The impact on mental health practitioners of the death of a patient by suicide: A systematic review

There is a growing body of research investigating the impact on mental health professionals of losing a patient through suicide. However, the nature and extent of the impact is unclear. This systematic review synthesizes both quantitative and qualitative studies in the area. The aim was to review the literature on the impact of losing a patient through suicide with respect to both personal and professional practice responses as well as the support received. A search of the major psychological and medical databases was conducted, using keywords including suicide, patient, practitioner, and impact, which yielded 3,942 records. Fifty-four studies were included in the final narrative synthesis. Most common personal reactions in qualitative studies included guilt, shock, sadness, anger, and blame. Impact on professional practice included self-doubt and being more cautious and defensive in the management of suicide risk. As quantitative study methodologies were heterogeneous, it was difficult to make direct comparisons across studies. However, 13 studies (total n = 717 practitioners) utilized the Impact of Event Scale, finding that between 12% and 53% of practitioners recorded clinically significant scores. The need for training that is focused on the impact of suicides, and the value placed upon informal support was often cited. The experience of losing a patient through suicide can have a significant impact on mental health professionals, both in terms of their personal reactions and subsequent changes to professional practice. The negative impact, however, may be moderated by cultural and organisational factors and by the nature of support available.


| INTRODUCTION
Approximately 135 people, on average, are exposed to each death by suicide (Cerel et al., 2016), where exposure is defined as knowing or identifying with the deceased, and it is estimated that 35% of the population experience moderate to extreme emotional distress as a consequence of a suicide death during their lifetimes (Feigelman, Cerel, McIntosh, Brent, & Gutin, 2018).Bereavement by suicide is also a recognized risk factor for suicide attempts (Pitman, Osborn, Rantell, & King, 2016).Among those exposed to loss of life by suicide are mental health practitioners.In the United Kingdom, it is estimated that 27% of those who take their own lives are individuals who had been in contact with mental health services in the 12 months prior to death (National Confidential Inquiry into Suicide and Homicide by People with Mental Illness [NCISH], 2014).For mental health practitioners, the tragedy of a patient's suicide has been described as a rare event, although one that is likely to be experienced at some point in their careers (Foley & Kelly, 2007) and one that is likely to evoke strong emotional distress (Feigelman et al., 2018).One study found that 58% (n = 704) of responding mental health practitioners working in institutions had experienced a patient suicide (Castelli Dransart, Heeb, Gulfi, & Gutjahr, 2015).
Recognizing and understanding the impact of this are necessary precursors to identifying how best to support health professionals who experience the suicide of a patient.This matters on a personal level, in order to ameliorate the level of distress and prevent longterm effects and in terms of professional practice, that is, ensuring risk is effectively managed and at a service level, for example, protecting against staff burnout and promoting staff retention.A growing body of studies has investigated the impact on mental health professionals of losing a patient through suicide, although the majority of these studies have investigated the effects on psychiatrists or psychologists (Foley & Kelly, 2007;Lafayette & Stern, 2004).In a previous literature review, Lafayette and Stern (2004) concluded that mental health professionals' reactions may be strong or overwhelming and may be similar to grief responses.They also highlighted that studies often differentiated between the impacts on personal life and those on professional practice.Foley and Kelly (2007) concluded that the impact on mental health professionals could be pronounced, prolonged, and profound.They noted, among other reactions, signs of stress, guilt, symptoms of post-traumatic stress disorder (PTSD), more defensive approaches to risk, and consideration of retirement.They also commented that the perceived risk of patient suicide as well as suicide itself contributed to exhaustion and depression among psychiatrists.
The latest synthesis of evidence was a literature review of studies completed up to 2012 (Séguin, Bordeleau, Drouin, Castelli-Dransart, & Giasson, 2014), which focussed on studies that had employed a psychological well-being outcome measure and concluded-in contrast to other reviews-that stress reactions or affective-related symptomatology were minimal.There was, however, an impact on the way professionals conducted their clinical assessments and reached subsequent treatment decisions (Séguin et al., 2014).
To our knowledge, there have been no systematic reviews of the research into the impact of losing a patient to suicide on mental health professionals more broadly.The (non-systematic) narrative reviews that have been published present mixed evidence for both the professional and personal consequences for mental health professionals of losing a patient through suicide, although the studies themselves do seem to converge on the idea that there is a notable effect on both areas.A thorough synthesis of all the evidence is important if organizations are to be encouraged to respond proportionately in supporting practitioners.We therefore undertook a systematic search of both quantitative and qualitative studies.Specifically, our three main aims were to synthesize the research evidence on (1) the impact of a patient's suicide on both personal (emotional/coping) and professional practice (confidence/behaviour/attitude) responses; (2) the support received; and (3) the factors or interventions that help to minimize negative sequelae.

