Physician Suicide: A Scoping Literature Review to Highlight Opportunities for Prevention

Suicide is estimated to occur at a higher rate among physicians than the general population and perhaps even other professions (Schernhammer and Colditz 2004; Gold, Sen, and Schwenk 2013); however, estimates by profession vary (Peterson et al. 2018). Death by suicide was the only primary topic in 108 (31.1%) of all articles. Non-opinion articles primarily sought to estimate epidemiology of suicidal behaviors among physician populations. In other words, death by suicide was most often studied among attending physicians because of availability of data on occupation and death, particularly in vital statistics or death certificates. However, the tools used to perform such estimates varied. Deaths by suicide were most frequently estimated based on 9th or 10th Revision of the International Classification of Diseases (ICD-9 or ICD-10) codes on death certificates.

Other commonly used sources of data included membership masterfiles from physician organizations or associations (Juel, Mosbech, and Hansen 1999), published obituaries, and charts from medical records or forensic reports (Table 2). These data are primarily limited by potential undercoding. Undercoding can result from deaths being coded as accidental deaths, for example, leading to underestimation of the actual incidence of suicides among physicians. In an editorial, one author notes, “Suicide is a way to die and not a cause of death. And there are several means to this end: the ICD-10 lists at least 31 different ways to perform a suicide” (Aasland 2013). Rimpelä et al. articulated this in 1987 also: “Differentiation between suicide, accident, poisoning, and violence as a cause of death is often difficult, and suicide might sometimes be falsely, even deliberately, classified as an accident and even differently in different occupational groups” (Rimpelä et al. 1987).

Suicides are typically reported as a suicide mortality rate (SMR) in epidemiologic literature, which is the number of deaths by suicide per 100,000 person-years. A suicide rate ratio can be calculated by dividing the SMR among physicians, or a subpopulation thereof, and dividing by a comparison group, such as the general population. Despite these standardized manners of suicide reporting, included papers showed inconsistencies in reporting, often reporting a crude mortality rate, which is calculated by the absolute number of physician suicides divided by the number of years in a study period, then reported per 100,000 physicians. For example, in 1968, Craig and Pitts counted 228 suicides among physicians based on the obituary materials collected by the Deaths Editor of the Journal of the American Medical Association between May 1965 and May 1967 (Craig and Pitts 1968). Combined with an estimation of approximately 296,000 physicians in the U.S. at the time, this resulted in a crude annual suicide rate of 38.4 per 100,000 physicians (Craig and Pitts 1968). Multiplying this crude annual suicide rate and the U.S. physician population of approximately 953,000, which was reported by a 2016 census from the Federation of State Medical Boards, leads to an estimate of 366 physician deaths by suicide annually.

Underreporting undermines the accuracy of physician suicide incidence estimates using death certificates, membership files, and medical charts. Surveys also can lead to underreporting. One study surveyed the deans of 116 medical schools in the 1970s about medical student suicides; 88 respondents reported 52 medical student suicides between 1970 and 1978 (Pepitone-Arreola-Rockwell, Rockwell, and Core 1981). Another approach to studying physician suicides is psychological autopsy, in which a psychological profile about a person is constructed after their death. Time-intensive and not a standard practice, psychological autopsy involves collecting information through interviews of relatives and healthcare professionals, along with data from forensic examinations, police investigations, psychiatry, medical and social agency records, and any other information available, such as suicide notes (Lindeman et al. 1998). Such an approach can elucidate potential contributors to an individual’s death, providing detailed contextual information, their life circumstances, and how they managed such circumstances up until their death. However, psychological autopsies are not systematically applied in the setting of deaths by suicide. Case reports also may describe a psychological profile, in addition to medical and clinical aspects of the case (Lindeman et al. 1998; Hawton, Malmberg, and Simkin 2004; Koekkoek et al. 2008), but this was also infrequently published. In 2013, Gold et al. examined data from the U.S. National Violent Death Reporting System (NVDRS), which is a combination of information and data from death certificates, coroner and medical examiner reports, toxicology reports, and law enforcement reports (Gold, Sen, and Schwenk 2013), the only study to triangulate multiple data sources to estimate suicide incidence. The wide variety of data sources and methods of estimation suggest that there is no standard.

