Long-term outcome studies of patients with borderline personality disorder have documented a high rate of suicide completion. In a 15-year follow-up of 200 patients treated at New York State Psychiatric Institute, Stone (2) found a 9.5 percent rate of suicide. A 15-year follow-up of 162 patients treated at a general hospital in Montreal found a rate of 9 percent (3,4); after 27 years, the rate of suicide completion in this group increased to more than 10 percent (5). A ten-year follow-up of 70 patients treated at a general hospital in Toronto also reported a 10 percent rate of suicide completion (6). Despite socioeconomic differences among the cohorts, these results are strikingly convergent. In McGlashan's 15-year follow-up of 81 patients treated at Chestnut Lodge (7), only 3 percent completed suicide, but McGlashan believed that the higher rates in other studies are more representative (personal communication, 1991). Thus one in ten patients with borderline personality disorder can be expected to complete suicide, a rate similar to those for patients with schizophrenia (8) and patients with major mood disorders (9).
Although suicide threats and attempts peak early in the course of illness in borderline personality disorder, when patients are in their 20s, this time frame is not necessarily when most completions occur. The mean age of those who completed suicide in the New York study was 30 years (2), and in the Montreal study it was 37 years (5). Thus patients with borderline personality disorder may not kill themselves when suicide threats and attempts most alarm therapists, but later in the course of their illness. Completions occur when the patient does not recover and when treatments have been unsuccessful.
Community studies have shown that rates of suicide peak between the ages of 18 and 30 years, and a diagnosis of borderline personality disorder can be made in more than a quarter of all cases of completed suicides (10). However, in these surveys, the majority of individuals who committed suicide were male, and most completions were first attempts. In contrast, most patients with borderline personality disorder who kill themselves are female, and most completed suicides in this group occur after multiple attempts.
Stone (2) found that among patients with borderline personality disorder, those who had comorbid substance abuse or major depression were more likely to complete suicide than those without these comorbid disorders. Three studies found that the number of previous attempts was predictive of completed suicide (11,12,13). Although a higher level of education was associated with completed suicide in a study in Montreal (11), this association was not found in a study in Norway (12). Studies of childhood experiences have not identified consistent predictors (11,12). None of these factors account for much of the variance in outcome, a general problem for research on suicide. Completed suicides are relatively rare and are difficult to predict with any accuracy (14).
Acute and chronic suicidality
The intensity of suicidality among patients with borderline personality disorder varies over time. When patients are acutely suicidal, they may also meet criteria for a major depressive episode. However, managing depression may be problematic, because the affective symptoms of patients with borderline personality disorder are phenomenologically different from those of patients with mood disorders and do not respond to antidepressants in the same way (15,16). Moreover, early-onset dysthymia is more characteristic of borderline personality disorder than are discrete episodes of depression (17).
Borderline personality disorder is characterized by affective lability, which has been hypothesized to be a primary factor underlying the disorder (18,19). This trait involves a high level of sensitivity to the environment, so that mood can shift over the course of days, or even hours. High levels of reactivity also explain why patients with borderline personality disorder are sensitive to changes in the social environment, as suggested by evidence that this disorder has become more prevalent in recent decades (20,21).
Suicide completers and suicide attempters, although overlapping groups, are distinct clinical populations (22). Completers tend to be older and male, to use more lethal methods, and to die on the first attempt. Attempters tend to be younger and female, to use less lethal methods, and to embark on a "suicidal career" (22). Completers are less likely to seek treatment before attempting suicide. In a psychological autopsy study of young adult suicides (10), fewer than a third had been in therapy at the time of their death, fewer than half had seen a therapist during the previous year, and a third had never been evaluated. Similar findings have been reported for suicide among patients with a variety of psychiatric diagnoses (23).