| Search strategy and screening of results
The protocol for this systematic review was registered on Prospero (registration number CRD42017052807).A keyword search of the major psychological and medical databases (Medline, PsychInfo, CIN-AHL, ERIC, and EMBASE) was conducted using the search terms detailed in Table 1.The screening process followed PRISMA guidelines (Moher, 2009).See Figure 1 for flowchart.

| Inclusion and exclusion criteria
The inclusion criteria were that (i) the study must be published primary research in the English language, and (ii) the studies must have reported on the impact on mental health professionals or teams of mental health professionals of having experienced the loss of a patient through suicide (either in active treatment or post-discharge).Studies were excluded if they were single case studies or personal accounts.
To be comprehensive, we included both qualitative and quantitative studies.

| Quality assessment
A quality assessment was performed on the included studies, and this was subsequently used to give a weighting to findings reported in the results section.A quality assessment framework was adapted from

Key Practitioner Messages
• The death of a patient by suicide can have a considerable, and lasting, emotional impact on mental health professionals most commonly manifested as guilt, blame, shock, anger, sadness, anxiety, and grief.
• The impact is comparable with that of other traumatic life events, and therefore, active monitoring of practitioners for symptoms of PTSD is recommended.
• There were notable impacts on professional practice including self-doubt and being more cautious and defensive in the management of suicide risk.
• More should be done to prepare and support mental health professionals for the event that they may lose a patient through suicide.
The quality assessment was carried out by the first author and then 13 (25%) were independently assessed by the third author.
There were discrepancies in rating 3 out of 92 (3%) individual assessment items, which equated to 3 differences in assigned quality ratings and yielded an intraclass correlation of k = 0.71.This compares well with the estimation of correlation of 0.8 made F I G U R E 1 Procedure for identifying, screening, and determining the eligibility of studies for inclusion in the review following pilot testing of the appraisal tool (Pace et al., 2010).The disagreements were resolved through discussion between the first and third authors (see Table 5).The data extraction was carried out by the first author using a data extraction form that had been agreed by all four authors.