Suicide clusters are suicides that occur near each other, typically with respect to time and geography. Suicide contagion refers to spread of information about a suicide via media and other channels, which can increase suicide risk in a community. Only three studies acknowledged these topics. Two opinion papers described suicide clusters that occurred in a short time span and in a focused geographic region, one in Winnipeg, Canada (Williams 1997) and another cited suicides in Australia that had been reported in news media (Clarke and McKee 2017). Both articles focused on intense or distressing working conditions as important factors in the suicide deaths, but did not elaborate further on suicide clusters. In fact, the Canadian article, along with one other case series of physicians who died by suicide while on probation in Oregon state (Crawshaw et al. 1980), used the term “epidemic” rather than “clusters.” Nevertheless, these three articles referred to the concepts of suicide clusters and suggest contagion, even though they did not use these terms explicitly. These were articles that were tagged with the public health theme.

Suicide clusters also appeared to follow physician subpopulations, although numbers were small in such case series and opinion articles. Other physician subpopulations included, for example, immigrant physicians (Richings, Khara, and McDowell 1986), physician pilots (Cath 1974), and physicians who experienced war either as victims or as wartime medics (Bock et al. 2016; Likeman 2015). In 2016, Bock et al. described a case series of 62 Hungarian Jewish dermatologists during the Holocaust: some emigrated, some died during Nazi rule, and others survived (Bock et al. 2016). Three of these dermatologists died by suicide, which was “very common among the doctors (under the Nazi ordeal), especially those physicians of the older generation” (Bock et al. 2016). Another analysis described issues relevant to physician terrorists (Marvasti and Vahidy 2008), who are physicians who die by suicide and also experience “feelings of hopelessness, frustration, anger, and helplessness, as well as other kinds of psychic pain.” Nevertheless, the authors contrast “clinical suicide” and “suicide bombing” by exploring their differing motivations (death of self vs. death and terrorization of others via death of self) and the relative lack of psychopathology and greater reliance on public perceptions of their actions among suicide bombers. This was not a population further described in this literature review.

Overall, suicidal ideation was the second most studied thesis among articles included in this literature review, especially suicidal ideation among medical students. To assess suicidal ideation and attempts, surveys were most common; overall, 61 cross-sectional survey studies were performed (Table 1). Validated surveys are available to assess suicidal ideation (Table 2); however, investigator-developed items and surveys were also used. Only four studies compared physician subpopulations between countries. In the Health and Organization among University Hospital Physicians in Europe (HOUPE) study, investigators sought to compare suicidal thoughts and work-related stressors between practicing physicians in Italy and Sweden (Fridner et al. 2009, 2011, 2012). Another study compared lifetime prevalence of suicidal thoughts among practicing physicians in Norway and Germany (Rosta and Aasland 2013).

Suicide attempts, in general, were poorly studied in any physician subpopulation, but suicidal behaviors among postgraduate trainees were the least studied of all, with resident physicians’ behaviors studied more than fellows (Figure 2). Among the articles reviewed, only two validated tools, Paykel’s Instrument and the Beck Hopelessness Scale, were identified that were specifically designed to assess suicide attempts; otherwise, investigators developed survey items to assess respondents’ self-reported attempts.

Figure 2

Physicians’ risk factors for suicidal behaviors, especially mental health disorders, are among the most common topics of study. Gold et al. examined the NVDRS data and identified certain risks unique to physicians compared to non-physicians: physicians were more likely to have a job problem preceding death by suicide and a greater likelihood of known mental health disorders, yet no accompanying increased likelihood of antidepressant therapy (Gold, Sen, and Schwenk 2013). Additional risk factors potentially include professional setbacks prior to suicide death, such as facing complaints (Crawshaw et al. 1980; Casey and Choong 2016), feeling overloaded, working long hours, being unable to cope with job responsibilities (Keith Hawton, Malmberg, and Simkin 2004), or experiencing disability as a result of medical illness (Cath 1974; Lindeman et al. 1998; Thomas 2014). Related to these conditions, physicians are also at high risk of burnout, which has been found to be associated with suicidal ideation in U.S. medical students (Dyrbye et al. 2008) and Dutch residents (Van Der Heijden et al. 2008), although no direct relationship between burnout and death by suicide has been established. To assess these risk factors, a variety of validated questionnaires were used to inquire about general health, burnout and stress, mental health (e.g. depression, anxiety or insomnia), substance use including alcohol use, and other measures of career, work, personality, and other life experiences (Table 2). Additionally, investigator-developed survey questions were also used, for example, to assess attitudes toward suicidal behaviors of a peer.