In contrast, chronically suicidal patients are treatment seeking. Persons with a history of repeated suicide attempts constitute a large proportion of visitors to psychiatric emergency departments and to clinics (24). Yet, as documented by follow-up studies (1,3), only a minority of completed suicides among patients with borderline personality disorder occur during active treatment.
The clinician who treats a chronically suicidal patient must assess acute risk. But if a patient repeatedly entertains suicidal ideas and frequently threatens suicide, at what point should the clinician be alarmed? No sharp line exists between chronic and acute suicidality. Clinical judgment can depend as much on the therapist's anxiety as on objective risk.
Research on completed suicide is paradoxically reassuring. Empirical evidence that clinical interventions have any systematic effect on completion is notably lacking (25). In large-scale studies of suicide, researchers can identify factors associated with higher risk and, given large samples, can identify statistically significant associations. But algorithms for guiding prediction have produced too many false-positive findings to be clinically useful (14,26). It is impossible to predict suicide in the case of an individual patient.
Suicidality among patients with borderline personality disorder can best be seen as a way of communicating distress. The object of this communication can be a significant other, a therapist, or both. Paradoxically, threats of suicide usually reflect some degree of attachment and involvement. Suicide completion, in contrast, is associated with a loss of connection.
Patients with borderline personality disorder are well known for taking overdoses or self-mutilating after quarreling with their intimates. Clinicians know that these attempts tend to be ambivalent, often containing an element of "protection." If the patient takes an overdose, the clinician commonly observes that someone has been telephoned, that another person was present when the attempt was made, or that a friend or relative was expected to arrive. Occasionally, when these attempts at protection do not work, the patient dies, more by accident than by intention. In most cases, the attempter is saved and brought to the hospital.
Although self-mutilation is common among patients with borderline personality disorder, it does not belong in the same category as suicide attempts. Although wrist cutting can precede serious attempts, its main function is to relieve dysphoria (19,27).
Hospitalization is expensive. Hospital resources should be used to carry out specific aspects of treatment plans that can be provided only in the hospital setting. Patients should be admitted for indications supported by empirical evidence.
For acutely psychotic patients, the rationale for admission is clear. Nor would anyone doubt the importance of hospitalizing suicidal patients with a classic melancholic or psychotic depression. In such cases, the efficacy of treatment is well established, and good results are often obtained within weeks. It makes sense to hospitalize patients to carry out these interventions and to institute suicide precautions to make sure the patient does not die before treatment takes effect.
In contrast, the management of chronic suicidality requires a different set of principles. Unlike patients with melancholic or psychotic depression, patients with borderline personality disorder rarely commit suicide while in the hospital. Although some suicide completions occur soon after discharge (13), most patients remain alive but chronically suicidal. Neither biological treatments nor short-term containment strategies provide a "quick fix" for this problem. Therapists may find it necessary to tolerate suicidality over extended periods (28,29).
The most common reasons for the hospitalization of patients with borderline personality disorder are psychotic episodes, serious suicide attempts, suicidal threats, and self-mutilation (30). It clearly makes sense to admit a patient for treatment of a brief psychosis. Specific treatment (neuroleptic medication) is available to control psychotic symptoms. The admission of a patient after a life-threatening suicide attempt also makes sense. At the very least, it provides a break to the family and to the outpatient therapist. Even if no active treatment is conducted in the hospital, a brief admission provides an opportunity to assess precipitating factors and review treatment plans.
The value of hospitalization is much less clear for suicidal threats, minor overdoses, or self-mutilation. Once hospitalization is introduced, admissions can become repetitive. If, as so often happens, the patient becomes suicidal again shortly after discharge, little has been accomplished. A patient who eventually recovered from borderline personality disorder illustrated this problem with the following admonition: "Do not hospitalize a person with borderline personality disorder for any more than 48 hours. My self-destructive episodes—one leading right into another—came out only after my first and subsequent hospital admissions, after I learned the system was usually obligated to respond" (31).