| RESULTS
For the current synthesis, we focused on the following areas: personal responses of practitioners, impact on professional practice, variables associated with the extent of the impact on practitioners, and support that helps minimize negative sequelae.Fifty-four studies were included (see Table 3), 21 were quantitative studies, 16 were qualitative studies, and a further 17 studies adopted a mixed-method design.
The majority (15 studies) investigated the impact on psychiatrists, and 13 studies focussed on psychologists or psychotherapists, with a further six studies including both psychiatrists and psychologists.Other professional groups represented within this review were social workers (five studies), nurses (four studies), general practitioners (GPs) (three studies), and counsellors (two studies).Nine studies were based on a mixed group of mental health workers.In terms of geographical location, 23 studies recruited participants from North America, nine from the United Kingdom, seven from Ireland, 11 from elsewhere in Europe, two from Australia, and one each from Thailand and China.
Intervals between the death and the time of the study varied markedly both within and across studies; for example, the range in Alexander, Klein, Gray, Dewar, and Eagles (2000) was between 1 month and 20 years (median 3 years).
Although the methodologies in the quantitative studies were heterogeneous, making comparisons difficult, 13 studies (total n = 717) utilized the Impact of Event Scale (IES or IES-R, Horowitz, Wilner, & Alvarez, 1979;Weiss & Marmar, 1996) and reported that between 12% and 53% of their samples recorded clinically significant scores in the time immediately following the suicide (see Table 2).Of the other quantitative and mixed-method studies, Finlayson and Graetz Simmonds (2016) included 12 items in their study questionnaire to capture intensity of emotions and reported that sadness was rated most highly, followed by shock, helplessness, anger, and pain.Gulfi, Castelli Dransart, Heeb, and Gutjahr (2010) used nine items from an adapted Long-term Emotional Impact Scale and reported low to moderate impact overall, with increased sensitivity to signs of risk and increased anxiety when working with  suicidal patients being the highest rated items.The most frequently endorsed emotional reactions by Midence, Gregory, and Stanley's (1996) study of nursing staff were sadness and helplessness, followed by shock, and feeling guilty and angry.Ruskin (2004) employed a 25-item scale based on the DSM IV (American Psychiatric Association, 2013) criteria for acute stress or PTSD to assess psychiatrists and psychiatry trainees and reported that 22% of their sample met criteria for acute stress disorder and 20% met criteria for PTSD.Gibbons et al. (2019) found that the majority of their sample of psychiatrists (92%, n = 105) rated the effect on their emotional well-being above 50 on a 0-100 scale anchored at 0 = "no effect," 50 = "some," and 100 = "a very severe response," with an average rating of 66.Finally, Wurst et al. (2010Wurst et al. ( , 2011Wurst et al. ( , 2013) used a 100-mm visual analogue scale to measure emotional responses in three studies.The items measured were grief, guilt, anger, relief, shock, shame, disbelief, feeling offended, feeling insufficient, and an overall rating for total distress.Their three respective samples of therapists reported mean ratings of overall distress of 62 (Wurst et al., 2010), 63 (Wurst et al., 2011), and 63 (Wurst et al., 2013) with shock and sadness being the highest rated emotional responses.
Nine studies attempted to quantify the proportions of practitioners who were affected after the loss of a patient through suicide, although the thresholds used to define this were unclear, and the estimates showed considerable variation.In one survey (Landers, O'Brien, & Phelan, 2010), 97% (n = 139) of psychiatrists reported some effect on personal life following their most distressing suicide loss, and 87% (n = 124) reported some disturbance following their most recent suicide loss.Just 9.7% (n = 8) of the sample in Murphy et al.'s (2019) study reported no impact, with 55% (n = 46) describing some impact, 24% (n = 20) quite an impact, and 11% (n = 9) reporting a major impact.For Saini, Chantler, While, and Kapur (2016), 66% of their sample reported being affected in some way.With Courtenay and Stephens (2001), emotional impact was considered severe in 24% (n = 40), moderate in 51% (n = 85), and absent or minimal in 14% (n = 23) of the participants.In Alexander et al.'s (2000) study, 33% (n = 54) reported being personally affected (lowered mood, poor sleep, increased irritability), whereas Dewar, Eagles, Klein, Gray, and Alexander (2000) reported that for 31% (n = 15) of clinicians, the suicide had an adverse impact on some aspect of their personal lives.Hendin, Lipschitz, Maltsberger, Haas, and Wynecoop (2000) reported 38% (n = 13) as being severely distressed.In contrast to the above, Halligan and Corcoran (2001) found that more than 80% (n = 84) in their survey noted no personal effect other than for guilt feelings (35% [n = 37], experiencing feelings of guilt), and in a further study (Pieters et al., 2003), 84% (n = 66) of the participants did not recall an adverse impact on personal life.

| Impact on professional practice
Thirty-four of the studies reported on the impact upon professional practice experienced by practitioners.As summarized below, practitioners reported a greater focus on risk assessment.They also became more cautious and adopted a more defensive management of suicide risk.Others also reported increased self-doubt related to their own judgement and decision-making.

| Risk assessment
Fourteen studies described changes that relate to risk assessment.
• Ruminations, search for understanding.
• Depression and demoralization of ward team.
• Anxiety and worry-happen again, best efforts not enough.
• Incidents though distant in time were still having an influence on practice in the present.
• Impact across hospital (not just ward of occurrence).
• Attentiveness to risk assessment.
• More rigorously pursued policies.
• Greater use of containment methods.
• Time patient known to staff.
• Strength of relationship.
• Availability or lack of support and aftercare directly after.
• Perception of whether DSP could have been prevented.
• Culture of blame.
• Pressure of work preventing time to deal with feelings.
• Shock-all the more devastating.• Preoccupation about the suicide and how it could have been prevented.
• Problems with anxiety, guilt, insomnia, and loss of confidence.39% reported impact on professional practice: • Increased anxiety and difficulty in making decisions, particularly when this involved patients with recognized increased risk of self-harm.Most prominent reactions: • Failure to hospitalize an imminently suicidal patient who then died • A treatment decision the therapist felt contributed to the suicide • Contact with patient's family-source of comfort for some but painful for others Wang et al. (2016) Semi-structured interviews • Shock, helplessness, and sadness however markedly higher.
• Relationship to the patient • 52% greater awareness of risk Factors that contributed to distress: • 21% either an inpatient or had been recently seen and assessed • 20% the effect on the patient's family, when the patient was a parent of young children.• Confrontation with the body or involvement in resuscitating the patient was traumatic.
• A long and intense involvement with the patient or his/her family Distress and symptom levels: • Training > post-graduation.
• Less social integration to professional network > more.
• Previous personal experience of suicide in their family or friends did not affect their scores on the acute stress disorder, PTSD, and IES.No significant difference in overall distress experienced was observed between professional groups and at different levels of care.
Highly distressed therapists were • More likely to be female • Felt less supported by their colleagues and institution • Had a higher fear of lawsuit • Low mood, poor sleep, irritability, difficulties at home, preoccupation 69 (42%) changes in professional practice: • A more structured approach to the management of risk • Increased use of mental health legislation.
• More cautious and defensive approach and use of observation • Heightened awareness 24 (15%) • Considered taking early retirement • Colleagues and family or friends were the best sources of help and team and • critical incident reviews were also useful • Of 56 consultants who had been aware of publicity in the media 27 found the publicity extremely or moderately distressing.
• 33 of the 159 were extremely or moderately distressed at the prospect of litigation.