Because the earliest publications about physician suicide suggested an increased incidence compared to the general population, certain theories have been applied to attempt to explain the increased suicide risk among physicians. For example, Fink-Miller applied the interpersonal psychological theory of suicidal behavior (IPTS) (Cornette et al. 2009; Fink-Miller 2015a) to physician suicide. The IPTS posits three necessary and sufficient precursors to death by suicide: (1) thwarted belongingness, a feeling of disconnection with others; (2) perceived burdensomeness, a miscalculation that one’s death would relieve burdens on others; and (3) acquired capability, habituation to previously provoked fear responses, including losing the fear of pain involved in taking one’s life. This can stem from repeated exposure to painful or provocative stimuli, including events triggering second victimization, such as patients’ poor outcomes, death, and suffering. This can then lead to desensitization when exposed to death in general.

Role strain describes physician risks as a result of their direct work environment or professional norms; a mismatch between social and institutional norms and the physician’s roles can manifest as an unrealistic expectation of perfect function at a maximum level of competence (DeSole, Singer, and Aronson 1969). Professional self-image and identity, along with self-stigma, are also considered relevant mediating themes (Cath 1974; Lindeman et al. 1998; Legha 2012).

Early studies of suicide rates by specialty suggested that psychiatry and anesthesiology had the highest suicide rates (Cornette et al. 2009). A 2014 case series of 87 physician suicides in Japan demonstrated that internists (24.1%), psychiatrists (18.4%), pediatricians (6.9%), and anesthesiologists (5.7%) were most commonly affected specialties, although 16% of cases had unknown specialty (Hikiji and Fukunaga 2014). Small case series in Thailand and China had to few cases to draw a conclusion about specialties (Wang, Liu, and Xu 2017; Yasri and Wiwanitkit 2017), a commonly encountered issue in comparing specialties’ suicide rates. Eighteen non-opinion articles published since 2010 described either cross-sectional survey studies of burnout and its correlation with suicidal ideation per specialty or additional case reports or series of physician suicide deaths. Generally, the findings about which specialties may be at higher risk than others have mixed findings due to data limitations in ascertaining suicide rates, small samples, and inappropriate comparisons or assumptions in data interpolation (Bruce et al. 1968; Bergman 1979; Neil et al. 1987; Swanson, Roberts, and Chapman 2003).

Means of suicide death was also a topic examined in some studies. In a U.K. study examining suicide means, 10 (50%) of 20 anesthesiologists died by overdosing with anesthetic agents compared to other doctors, for example, 2 (16.6%) of 12 surgeons and 10 (4.5%) of 223 physicians overall died by overdose (Hawton et al. 2000). In the 2014 case series published from Japan, authors found a similar trend in studying suicide deaths across specialties (Hikiji and Fukunaga 2014); four of five anesthesiologists (of a total of 87 physicians) used drugs as means of suicide. Based on the IPTS, one proffered but unstudied hypothesis is that some medical specialties may possess more acquired capability, or acquired fearlessness about death, than others (Fink-Miller 2015a, 2015b).

The access and knowledge hypothesis is a commonly discussed risk factor for suicide death among physicians. Physicians acquire specialized medical knowledge of the human body and have access to the means (e.g. prescription drugs) that can cause lethal self-harm. Observational studies of means of suicide death suggest that firearm, prescription drug overdose, and hanging are among the most common methods used by physicians (Cornette et al. 2009; Rimpelä et al. 1987; Swanson, Roberts, and Chapman 2003), although this varies by country. Psychological autopsies are consistent with the access and knowledge hypothesis. One set of psychological autopsies in the United Kingdom examined 38 physician suicides, finding that 28 of the physicians died by self-poisoning and 11 by self-injury (other than poisoning), as 1 physician used both methods (Keith Hawton, Malmberg, and Simkin 2004). In a Finnish psychological autopsy of seven physician suicides, only two had previous suicide attempts, suggesting that physicians may be more likely to cause self-harm leading to death by suicide; in other words, physicians who die by suicide may not have had any prior suicide attempt, which is a risk factor in the general population (Lindeman et al. 1998). Self-treatment or self-prescription is also a concerning contributor due to specialized access to controlled substances, as “medical doctors exploit their profession for purpose of self-treatment” (Lindeman et al. 1998).