Hospitalization is a two-edged sword. When patients with borderline personality disorder are admitted, they enter an environment that reinforces the very behaviors that therapy is trying to extinguish. Procedures developed for treatment of acute suicidality among patients with mood disorders are rarely appropriate for chronic suicidality. Most clinicians recognize the scenario in which a patient with borderline personality disorder escalates suicidal or self-mutilating behaviors in the hospital. Two mechanisms account for this phenomenon. First, for patients with poor social supports, a week on a ward, or even a night in an emergency department, provides a reinforcing level of social contact. Second, the environment of a psychiatric ward acts as a reinforcer, since patients who cut themselves or who carry out parasuicidal actions receive more, not less, nursing care.
Linehan (19), applying the principles of behavioral psychology, has discouraged admission of patients with borderline personality disorder, tolerating at most an overnight hold. Dawson and MacMillan (32) took an even more radical position, arguing that such patients should never be hospitalized. Although one should never say never, admissions can be ineffective and counterproductive.
When therapy spirals out of control and the clinician needs the help of a specialized team, partial hospitalization may be useful. Unlike full admission, day treatment has been empirically demonstrated to be effective in two cohorts of patients with borderline personality disorder (33,34). Partial hospitalization is particularly effective in the treatment of this disorder because of its highly structured program. Patients with borderline personality disorder typically show greater pathology in an unstructured environment (35). In a partial hospital program, where activities are scheduled for every hour, little time remains to slash one's wrists. Regression is further limited by the fact that the patient goes home at night. With no evidence that full hospitalization prevents suicide completion, suicidal risk is not a contraindication for day hospital treatment.
Many clinicians are aware of the limited value of hospitalizing patients with borderline personality disorder. Yet, because of fear of litigation, they feel compelled to hospitalize a patient who threatens suicide.
Most mental health professionals either have had to endure the completed suicide of a patient or know of such cases among their colleagues. Surveys have shown that a patient's death by suicide occurs at least once in the career of 50 percent of psychiatrists (36) and 20 percent of psychologists (37). These general figures mask the relationship between prevalence and specific clinical settings. A therapist with an office practice may avoid completed suicides among his or her patients, but in hospitals or community clinics, it is hard to find a practitioner who has never lost a patient.
Suicide is the leading cause of lawsuits against mental health professionals, accounting for 20 percent of cases (38). Yet data drawn from various jurisdictions around the United States (39,40) suggest that only a very small fraction of completions lead to litigation. Moreover, only a minority (about 20 percent) of such lawsuits end with a decision against the practitioner. Most practitioners will never have to face litigation, and those who do can expect a decision in their favor. Nonetheless, malpractice litigation causes enormous stress. The anticipation of this possibility leads practitioners to treat suicidal patients with great caution.
The vast majority of cases brought to the courts have involved inpatients treated for major axis I disorders (38). Lawsuits usually focus on whether patients were discharged too early, not on whether they should have been hospitalized in the first place. In contrast, litigation after a completed suicide of a chronically suicidal patient is rare.
Lawsuits in which clinicians are found liable are almost never based on the fact of suicide alone (38). Most courts know that suicide cannot always be prevented and do not routinely hold clinicians responsible when it happens. Liability depends on gross clinical misjudgments, the failure to assess patients carefully, and the absence of adequate records documenting the management plan (40). Therapists must make notes on chronic risk, on reevaluations when patients threaten suicide, and on the rationale for avoiding hospitalization.
Involving the family in the treatment of chronically suicidal patients makes litigation less likely. Packman and Harris (40) have recommended that therapists inform a suicidal patient early in treatment that the family will be contacted if the patient is at risk. This strategy should be presented as a necessary parameter of treatment that need not involve any breach of confidentiality. Such procedures closely parallel the relationship of professionals with the families of psychotic patients.