Semi-structured interviews
Personal life: • Shock, sadness, anxiety, feeling upset, grief, anger, and guilt, but these were mainly described as short-term effects.Professional life: • Increased vigilance and awareness of suicide risk • Being more proactive in assessment (e.g., asking about suicide ideation) and management (e.g., try harder to contact patient) Factors most associated with an impact upon personal life: • Being female • Men seemed more likely than women to report no effect on their functioning.
• Loss of sleep, irritability, and lack of concentration reported equally by men and women.
• Professional self-doubt was expressed more often by women.• Personal sadness, low mood, and self-doubt.13 (32.5%)professional lives affected: • Sense of helplessness, making decisions, reluctant to discharge patients.

(57.5) future management:
• More aware of risk assessment and the importance of meticulous documentation.
• More anxious to establish links with a patient's family, to admit patients to the ward, and to prescribe antidepressant medications Rated "very helpful" or "helpful": • 34 (85%) family • 29 (73%) peers • 25 (62.5%)team support • 26 (65%) team meetings • 10 (47.5%), patient's family Also: • Other consultants whose patients had died by suicide helpful • Need to ensure no blame or scapegoats at team meetings • 15 (37.5%) aware of publicity in the media.However, the majority of consultants did not find this distressing.• 6 (75%) described "shock" as their initial reaction.After shock, they mentioned (in order of frequency) guilt or shame, denial or disbelief, feelings of incompetence, anger, depression, a sense of being blamed, relief, and fear.

| Acceptance
The experience of losing a patient to suicide resulted in some practitioners being more accepting of suicide as a possibility (Kleespies et al., 1990;Linke et al., 2002) or having an increased awareness of the limits of their professional ability to prevent suicide (Goldstein & Buongiorno, 1984;Gulfi et al., 2010;Ting et al., 2006).Other studies detected increased acceptance of suicide itself (Finlayson & Graetz Simmonds, 2016) or feelings of understanding or acceptance of suicide as an option (Rothes, Scheerder, Van Audenhove, & Henriques, 2013;Ting et al., 2006).

| Suicide risk (for staff)
One study concluded that the experience of a patient's death may increase suicide risk for the practitioner, noting suicidal behaviours or ideation (Kleespies, 1993) as a consequence.Castelli Dransart et al. (2015) reported that 10% of the people in their study experienced their own suicidal ideation following the death by suicide of a patient, but no causal link was investigated.

| Change over time
The quantitative and mixed method studies included in this review varied in the timeframes of reference when collecting the retrospective data on the impact of a patient's death by suicide.Most asked participants to report on their reaction immediately following hearing of the death, whereas some also collected information about the impact at the time the study was conducted.The qualitative studies included in this review did not specify a time point and explored reactions more broadly across time.Findings from Sanders, Jacobson, and Ting (2005) and Bowers et al. (2006), for example, indicated that reactions persisted over time to the extent that those deaths that were some time in the past were still having an influence on practice in the present.Similarly, another study indicated that among a wide range of impacts, only disbelief showed a significant reduction with time (Saini et al., 2016).
Of the 13 quantitative studies that collected retrospective information on the period immediately following the incident and again at the time of the study, all indicated a reduction in the severity of impact over time, although Yousaf et al. (2002) reported that this reduction was not significant (Table 2).

