Introduction

Suicide is a major cause of death among individuals with eating disorders; however, the vast majority of studies published on the topic until now have been on clinical samples [1–16]. The current study aims to report on the risk of suicide among females with eating disorders based on a population-based sample of female adolescents in Israel, utilizing data on diagnoses routinely collected in a standardized method by the Israeli military and diagnoses given in mental health examinations among adolescents who sought or were referred to mental health treatment. Later, suicide was ascertained using the Israeli Death Registry, which includes records of all deaths in the country. We hypothesized that females with eating disorders would be at increased risk for suicide compared to a population-based cohort of females without eating disorders.

Materials and methods

Psychiatric screening

Service in the Israeli military is mandatory for males and females and all Jewish Israeli citizens are screened by the military at age 17. At age 17, all females are screened for psychopathology as part of the assessment of their ability to serve in the military. They are examined by a physician, who has a letter from their pediatrician reporting all significant and/or chronic illnesses. Based on this letter and the physician’s examination, those suspected of suffering from behavioral disturbances or mental illness are referred for in-depth assessment by a mental health professional. The mental health assessment is administered using a semi-structured interview performed by a clinical social worker or psychologist, who inquires about personal and family history, previous psychological and psychiatric treatments, interpersonal relationships, self-injuries and antisocial acts, and functioning within the family and school. Based on this assessment, if the clinician suspects that the adolescent suffers from psychopathology, a provisional diagnosis is suggested (based on ICD-9 and ICD-10 criteria), and the adolescent is referred for further evaluation. The data in this study include all adolescents who underwent screening at the draft board, even those who were subsequently deferred from service based on this screening.

In addition to screening, we also obtained data on females diagnosed with eating disorders during their mandatory 2-year service. Eating disorders are recorded according to severity which was coded as mild, moderate or severe according to the criteria specified in Table 1. This study focused on last diagnoses of eating disorders of any severity. Furthermore, we looked at change in the severity of the eating disorder between the draft board assessment (screening) and the last diagnosis given during military service.

Table 1 Criteria defining the severity of eating disorders
Full size table

Comorbid psychopathology was defined as having additional diagnoses of psychiatric disorders given by the clinician, including personality disorders, neurosis, non-affective psychosis, organic disorders, antisocial personality disorders, minor affective and anxiety disorders, posttraumatic stress disorder, adjustment disorder, major affective disorders (which include affective disorder with or without psychotic features), mental retardation, autism, substance abuse and body perception disorders.

The Central Bureau of Statistics

For all deaths in the country, the treating physician must fill out a form describing the cause of death. The Central Bureau of Statistics receives these reports and encodes the causes of death, including suicide, using standard ICD criteria. As in other similar studies [17], death by suicide was defined to comprise both those deaths which were officially classified as suicide/intentional self-harm (ICD9 E950-959; ICD10 X60-84) and also deaths ‘undetermined whether accidentally or purposely inflicted’ (ICD9 E980-989; ICD10 Y10-34). This was done to reduce misclassification bias. Furthermore, as has been done in other studies (e.g., [18]), in a further attempt to reduce misclassification bias (in which death by suicide is reported as other causes of death), hence under-reporting suicide, we excluded female adolescents who later died for reasons other than suicide from the analysis. Hence, the analysis compared risk for suicide in females with eating disorders with risk for suicide in females without eating disorders. The rate of death by suicide reported in this study for the population-based cohort of female Israelis was 0.3 %, which is similar to the rates reported by the Israeli Ministry of Health for the relevant years among Jewish and non-Jewish females in Israel [19].

Study population

After receiving approval from the IDF Medical Corps IRB, the draft board data were linked to the death registry, and the Ministry of Health provided mortality data including the causes and dates of deaths. Utilizing data from the draft board and from cases reported during mandatory service, we identified 1,356 females (0.2 % of the population-based cohort of female adolescents consecutively screened) who had a last military diagnosis of an eating disorder of any severity (mild cases: n = 358, moderate-severe: n = 998). We compared their rates of suicide to those of the entire population-based cohort of 577,399 female Israeli adolescents, who were consecutively screened by the Israeli draft board during the same time period. Mean follow-up time from first contact with the draft board until date of death or date of merger was 8.5 years (SD = 5.34; range 0–28 years).