The goals of meeting with family members are to inform them of the rationale behind treatment, to educate them about the management plan, and to obtain cooperation with therapy. Family members, who themselves have had to endure a patient's suicidality, will feel supported by being brought into such an alliance. If there is an unfavorable outcome, they will have less reason to feel angry and excluded. Finally, if a suicide occurs, the therapist can provide "postvention" for the family (40), meeting with relatives soon after the death of a patient to help them deal with bereavement.
An excessive focus on suicide prevention can prevent therapists from doing their job. When clinicians feel forced to do almost anything to prevent suicide completion, the therapeutic relationship becomes characterized by "coercive bondage" (41,42) and the quality of the patient's life becomes compromised by overzealous concern. Ultimately, treatment must help patients solve problems. When clinicians spend too much time worrying about suicide completion, this treatment process becomes derailed. Clinicians must focus on the task at hand and respond to the distress behind the threat.
To quote again the testimony of a patient who recovered from borderline personality disorder: "When you as a service provider do not give the expected response to these threats, you'll be accused of not caring. But what you are really doing is being cruel to be kind. When my doctor wouldn't hospitalize me, I accused him of not caring if I lived or died. He replied, referring to a cycle of repeated hospitalizations, 'That's not life.' And he was 100 percent right!" Or, in the words of one expert on suicide: "We cannot afford to be so afraid of litigation as to deny our patients the right to learn to live."
Managing chronic suicidality in outpatient therapy
Chronic suicidal ideation "goes with the territory" of borderline personality disorder. These symptoms remit late in the course of treatment. Clinicians who treat patients with borderline personality disorder are advised to accept these symptoms and to get on with the task of addressing their underlying causes.
Once the communicative function of suicidality is understood, treatment sessions can become calmer and more focused. In outpatient psychotherapy, suicidal ideas and attempts require behavioral analysis (19). First, the therapist listens to the emotional content of the suicidal ideation and validates dysphoric feelings that tempt the patient to act impulsively. The second step is to identify the circumstances leading the patient to experience dysphoria. The third step is to establish a dialogue with the patient to develop alternative solutions to the problems lying behind these emotions. Once the causes of suicidality are established, therapy can focus on change. These basic principles have been clearly described in dialectical-behavioral therapy (19) and are used by clinicians of various theoretical orientations (43).
Linehan (19) has emphasized that therapists should be aware that their responses to suicidality can increase or decrease the frequency of the patient's suicidal thoughts or attempts. In dialectical-behavioral therapy, patients are positively reinforced—through telephone contact—for communicating suicidality as a problem to be solved. In contrast, they are negatively reinforced—through temporary loss of access to sessions—for cutting themselves or taking an overdose. Although patients can go to the emergency department, it is not recommended.
Linehan has recommended dealing with suicidality as the first priority in a session. Paradoxically, her aim is not to prevent completion. Rather, because the patient cannot discuss other important issues until this issue is dealt with, the technique negatively reinforces suicidal ideation. In Linehan's original clinical trial, which lasted for a year and involved 22 patients, hospitalization was discouraged, and only one suicide occurred (44).
Clinicians need to respond to suicidal thoughts empathically, while avoiding overly anxious questions about intent. When a patient is suicidal, something similar to the following response may be most appropriate: "You must be feeling particularly upset to be thinking along those lines. Let's figure out what is making things worse and see if we can find a way of dealing with the problem."
As shown in prospective follow-up studies of patients with borderline personality disorder (45), suicidal ideas wax and wane over the course of treatment, depending on concurrent life events. As demonstrated in several studies of psychotherapy for patients with borderline personality disorder (44,46,47,48), when treatment begins to work, suicidality gradually drops out of the clinical picture.
Some patients with borderline personality disorder actually need to be suicidal (28)—when they feel no power over life, they retain the choice of death. Clinicians should therefore be cautious about removing this coping mechanism too soon. Patients with borderline personality disorder may remain suicidal until they feel in control of their fate. The knowledge that they can choose to die allows them to go on living.