| Duration of impact
Ten studies collected data on the duration of the initial impact following the death by suicide of a patient although the manner of recording and reporting varied.For example, Kleespies et al. (1990) found that the duration of the initial emotional impact was reported to have lasted a week or less for 37% (n = 3) of practitioners, between 1 and 4 weeks for 50% (n = 4) and between 1 and 4 months for 13% (n = 1) of the practitioners.However, longer-term effects related to professional practice were reported as still being present at the time of the study in 88% (n = 7) of practitioners.In Alexander et al. (2000) paper, 8% (n = 4) of psychiatrists stated that the effects had lasted up to a week, 31% (n = 15) up to 1 month, 31% (n = 15) up to 3 months, and 29% (n = 14) over 3 months.Gibbons et al. (2019) found that 39% (n = 45) reported a detrimental effect on professional duties, which lasted between 1 week and 6 months; 21% (n = 18) felt that the effects lasted between 6 months and 2 years; and 13% (n = 11) reported having ongoing effects.Murphy et al. (2019) indicated that although the emotional impacts predominately lasted less than 6 months, for nearly 10% (n = 8), the impact lasted more than 12 months.Finlayson and Graetz Simmonds (2016) reported that for nearly a quarter of their sample, 24% (n = 13), the feelings lasted less than 1 week.However, for 41% (n = 22), feelings lasted from between 1 week and 1 month, for 14.8% (n = 8) between 1 and 3 months, for 7% (n = 4) between 3 and 6 months, and 13% (n = 7) experienced the feelings for more than 6 months (see Table 3).

| Variables associated with the extent of the impact on practitioners
The studies included in this review examined a wide range of factors that may account for variability in impact on practitioners.The most frequently identified factors are summarized under five broad themes: practitioner characteristics; the therapeutic relationship; patient characteristics; the response from others, and fear of litigation and publicity.
Although the studies covered a wide range of professions, as the majority focussed on single professional groups, it was difficult to make direct comparisons.Castelli Dransart et al. ( 2014), Gulfi et al. (2010), Grad et al. (1997), andMcAdams andFoster (2000), however, reported that impact was unrelated to work role.
The practitioner's perception that they were in some general way accountable or responsible for the death, (Castelli Dransart et al., 2014;Dewar et al., 2000;Finlayson & Graetz Simmonds, 2016;Gulfi et al., 2010;Robertson et al., 2010) or that a decision they had made had contributed to the death (Hendin et al., 2004) emerged as a factor that was associated with a negative impact.This was also the case with the perception that the death could have been prevented (Bowers et al., 2006;Landers et al., 2010;Rothes et al., 2013) or was unexpected (Davidsen, 2011, Finlayson & Graetz Simmonds, 2016).Two studies reported that for some practitioners, reduced impact was associated with holding the belief that suicide was the client's choice and outside of the practitioner's control (Darden & Rutter, 2011;Draper et al., 2014;Rothes et al., 2013).Linke et al. (2002) found that the majority of their sample had felt inadequately prepared for dealing with a suicide by their initial professional training, although they all felt trained in risk assessment.However, in another study, previous training in suicide risk assessment was associated with practitioners reporting a reduced sense of burnout (Murphy et al., 2019). Castelli Dransart et al. (2015) categorized the group within their respondents who reported low impact as "anticipators with support," that is, people who had been aware of the risk of losing a patient to suicide and were also well supported following the actual loss.Cotton et al. (1983) reported that professionals whose previous experience had been more likely to expose them to the death of patients, for example, those with previous medical experience, suffered less than other disciplines such as mental health workers.

| Therapeutic relationship
Elements of the therapeutic relationship were cited in various studies as being related to differential reactions after a death.Litman (1965) noted in general terms that the impact was dependent on specifics of the relationship but without indicating what these specifics were or the direction of the association.The relationship to the patient (Castelli Dransart et al., 2015), the length of time the patient was known to staff (Bowers et al., 2006), the strength of relationship (Bowers et al., 2006), emotional closeness or attachment, (Castelli Dransart et al., 2014;Gulfi et al., 2010), and either knowing the patient well or disliking the patient (Dewar et al., 2000) were all reported as increasing the negative impact of the death.Likewise, difficulty with emotional contact with the patient (Davidsen, 2011), a long and intense involvement with the patient or their family (Murphy et al., 2019;Pieters et al., 2003), or a particularly difficult therapeutic relationship (Pieters et al., 2003) were all linked to an increase in negative impact.A small number of studies noted a link between characteristics of the person who died and the impact on the practitioner, finding that the impact was greater when the patients were younger (Dewar et al., 2000;Murphy et al., 2019;Pieters et al., 2003;Wurst et al., 2010) and when patients had young children (Dewar et al., 2000;Landers et al., 2010;Murphy et al., 2019).Chemtob et al. (1989), found that working with people with problems with substance abuse was associated with lower impact, and whilst recommending that this warranted further research, they questioned whether this may relate to practitioners being more prepared for the suicide with the knowledge that there was ready access to lethal means and to them believing that they had less influence over their patients.