Statistical analyses

The crude mortality ratio (N suicides ÷ N sample × 100) of suicides is presented for those with eating disorders and for the control group of females without eating disorders who served in the military during the same years. Rate ratio (observed suicide rate per 100 person-years ÷ expected suicide rate per 100 person-years) was then calculated to examine the risk of suicide among those with eating disorders compared to the expected rate according to the population-based control group. Females with eating disorders were divided according to the severity of the eating disorder (mild vs moderate–severe) and crude mortality ratios and rate ratios were calculated for each group and compared to the population-based control group. Additionally, we examined the number of completed suicides according to change in eating disorder severity between the draft board assessment (screening) and the last diagnosis given in the military. Finally, an additional analysis compared females with eating disorders (divided according to severity) with and without comorbid psychopathology in terms of their risk for suicide.

Results

An examination of crude mortality ratios (%) for suicide indicated that females diagnosed with an eating disorder of all severities had higher rates (0.22 %, n = 3) and risk of completed suicide (RR = 10.00, 95 % CI = 3.16–31.19) compared to the population-based control group (females without eating disorders; 0.03 %, n = 166). All three cases of suicide among individuals with eating disorders were defined as self-harm (rather than ‘undetermined whether accidentally or purposely inflicted’), indicating that they were definite cases of suicide.

An examination of the crude mortality ratios (%) according to the severity of the eating disorder (mild, moderate-severe) indicated that none of those with mild eating disorders had committed suicide, in contrast to 0.30 % (n = 3) of those with moderate–severe eating disorders. Furthermore, significantly increased risk of suicide (in comparison to the population-based control group) was only found among those with moderate–severe eating disorders (RR = 12.50, 95 % CI = 3.86–38.09; Table 2).

Table 2 Summary of crude mortality ratios, suicide rates, and rate ratios for participants with an eating disorder in comparison with the population-based control group
Full size table

An examination of change in severity of the eating disorder between the draft board assessment and the last diagnosis given in the military revealed that of the 1,356 female adolescents with a last diagnosis of an eating disorder, 64.6 % did not have a diagnosis of an eating disorder at screening, 33.4 % did not change in terms of severity, 1.4 % worsened, and 0.6 % improved. Two of the cases of suicide were found among those who did not change in terms of severity and one case was found in a female adolescent who was not reported as having an eating disorder at screening, but had a moderate–severe eating disorder diagnosis according to the last diagnosis received in the military. No cases of completed suicide were found among those whose eating disorder improved over time.

An examination of comorbid psychiatric diagnoses among female adolescents with eating disorders (the percentage of eating disorder patients with the different comorbid psychiatric disorders can be found in Table 3) indicated that there were no significant differences in risk for suicide between those with moderate–severe eating disorders who had comorbid psychopathology (RR = 13.83, 95 % CI = 1.77–108.85) and those with moderate–severe eating disorders who did not have comorbid psychopathology (RR = 11.00, 95 % CI = 2.75–24.29).

Table 3 Percentage of eating disorder patients (n = 1,356) with comorbid psychiatric disorders
Full size table

Discussion

This study found that females with an eating disorder were approximately ten times more likely to commit suicide in comparison with the population-based control group of females who were not found to have eating disorders when screened at the draft board or during their military service; this increased risk of suicide was only present in moderate–severe cases of eating disorders.

Past studies have varied in the percentage of suicides reported among individuals with eating disorders, with crude mortality ratios (N of suicides/N of sample × 100) of suicide ranging between 0.7 and 4.5 % among individuals with various eating disorders (e.g., [1, 2–6, 9, 12, 16, 20]). This present study found that 0.2 % of the females in this sample who had an eating disorder diagnosed at screening or during their military service had committed suicide, a rate lower than most previous reports, perhaps due to the limited follow-up period (mean follow-up time: 8.5 years). A possible explanation for the lower crude mortality ratio reported in this study among females with eating disorders is the population-based nature of the study. In contrast with other studies which focused on persons with eating disorders receiving treatment, this current study includes clinically significant cases which might not necessarily seek medical attention in a community setting. This is particularly important in light of the data indicating that in community settings, relatively few eating disorder cases come to medical attention. For example, a survey performed in the USA reported that fewer than 20 % of the adolescents with eating disorders received treatment for their disorder [21]. Screening for case detection is very intensive in the military, since there is a low tolerance for odd behaviors differing from the norm and, thus, a high threshold for referral to a mental health examination. Thus, one might assume that the cases identified in this study may have been milder overall than those that come to medical attention and receive treatment, which have been the focus of previous studies on eating disorders and suicide. However, because of the secretive nature of eating disorders, a high threshold of referral for mental health examination may not be sufficient to identify eating disorder cases. Thus, the underdiagnosis of eating disorders remains a possibility.

