When treating patients at risk for suicide, clinicians often struggle to identify signs, symptoms, or precipitating events that might afford opportunities for them to intervene. Clinically, we remain largely unable to accurately distinguish between patients who will attempt or die by suicide and patients who will not [1, 2]. Clinicians’ emotional responses to patients (broadly speaking, their countertransference) have long and increasingly been recognized as an important factor in treatment outcome [3, 4], however they have received relatively little attention in the literature on suicidal patients. Rather, current research on acute suicide prediction has focused largely on warning signs that are patient-dependent, such as precipitating events [5–9], behavior changes [10, 11], or intense affective states [12–19]. Yet, even though they are easily identified retrospectively, such findings may be difficult to utilize clinically; these markers may be masked and/or minimized by the patient, or misattributed/misinterpreted by the therapist [20–22], and in some cases overzealous efforts at intervention, such as those that prematurely push an unready patient towards independence, even appear to precipitate patient suicide [5].

A potential factor contributing to these difficulties, beyond the general difficulty of predicting human behavior, and external constraints of the current mental health care system (e.g. [23]), may lie in clinicians’ own emotional responses to the suicidal patient. While clinical judgment is ultimately a conscious process, the suicidal patient elicits powerful responses that may not become directly conscious [4, 24]. Indeed, neuroimaging studies suggest activation of brain regions primarily involved in unconscious processing during emotional as compared to cognitive empathy tasks [25]. Without (and even with [26]) tremendous experience, unaided conscious integration of unconscious emotional responses is likely to fail. A systematic assessment of these responses, however, has the potential to ameliorate the inherent distortions of the clinician’s judgment without discarding the data inherent in his or her interpersonal experience with the patient.

Clinician-focused research supports distinctive patterns of reaction to various patient types [27–29], and there is a relatively large body of literature examining clinicians’ reactions following patient suicide (e.g., [22, 30–32]) and unexpected death [33] which observe prominent reactions of grief and mourning on the one hand [30], and guilt and anger on the other [32, 34]. Similarly, clinicians confronted with patients’ desire for death in studies of physician assisted suicide (also only qualitative), elicited anxious, helpless, and overwhelmed responses most prominently [35]. Clinicians’ emotional responses to suicidal patients have not been the subject of many research studies. Since Maltsberger and Buie’s seminal 1974 paper [24], which elaborated an array of emotional experiences and behaviors rooted in different defense responses to negative countertransference towards the suicidal patient, only a few empirical studies have been conducted. These retrospective clinical studies have focused almost exclusively on countertransference hate and/or negative countertransference in general, finding feelings of anxiety and hostility as those most prominently elicited by suicidal patients [34, 36]. The studies share a common conclusion that emotional responses must be recognized and acknowledged, and present evidence that the management of the clinicians’ emotional response is correlated with therapeutic outcome [3, 22, 37]. Quantifying clinicians’ emotional responses may thus potentially enhance suicide risk assessment.

The present preliminary study, though conducted retrospectively, assessed clinicians’ reported emotional responses toward their patients in the encounter preceding their suicide attempt, completed suicide, or unexpected (non-suicide) death, with a focus on quantifying differences in the patterns of clinician response to patients with differing levels or types of suicidality. The goal was to identify potential significant differences in clinicians’ emotional responses to the patients that were at imminent risk for suicidal behavior, compared with those who were not. Ultimately, a thorough understanding of characteristic emotional responses to imminently suicidal patients might allow clinicians to better recognize those responses to their patients that might interfere with taking appropriate measures to prevent imminent suicidal actions, or that in themselves may serve as warning signs of imminent suicidality.

An anonymous web-based survey (implemented through the surveymonkey.com website) was distributed to psychiatrists, psychologists, and social workers at the Beth Israel Medical Center in New York City via a department-wide email message requesting participation including the link to the anonymous survey. Participation occurred on a voluntary basis and participants had the ability to discontinue at any time. Participants were informed of the nature of the study in the email message inviting them to complete the survey. The study was approved by the Beth Israel Medical Center Institutional Review Board.