| Response of others
The fear of blame by relatives or fear of their reactions (Dewar et al., 2000;Landers et al., 2010;Wurst et al., 2011Wurst et al., , 2013) ) or the fear of blame more broadly (Alexander et al., 2000;Bowers et al., 2006;Dewar et al., 2000;Hendin et al., 2000;Kleespies et al., 1990;Landers et al., 2010) was associated with higher levels of distress.Furthermore, Pieters et al. (2003) and Gibbons et al. (2019) found that actual confrontation with family members contributed to increased adverse impact.Castelli et al. (2004), however, described that overall, their sample experienced little blame (11% reporting blame from relatives), and they did not report any association with level of impact.Less specifically, managerial responses, (Bowers et al., 2006), negative reactions from health service executive staff (Landers et al., 2010), and negative reaction from the institution in which the practitioners worked (Gibbons et al., 2019;Hendin et al., 2004) were cited as factors that increased the negative impact.Five studies noted that publicity in the media contributed to the impact that practitioners experienced (Alexander et al., 2000;Dewar et al., 2000;Landers et al., 2010;Midence et al., 1996;Sherba et al., 2019), with all being in agreement that awareness of publicity heightened the distress experienced.

| Support that helps minimize negative sequelae
The vast majority of the studies that reported on the support accessed following the death of a patient cited informal support through colleagues, peers, family, or friends as having been the most helpful (e.g.Alexander et al., 2000;Cotton et al., 1983).In terms of more formal provision, supervision was also reported as offering valued support (Courtenay & Stephens, 2001;Kleespies et al., 1990;Ruskin, 2004;Trimble et al., 2000), but evidence for other formal procedures was more mixed.Team meetings were found helpful by some (Alexander et al., 2000;Courtenay & Stephens, 2001;Cotton et al., 1983;Pieters et al., 2003) but either unhelpful by others (Hendin et al., 2000), or the need for careful management of meetings was noted (Kelleher & Campbell, 2011;Linke et al., 2002).Some studies found that critical incident debriefs or case reviews were useful (Alexander et al., 2000;Kleespies, 1993;Kleespies et al., 1990;Landers et al., 2010;Pieters et al., 2003;Rothes et al., 2013;Sherba et al., 2019), but other studies described them as unhelpful (Bowers et al., 2006;Hendin et al., 2000), specifically if experienced as insensitive or persecutory (Gibbons et al., 2019).One study reported that the use of external reviewers to guide debriefing was unhelpful (Courtenay & Stephens, 2001).
More generally, Bowers et al. (2006) found that any support was received positively, and Castelli Dransart et al. ( 2014) found that receiving insufficient support significantly increased the overall impact in terms of emotional impact and trauma symptoms.

| DISCUSSION
This review had three main aims, to synthesize the evidence on (1) the impact of a patient's suicide on both personal and professional practice responses; (2) the support received by health professionals; and (3) which factors helped to minimize negative sequelae.Our key findings were that the death of a patient by suicide can have a considerable, and lasting, emotional impact (most often manifest as shock and/or guilt) on mental health professionals.This impact is comparable with that of other traumatic life events.There were notable impacts on their professional practice including self-doubt and being more cautious and defensive in the management of suicide risk.Informal support was highly valued, however more needs to be done to help to prepare and support practitioners for this rare but likely experience.

| Impact on personal and professional practice
There is extensive evidence in the literature that the loss of a patient through suicide has considerable impact on many practitioners.The reported incidence and severity of this impact varied considerably across the studies included in this systematic review.Those studies that reported on change in reactions over time all concluded that there was a reduction in impact over time.This would be expected given what is known about the processes of grief and of recovery following trauma (Kleber, 2019;Zisook & Shear, 2009).
Many studies used a PTSD outcome measure, the IES/IES-R, to assess the emotional impact of losing a patient to suicide.This seems reasonable given that such an experience would meet the event criteria of diagnostic guidelines for PTSD (American Psychiatric Association, 2013).A previous review of these studies (Séguin et al., 2014), concluded that, on the basis of the IES/IES-R scores, overall emotional impact on practitioners was low.However, in the current systematic review, for the studies utilizing the IES or IES-R, between 12% and 53% of their samples recorded clinically significant scores in the time following the suicide (see Table 4).To place these findings in context, although the majority (90%) of the general population will experience a traumatic event in their lifetimes, a relatively small proportion (8%) will develop PTSD (Kilpatrick et al., 2013).In light of this, it is fair to conclude that the emotional impact on practitioners of a loss of a patient to suicide is significant and comparable with the impact of other traumatic life events.Current guidelines for the prevention of PTSD (National Institute for Health and Care Excellence [NICE], 2018) recommend that individuals presenting with subthreshold symptoms (i.e., on a PTSD measure such as the IES-R) in the month following a traumatic event should be actively monitored, and those with above threshold symptoms should be offered trauma focussed treatment within 1 month the event.Psychological focussed debriefing is not recommended.
Although many of the studies included in this review reported the impact on their professional practice, often the studies did not differentiate between those changes that were beneficial (e.g., improved patient care and better risk management) or detrimental (e.g., defensive practice and avoidance) to patient safety and well-being.Some studies, however, did report positive changes as a consequence of the adversity of the experience, described in Cotton et al. (1983) as "new growth on emotional scars."Some studies concluded that to achieve such growth requires a facilitative working environment that recognizes the importance of learning through experience (Courtenay & Stephens, 2001;Gulfi et al., 2010;Kolodny et al., 1979).

