getting better.
One third of our younger psychiatrists, with less experience but probably more enthusiasm and emotional involvement, judged patients' conditions to have worsened. This judgment did not, however, result in any change in the drug regimen or in any other clinical intervention (e.g. hospitalization). This may be due to less training on the potential suicidal development of certain clinical conditions. Conversely, the more senior psychiatrists with more expertise, but who were possibly more detached and caught up by routine or experiencing burn-out, did not report any change at the last appointment. In any event, there is scant evidence in the literature of a relationship between physician personality traits or demographics and suicide inquiry.26
The female psychiatrists in our sample, irrespective of age, more frequently changed the drug regimen at the last contact than did their male colleagues, even when there was no difference in how they judged their patients' mental conditions.
Evaluation of suicide preventability
Our physicians also judged most of the intervening suicide attempts to be unpreventable. This is consistent with the data reported by Appleby et al.2 where the percentage of completed suicides deemed preventable was between 13% and 21%. The fact that psychiatrists frequently judged attempts to be unpreventable might also be explained by conscious or unconscious underestimation of patients' real suicide risk. However, the commonly-held belief that suicidal acts are somehow inevitable deserves some comment. While it is reasonable to assume that some attempted suicides are really unforeseeable and that psychiatrists do not possess reliable strategies to prevent suicidal behavior, assessment of preventability cannot be separated from assessment of suicidal risk and developing a better approach to risk assessment (more specific parameters and more attention). Accordingly, therefore, therapists must monitor the presence of suicidal ideation throughout the entire care program, for example using scales to measure suicidality. Monitoring should be carried out on a routine basis at each psychiatric evaluation and not just during crises with frank signs and symptoms of psychopathological breakdown. This is the approach recommended by all suicidology manuals. Nonetheless, as Lewis27 pointed out, a great deal of research has demonstrated the persistent ubiquity of error in healthcare, wide and unjustifiable variations in practice, and the minimal impact of decision aids such as clinical practice guidelines.
To our surprise, few of the psychiatrists involved in this study were using the ‘no-suicide contract’ (eight out of 28 psychiatrists), a widely recommended tool for both suicidality assessment and intervention purposes.28,29 It can be used in different ways, as in a strictly ‘behavioral’ setting (i.e. to inhibit behavior) or in order to protect the physician from malpractice lawsuits. This contract may also be a powerful cohesion and cooperation factor that can improve the therapeutic alliance. For example, the no-suicide contract may contribute to improving suicidal care, as it enables patient and therapist to make suicide prevention a focal issue within the therapeutic plan. It is worth pointing out that giving patients the opportunity to ‘reveal’ their suicidal ideation to the therapist is a way of enhancing communication and the therapeutic alliance.
Nonetheless, adoption of this technique should be viewed as only one part of a comprehensive suicide-risk-management strategy.30 In this respect, a minority of the psychiatrists in our survey provided general advice on improving preventability. None of them suggested using other simple suicide prevention strategies that are widely documented in the literature (although their efficacy remains controversial), as the ‘green card’ or ‘crises lines’ between one clinical assessment and the next.31,32 Finally, the psychiatrists at the MHD usually had 20 min per patient appointment, which is probably too short a time to use any no-suicide strategies.
In preventing AS, our study emphasizes the importance of suicide ideation monitoring, which helps the patient express suicidality to the therapist, and helps the latter take any necessary action. Attention by the psychiatrist to this topic is a fundamental step towards reducing the risk of suicide/attempted suicide because, as emphasized by Norko and Baranoski, ‘once someone has been placed in a high-risk group, his or her risk usually decreases’.33
Building awareness about available preventive strategies (such as the ‘no suicide contract’, the green card, etc.) may extend psychiatrists' clinical know-how by providing them with more appropriate strategies for each individual patient. It is also worth stressing that in well-conducted, controlled, randomized studies, no scheme has ever proved effective in reducing suicide rates.13 It is therefore likely that managing multidetermined phenomena, such as suicide and AS, requires a complexity of treatments and therefore strategies.
This study presents some methodological shortcomings. Firstly, information was collected retrospectively by the participating psychiatrists. This may constitute a report bias for some variables, as suicide risk behavior evaluation (i.e. the psychiatrists were asked to assess their approach in the period prior to the AS). Nevertheless, it provides a panorama of the opinions and beliefs of psychiatrists operating in a healthcare setting. As in all studies that collect personal, nonobjective opinions, we were unable to review the reliability of what respondents reported in their questionnaires. Furthermore, a prospective study aimed at monitoring clinicians' ability to predict suicide risk in patients over time could draw more attention to risk and very likely improve predictive ability. Instead of being observational, the study would then become an assessment of intervention efficacy.
Another limitation stems from the fact that while the psychiatric diagnoses were longitudinal, they were not based on structured interviews. The stability of a diagnosis over time is a criterion of reliability of any diagnostic approach and has important therapeutic implications. The importance of monitoring and confirming the diagnosis over time is prompted by evidence that the temporal consistency of mental disorders is poor, ranging from 28% for specific personality disorders to 70% for schizophrenia.34
In addition, as our sample was composed of suicidal patients already in care at an MHD, our results are not representative of the general suicidal population.
The sample was relatively small, the study unicentric and conducted over a 12-month period, with no control group of non-suicidal people, so our results cannot be generalized to other MHD in other countries. The organization of psychiatric services varies, even considerably, not only between different countries, but also in different areas of the same country.
Further, we cannot be sure that our sample covered all of the people who attempted suicide within the catchment area. Nonetheless, the method used to survey the AS population was such that the escape rate was probably very low. Moreover, people who attempted suicide monitored in a similar geographical area yielded similar AS rates in a previous study.35 It is also plausible that some subjects were in private care before attempting suicide.
We thank all the psychiatrists and staff of the MHD who made this study possible.