An advantage of the population-based nature of this study was that it enabled a comparison of the risk of suicide among females with eating disorders of different severities (despite the small number of participants with mild eating disorders). Findings indicated that having a mild eating disorder did not significantly increase the risk of suicide, whereas having a moderate–severe eating disorder increased the risk of suicide by 13-fold. This magnitude of risk for suicide is similar to that reported in previous studies on risk of suicide among individuals with different eating disorders (ranging between 4.7 and 58.1; e.g., [1, 2, 6, 9, 12–16, 20]). This finding has important clinical implications for parents and families of adolescents with eating disorders, and the clinicians who treat them, in that risk for suicide seems to be limited to the more severe cases.

Furthermore, an additional analysis indicated that both females with moderate–severe eating disorders with comorbid psychopathology and those with moderate–severe eating disorders without comorbid psychopathology were at increased risk of completed suicide and no significant difference was found between the two groups, indicating that having a comorbid psychiatric disorder in addition to an eating disorder does not seem to increase risk of suicide beyond the increased risk incurred by having an eating disorder.

Limitations

It is important to note that the findings of this study are based on only three deaths by suicide among individuals with eating disorders; hence, the confidence intervals are very wide and limit the power of the analysis and the ability to draw generalizable conclusions from the results. In addition, the exclusion of individuals who died by other causes of death to reduce misclassification bias may have led to an underidentification of suicide cases. Thus, the results should be interpreted cautiously.

While there are no published data on the sensitivity and specificity of the vital statistics reported in the Israeli Death Registry, the process of ascertaining death in Israel is systematic. When death occurs in a hospital and there is an ECG with a straight line, the doctor fill outs a death certificate which is sent to the Ministry of Health. In cases of death which occur outside the hospital, a doctor must observe the body, confirm death, and fill out a death form. In addition, in all cases of death which occur outside of the hospital, the police are notified and must examine the body and the circumstances of death to ascertain whether the death was unnatural. This rigorous procedure leads us to be confident that the vital statistics recorded in the Israeli Death Registry are valid.

Furthermore, the point prevalence of eating disorders reported in this study (0.2 %) is in fact relatively low compared to other studies. However it is within the lower part of the range reported in previous studies [22]. One explanation is that since the Israeli military is not focused on identifying eating disorders per se, but rather on screening for psychopathology that significantly compromises functioning in the military, milder cases of eating disorders might not have been identified or may have been ignored. Unfortunately, there is no existing information on the sensitivity and specificity of the screening mechanisms used in this study. Second, the Israeli military does not differentiate between eating disorder subtypes (i.e., anorexia nervosa vs bulimia nervosa). However, an examination of underweight BMI as a proxy for anorexia nervosa did not indicate increased risk for this group (data not shown), in contradiction with other studies (e.g., [16]). Thirdly, this study used the last diagnosis of an eating disorder in the analyses, which does not account for diagnostic variation over time. However, an examination of change in eating disorder severity between the screening and the last diagnosis given in the military indicated that there were no cases of suicide among those whose eating disorder severity improved over time. Rather, two of the three cases of suicide were among female adolescents with stable eating disorder severity and one case of suicide was that of a female adolescent who was not reported as having an eating disorder at screening, but had a moderate–severe eating disorder diagnosis according to the last diagnosis received in the military. Fourth, this study had a limited follow-up period of 8.5 years on average, and it is certainly plausible that additional cases of suicide occurred after follow-up. Finally, this study focused on the association between eating disorders and risk of suicide among females and not males, since there was a small number of males who were diagnosed with eating disorders (n = 212), none of whom had committed suicide (data not shown).

Despite these limitations, this study is the first to examine the risk of suicide in a non-clinical population-based sample of females with eating disorders, providing an assessment of the risk of suicide among females with eating disorders from a different angle than examined previously. This methodology enabled an examination of the risk of suicide among females with mild and moderate, rather than only severe, eating disorders. However, the majority of those with severe eating disorders are probably known to the mental health system. Findings indicated increased risk of suicide among females with eating disorders, though this risk was limited to moderate and severe cases of eating disorders. In conclusion, females with moderate–severe eating disorders have greatly increased risk of suicide and should be monitored for suicidal intent.