| Support received by health professionals
All studies concluded that further action is needed to prepare and to support practitioners, and a number of studies recommended that efforts to prepare staff for the possibility of losing a patient to suicide could be beneficial (Bowers et al., 2006;Chemtob et al., 1989;Gibbons et al., 2019;Hendin et al., 2000;Jacobson et al., 2004;Sherba et al., 2019;Wang et al., 2016;Wurst et al., 2011).It is conceivable that suicide prevention training itself may not fulfil the function of preparing practitioners for the death of a patient unless specific information is included about the probability of experiencing the suicide of a patient at some stage in their career: the likely impact, the procedures to be followed, and the support available.Increasing awareness of the incidence of suicide could improve risk management whilst also serving a protective function for the practitioner in the event of a death.
The majority of studies indicated that informal supports were most helpful, although it is not clear if this is partly due to an inadequacy of more formal supports or a preference for a more individualized approach.When formal support was reported, individual supervision appeared to be the most valued support whilst the responses to group procedures such as team meetings and debriefs were more mixed.However, generally not enough detail was provided to make meaningful comparisons between different group supports.
Although the current evidence points most clearly towards a preference for individual support, either informally or through supervision, this should be treated with caution.Group procedures have the benefit of facilitating the sharing of learning and of heightening awareness of risk, therefore helping practitioners to be better prepared in the event of losing a patient through suicide.If the emphasis is on individual support, then other ways of achieving these functions would need to be found.

| Factors that helped to minimize negative sequelae
There was no strong evidence that the demographic characteristics of practitioners (e.g., gender, age, profession, and length of experience) were associated with the impact of a suicide.It is possible that further research could be warranted so that certain groups of practitioners could be targeted by interventions to help reduce impact; however, it is more likely that such measures would best be directed to the whole workforce.The findings from this review support Castelli Dransart et al.'s (2014) conclusion that we need more research focussed on the relationship with the patient, rather than limiting our attention to the socio-demographic characteristics of the professionals and their patients.This association between the therapeutic relationship and impact is further supported by Cerel et al.'s (2016Cerel et al.'s ( , 2017) ) research indicating that perceived closeness to the deceased increased the chance of depression, anxiety, and PTSD.Consequently, one's perceived closeness to the deceased is likely to be an important factor in identifying those at greatest risk of adverse impact.
The finding that more negative impacts were associated with practitioners' perceptions that they were accountable or responsible (in some way) for the patient's death could be linked to the common personal reactions of guilt and fear of blame and the experience of reduced professional confidence and uncertainty.It would be beneficial to gain a better understanding of this relationship.For instance, is guilt exacerbated by reduced confidence in one's ability to assess and manage risk, or inversely, is greater confidence in one's ability protective against ruminations such as "did I miss something?" or "I should have known?"Further research would also be beneficial to investigate the influence of blame (self-blame, fear of blame, and actual experience of blame) and to investigate if blame is moderated by, for example, cultural factors, institutional factors, the frequency of suicide within an institution, attitudes towards suicide (e.g., if there is a more prevalent cultural belief that suicide is a valid option), and confidence in one's ability to assess and manage risk.
Castelli Dransart et al. (2014) suggested that a number of factors might have accounted for what they described as the relatively low impact of the death of a patient in their study.These included whether (i) the majority of their sample had experienced several patient suicides (with potential habituation effects), (ii) practitioners worked in team settings, (iii) they reported receiving sufficient support, and (iv) they worked in a socio-cultural setting in which they felt little sense of blame or fear of litigation.