The survey consisted of the Therapist Response/Countertransference Questionnaire (CQ) – a 79-item self-report measure designed for clinicians which provides a validated instrument for assessing countertransference patterns in the psychotherapeutic setting [29], as well as questions regarding the demographic and clinical characteristics of the clinicians and the patients they reported on. The CQ has eight defined subscales (found to be independent of clinicians’ theoretical orientation): overwhelmed-disorganized (coefficient alpha = 0.90) “marked by items indicating a desire to avoid or flee the patient and strong negative feelings, including dread, repulsion, and resentment”, helpless-inadequate (coefficient alpha = 0.88), “describing feelings of inadequacy, incompetence, hopelessness, and anxiety”, positive (coefficient alpha = 0.86), “indicating the experience of a positive working alliance and close connection with the patient”, special-over-involved (coefficient alpha = 0.75), “describing a sense of the patient as special, relative to other patients, and … ‘soft signs’ of problems in maintaining boundaries”, sexualized (coefficient alpha = 0.77), “describing sexual feelings toward the patient or … sexual tension”, disengaged (coefficient alpha = 0.83), “describing feeling distracted, withdrawn, annoyed, or bored”, parental-protective (coefficient alpha = 0.80), “describing a wish to protect and nurture the patient in a parental way… beyond normal positive feelings”, and criticized-mistreated (coefficient alpha = 0.83), “describing feelings of being unappreciated, dismissed, or devalued” [29]. The CQ was used to assess countertransference in clinicians across four different patient categories: suicide completers, high-lethality suicide attempters (as indicated by clinical judgment and/or necessity for hospitalization), low-lethality suicide attempters (as indicated by clinical judgment), and patients who suffered sudden (unexpected) non-suicide death. The order of patient category presentation was randomized for each respondent. In each patient category the clinicians were prompted to fill out the questionnaire based on their experiences in regard to “the patient you remember best” in the last session preceding their suicide attempt or death. This prompt was chosen to elicit what, in the absence of prospective data, should be the most reliable. [38] If a clinician reported having treated a patient in more than one category, a separate CQ was filled out for each patient category individually. Clinicians were instructed to rate each item on the questionnaire as 1, 3, or 5, based on the extent to which it was true in their work with the patient in question; 1 = not true at all, 3 = somewhat true, and 5 = very true.

To the best of our knowledge, this is the first published study to provide a quantitative comparison of clinician responses to acutely suicidal patients versus non-attempters and to patients who made high lethality versus low-lethality suicide attempts.

Such investigation is important, as problems in the management of countertransference (or emotional reactions in general) to patients may hamper treatment efficacy and even contribute to patient suicide in a small but significant proportion of cases [5, 36]. To date though, the literature has focused almost entirely on the development of qualitative treatments of the subject. A thorough literature search using the PsycINFO database resulted in our conclusion that there are no analogous studies in the literature. (Searches conducted using varied combinations of terms including “countertransference”, “suicide”, “therapist response”, “clinician response”, “predict”, “prevention”, “comparison”, and “quantitative” identified no peer-reviewed publications reporting on quantitative comparisons of clinician responses to suicidal versus non-suicidal patients or of patients with differing levels of suicidality). The only quantitative comparative work we have been able to find on the subject has been a small series of unpublished dissertations, which found no significant differences in negative therapist responses to suicidal versus “difficult” non-suicidal patients [42]. While rich qualitative data are an essential starting point, this preliminary study aimed to pilot a much-needed quantitative and comparative approach using a validated instrument and easily replicable quantitative methodology.

This study found that clinicians treating imminently suicidal patients recalled, on average, moderately positive feelings towards these patients (though less so than for non-attempters), with higher hopes for treatment, while finding themselves more overwhelmed, distressed by, and, at low levels, avoidant of them. Further, we found that the specific paradoxical combination of hopefulness and distress/avoidance was a significant discriminator between suicidal patients and those who had unexpected non-suicide deaths, and cross-validated classification by discriminant analysis remained statistically significant both when a past history of suicide attempt was present and when it was not. This finding of ‘paradoxical response’ is consistent with the higher scores observed on the “overwhelmed-disorganized” subscale of the CQ in clinician recollections of their encounters with suicide attempters.