Personality and life experience prior to medical training might also contribute to physician suicide risk, although they are poorly studied. In a prospective study of Norwegian medical students, a version of the Basic Character Inventory, originally developed in 1920, was used to measure major personality dimensions, including neuroticism, control, and extroversion (Tyssen et al. 2001, 2004). The control personality trait, or the degree of compulsiveness, was independently linked to suicidal ideation; and more neuroticism as a medical student, which is linked to the vulnerability personality trait including sensitivity to other people’s opinions and criticism, independently predicted more serious suicidal ideations and planning in the first two postgraduate years (Graves and Thomas 1991; Tyssen et al. 2004, 2001).The reality weakness personality trait, more commonly associated with serious psychiatric pathology such as paranoid traits, predicted a transition from suicide ideation to planning (Tyssen et al. 2004).

One editorial speculates about the influence of adverse childhood experiences on risk for physician suicide (Johnson 1991). A South Korean study found a more than threefold risk of lifetime suicidal ideation, planning, and attempts among medical students who experienced emotional abuse early in life, characterized by “a continuously cold and uncaring parental attitude” (Jeon et al. 2009). In China, medical students were surveyed about parental relationships and parenting communication styles, hypothesizing that this could be a highly influential aspect of the student’s character (Sun et al. 2017). The study found that for Chinese medical students, a good relationship with parents was statistically significantly associated with less suicidal ideation, plans, and attempts.

Some studies sought to identify risk factors unique to their cultural context. Researchers in the United Arab Emirates collaborated with Eskin (Eskin et al. 2011) to again assess medical students’ religiosity and their attitudes toward suicide (Amiri et al. 2013). In this study, investigators suggested that the Islamic faith may serve as a protective factor against suicidal thoughts and death by suicide among their medical students. In Pakistan, a study of suicidal ideation among medical students included religion as a demographic characteristic, but did not otherwise include religion in further analyses (Osama et al. 2014). Among opinion articles, certain conceptualizations of suicide also varied. For example, in one commentary from Japan, two concepts directly link overwork with death: karoshi (death due to overwork) and karojisatsu (suicide due to overwork) are considered causes of death in the Japanese culture (Hiyama and Yoshihara 2008), but nowhere else in the world.

Early studies in the 1970s and earlier explicitly excluded female and minority physicians (Thomas and Greenstreet 1973; Ullmann et al. 1991; Graves and Thomas 1991). Results published in 1999 from the Women Physicians’ Health Study, which surveyed a sample of women physicians from the American Medical Association masterfile (Frank and Dingle 1999), described previous studies that reported an odds ratio as high as 4 for women physician suicides compared to other categories of women, but that such studies were based on small numbers of suicide deaths. This study found that women physicians had similar rates of depression (19.5%) compared to women in the general population, but 1.5% of women physicians attempted suicide compared to 2.9% of women in the general population (Frank and Dingle 1999); the authors postulated that women physicians may have less suicidal intent, and therefore less suicide attempts, or it could be that they have a higher rate of completion than women in the general population, and therefore also less suicide attempts since attempts more often resulted in death.

In 2004, Schernhammer performed a systematic review and meta-analysis, concluding that the suicide rate ratio for women physicians was 2.27 compared to the general women population and for male physicians was 1.41 compared to the general male population (Schernhammer and Colditz 2004). In 2013, a Polish study of perimenopausal female physicians examined sociodemographic and lifestyle variables’ relationship to participants’ subjective sense of health, including suicidal ideation. Eight of 221 participants reported suicidal thoughts, which was statistically significantly correlated with poorer subjective sense of health (Gojdź et al. 2013). Another study in Pakistan appeared to show no gender difference in terms of rates of suicidal ideation among medical students (Ghazanfar et al. 2015).

Little data is known on suicide among physicians of gender minorities, as there is only one perspective piece written by a nurse who wrote about her transgender child who died by suicide (Dippo 2007).