| Professional implications
Feeling unprepared for the loss, also potentially related to the sense of shock, involved several different aspects: first, a general lack of awareness of the likelihood that the practitioner's profession may expose them to such an event; second and more specifically that the death by suicide of an individual patient in their care was unexpected, and finally uncertainty of the implications of the loss in terms of personal impact and service procedural requirements.Whilst awareness would be best raised during initial training, given the relative rarity of the event, sharing of learning following a death could help to maintain the level of awareness throughout a practitioner's career.
The evidence regarding the acceptability and benefits of formal support was mixed; however, several recurring themes were identified that could provide useful targets for additional focus when supporting health professionals who have lost a patient to suicide.Moderating the impact of guilt, a ubiquitous response across the studies included within this review, and improving understanding of how services can create an environment that reduces the fear of blame and promotes an atmosphere of learning from adverse events are both areas that could be improved.
Given the relative homogeneity of the responses reported in the studies, it is likely that the synthesis of these findings would generalize to a broader range of mental health practitioners than those already covered.However, further research is required to confirm this.

| Strengths and limitations
The findings of the current review must be interpreted within the context of its limitations.Only English language publications were included, and unpublished papers or research from the "grey literature" were excluded.Inclusion and exclusion criteria were deliberately set to cover a broad range of studies, but comparison of findings was challenging given the varied methodologies employed.The reference lists of all included studies were checked, but no additional studies meeting the criteria for this review were found.All included studies were retrospective cohort studies, based on self-report measures.
Methodologies of the studies varied: Some required participants to report on their reaction to their first, last, or most distressing experience, whereas for other studies, this was not specified at all.
The time since the participants experienced the death of a patient also varied greatly.Furthermore, studies ranged in the time point at which participants were asked to report on their reactions, that is, whether to retrospectively describe reactions immediately following the event and/or at the time of the study.Understandably, reports of impact were likely to be subject to recall bias (Spencer, Brassey, & Mahtani, 2017).
Like most research of this kind, all studies had possible selfselection bias, because participation in the studies was voluntary.
Practitioners with more or less severe stress reactions might have chosen not to participate in such studies because they anticipated taking part to be distressing, or they may have thought that they had little to contribute.In terms of sampling frames, none of the studies reported information on those practitioners who had experienced a patient suicide but who did not take part in the study.As a result, it is not possible to comment on the representativeness of participants included within this systematic review.Nonetheless, the current review has a many strengths.The review was systematic and included both quantitative and qualitative studies.We also used a validated quality assessment tool (Pace et al., 2010), which allowed the quality of both the qualitative and quantitative studies to be appraised.Additionally, the protocol for our systematic review was pre-registered on Prospero to facilitate transparency.A further strength of our review is that whereas other reviews, for example, Foley and Kelly (2007) and Leaune et al. (2019), focused primarily on psychiatrists or psychologists, we included studies with a range of mental health professionals, broadening the potential utility of our findings.

| CONCLUSION
The experience of losing a patient through suicide can have considerable impact on mental health professionals, both in terms of their personal reactions, which are typically characterized by feelings of guilt and shock and subsequent changes to professional practice that may be potentially positive or negative.Demographic characteristics of practitioners did not appear to be associated with impact.The most significant risk factors for negative impact were a sense of fear or blame and feelings of self-blame.In a small but important proportion of practitioners, the personal impact met the Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnostic criteria for PTSD and could be long lasting.This impact, however, may be moderated by tailored training (awareness of the occupational risk of loss through suicide), characteristics of the employing organization (i.e., a non-blame seeking approach, a culture of learning through adverse events) and by the level and nature of support available (tailored to the individual, opportunities for informal support).

FUNDING INFORMATION
DMS is completing a PhD funded by Cumbria Partnership NHS Foundation Trust.OJK is supported by a fellowship from an FWO Odysseus grant (Myin-Germeys, FWO GOF8416N).
Abbreviations: DSP, death by suicide of a patient; IES/IES-R, Impact of Events Scale-Revised.

•
and friends • Avoidance of suicide scene Quantitative studies n = 21 DSP = death by suicide of a patient Study, country, quality assessment (Psychiatric hospitals and outpatient psychiatric services, social and medical services, residential homes for people with mental health or addiction problems, care homes for elderly, and prisons IES-R (Impact of Event Scale-Revised) 5 pt scale, AEIS (Acute Emotional Impact Scale) 7 pt scale from Kleespies, 1993 • Low emotional response on the AEIS (M = 2.08, SD = 0.59).15.5% (n = 40) of the respondents were above the cut off.

•
studies n = 21 DSP = death by suicide of a patient Study, country, quality assessment (Psychiatric hospitals, outpatient psychiatric services, social and medical services, residential homes for persons with mental health or addiction disorders, homes for the elderly and prisons • Private practice IES-R Impact categorized into groups based on mean IES-R total score: (M = 20.0) and avoidance (M = 13.4) scores nearly equivalent to those of a patient sample with bereavement.
T A B L E 1 Search strategyEligibility:Databases searched from inception to 2019, search not limited by study design, studies restricted to English language