In our second comparison, we found no clear evidence of differences between clinicians’ responses in encounters with patients preceding either completed suicides or highly lethal suicide attempts and their responses in encounters preceding low lethality suicide attempts. Despite a trend towards a slightly more positive emotional responses overall, clinicians also recalled experiencing more sadness in encounters with patients preceding either successful or highly lethal suicide attempts than in encounters preceding low lethality suicide attempts. This difference in recalled sadness was found to be a modest discriminator between patients that went on to exhibit high and low lethality suicidal behavior. It is worth noting, however, that this difference appears attributable specifically to recalled responses to patients with a history of previous attempts; among these patients sadness in session was a significant discriminator of attempt lethality while among first-time attempters clinicians’ experience of sadness in the session prior to suicide attempt did not differentiate between lethality levels. This finding has not been supported or opposed in the literature, as the difference in emotional response to high and low lethality suicide attempters has not previously been explored. Further, interpretation is limited by significant risk of type-1 error given the small n’s and multiple comparisons involved.

Thus our findings, while grossly consistent with the qualitative literature findings of negative responses to suicidal patients [34, 36], differed in the important respect that the levels of recalled negative reactions to patients prior to their suicide attempts were, on average, fairly low, and even when significant, the magnitude of the differences in positive and negative responses elicited by suicidal versus non-suicidal patients was small. Maltsberger and Buie [24] were the first to describe in detail the negative countertransference (“countertransference hate”) that clinicians may experience in response to suicidal patients, and provided a theoretical framework which might account for our quantitative findings. First, as noted, we found that clinicians recalled fairly low levels of negative feelings towards their suicidal patients, though positive response was attenuated compared to non-attempters. This finding may be consistent with their predictions of repression of “countertransference hate”. On the other hand, our findings of distress and self-directed negative feelings combined with paradoxical hopefulness may be consistent with their predictions of turning of the countertransferential hate against the self and of reaction-formation against it, respectively. Indeed, our findings seem to suggest that the defense mechanisms described by Maltsberger and Buie may operate in concert.

Our findings point to the potential clinical utility of self-assessment of emotional response in the treatment of suicidal patients. This is a matter of some importance as both Modestin [36] and Marcinko et al., [22] have used observational evidence to support the theory that emotional responses to suicidal patients that are not properly managed can have harmful consequences. The latter group concluded that negative emotional response probably contributes to or correlates with negative patient outcomes [22], while Modestin, further indicates how the failure to control these reactions (hostility, hate, and aggressiveness in particular) may in some cases help push patients to suicide [36].

We should note, however, that while both emotional responses and judgments of suicide risk reside in the clinician, they are not the same. Indeed clinical judgment has been found to be a poor predictor of critical patient behavior such as suicide [1] and violence [43]. While clinical judgment is ultimately a conscious process, emotional responses may not become directly conscious [4, 24]. Thus systematic assessment of these responses, even in using self-report measures may reveal patterns generated by the clinician’s unconscious processes such as the “paradoxical hopefulness” identified using discriminant analysis. Quantitative self-report assessment may thus reveal data inherent in the clinician’s interpersonal experience with the patient that could potentially augment suicide-risk assessment.

We find preliminary quantitative evidence consistent with Maltsberger and Buie’s theory of countertransference hate in the treatment of suicidal patients. Though our study does not speak to the ability of the differences in response to influence or predict a patient’s outcome, it is the first to quantify the differences in clinicians’ emotional responses to suicidal patients versus non-attempter patients. Our findings thus provide a starting point for further research that may change the way that clinicians assess their suicidal patients’ acute risk, and may justify further research on the use of the CQ or other conceptually similar scales as predictors of suicide risk.

Jessica Briggs is co-first author.