**Table 1.** Characteristics of the population †.

† CI—Confidence Interval, BMI—Body Mass Index (definition of low BMI: for ages <18, below the 10th percentile for age and sex, and for ages >18, below 18.5 kg/m2).

**Table 2.** Crude and age/sex-adjusted effect of the self-reported major mental health problems on mortality †.


† CI—Confidence Interval, HR—Hazard Ratio, PTSD—Post-Traumatic Stress Disorder, BMI—Body Mass Index (definition of low BMI: for ages < 18, below the 10th percentile for age and sex, and for ages >18, below 18.5 kg/m2). \* Indicates statistical significance at the *p* < 0.05 level.

For purposes of comparison, we generated unadjusted and age/sex-adjusted HR for other self-reported history of mental health disorders and low BMI in the same population. As shown in Table 2, the unadjusted HR for schizophrenia/psychosis (1.44, 95% CI 1.04–1.99) was higher than the crude HR for ED, however, the age/sex-adjusted HR (2.02, 95% CI 1.24–3.31) was smaller than the age/sex-adjusted HR for EDs. Similarly, the age/sex-adjusted HR for the self-reported history of mood disorder and PTSD was lower than that of eating disorders. For comparison, age/sex-adjusted HRs for other potential determinants of mortality are also presented in Table 2: binge drinking, smoking to cope with stress (which is here considered a proxy for smoking), and low BMI.

### **4. Discussion**

To our knowledge, this is the first study that evaluated all-cause mortality of selfreported lifetime history of all EDs in a general population. This study confirms the high mortality in individuals with a lifetime history of ED that is not due to the selection of severely ill respondents as in prior studies of clinical individuals. However, as the prevalence of EDs was low, it is likely that these general population estimates also reflect a subset of the entire population with EDs. Previous studies reported mortality in a broad category of EDs [8,16,17] or with DSM or ICD diagnosis of anorexia nervosa [37,38], bulimia nervosa [37,38], other specified feeding or eating disorders [37], or binge eating disorder [14] based on a selective population, such as health-administrative data, inpatientssettings, or outpatients care. Therefore, the generalization of these results to all patients with EDs was limited only to those in treatment settings [14]. The current study helps to confirm that mortality is substantially elevated in members of the general population who report that they have been diagnosed with an ED, including those with subclinical eating disorders or those who failed to engage with treatment once referred. There is widespread stigma toward people affected by EDs by the general public, medical professionals, and service users due to the attribution of personal responsibility for illness behaviors [39]. Therefore, the seriousness of EDs may be underreported as a result of individuals who trivialize and minimize associated psychological and medical challenges [40]. There is also a self-stigma in people with EDs that leads them to deny, minimize, or assign positive meaning to their behaviors [39,41]. The effect of public/self-stigma may lead people to be reluctant to seek treatment, leave treatment prematurely, or experience a loss in the necessary components of recovery [42]. The current findings could be integrated into psychoeducation material for adolescents/adults with ED and their caregivers to help challenge the view that these illnesses are not serious, necessitating treatment. Public health campaigns could further highlight the high mortality rate of EDs [43].

Evidence has shown that mortality in EDs is influenced by age, sex, and case severity [6,9,16,44]. Generally, older individuals have an elevated risk of mortality for all types of EDs compared to the mortality of younger individuals [8], probably reflecting the strong effect of age and chronicity of illness on mortality. Consistent with this, age-adjusted estimates show a strong association with mortality. In a dataset of English national Hospital Episode Statistics, the SMR for the 15–24 age group diagnosed with anorexia nervosa or bulimia nervosa was found to be 11.5 and 4.1, respectively [8]. The SMR for the 25–44 age group diagnosed with anorexia nervosa or bulimia nervosa was found to be 14.0 and 7.7, respectively [8]. A number of studies have also reported sex differences in ED mortality [44–46]. Although the lifetime prevalence of EDs among males is lower than females, the CMR and SMR observed in males are almost two-fold higher than in females [16,17]. These provide evidence for the potential impact of age and sex as covariates in the all-cause EDs mortality assessment. Therefore, in the current study, the HR for eating disorders was reported as adjusted for age, sex, and both age and sex.

A naïve interpretation of crude HRs suggests only a weak effect of EDs (1.35, 95% CI 0.70–2.58, since the low end of the interval is below one); however, the age/sex-adjusted HR for EDs is much higher than that of other major mental health problems. This indicates that the unadjusted estimates were strongly confounded by age and sex, information that will be useful in planning the analysis of future studies. The direction of the confounding is predictable, since EDs were reported more often by younger people and by women (see Table 1), both of whom have lower rates of mortality.

In this study, only for the purpose of comparison, the mortality associated with some self-reported lifetime major mental health problems, and low BMI calculated based on selfreported height and weight were also evaluated. Evidence shows that low BMI associated with EDs may affect mortality [4,7,47]. About 12% of the individuals with a self-reported lifetime history of ED and 2% of the entire population reported a low BMI. However, the causes of this low BMI are not distinguishable between those who were initially underweight and those who became underweight due to malnutrition, over-exercising, or other comorbidities [48]. The present study aimed to evaluate all-cause mortality of self-reported history of EDs; however, numerous previous studies have reported a different degree of malnutrition in patients with EDs [49–51]. Several lines of evidence on the effect of low BMI caused by malnutrition on mortality meets the key Bradford Hill's criteria to establish this causal relationship [52–55].

In the current study, the age/sex-adjusted HR for lifetime history of EDs was over twofold higher than the age/sex-adjusted HR for self-reported lifetime schizophrenia/psychosis. In this study, the self-reported lifetime schizophrenia/psychosis was based on the question asking whether they have schizophrenia or any other psychosis as diagnosed by a health professional (response options, yes or no). A previous study on the CCHS1.2 survey has shown that the prevalence estimates of these two self-report survey items provide what appears to be a plausible epidemiologic pattern [56]. Accordingly, the mortality associated with these two items might also follow the same pattern. A nationally representative cohort study in the UK using primary care electronic health records on over 11 million people reported a very similar adjusted HR (accounting for age, gender, calendar year, area-level deprivation, ethnicity, and the average number of visits to the physician per year of follow-up) for schizophrenia (2.08, 95% CI 1.98–2.19) to our study [57]. In the current study, the results in the general population are in line with the previous studies on the clinical samples that observed lower crude HR and higher sex/age-adjusted HR in EDs than schizophrenia [8,17,58,59].

Although the sex ratio of the participants in this study was 1:1, males reported a history of lifetime ED nearly eight times less than females (11.86% vs. 88.98%). This difference is much larger than the existing literature on the lifetime prevalence of EDs among different sexes measured with a standardized tool (2.2% vs. 8.4%) [1]. This discrepancy may be associated with the difference between females and males in diagnosis, self-identifying, and self-reporting lifetime EDs due to the immense stigmatization toward males with EDs, stereotypes linked to EDs, and a misdiagnosis by a specialist treatment center [60–62]. These current findings may indicate the importance of gendered issues in diagnosing and treating patients with EDs and the necessity of tailored services to the patient's need, such as the same-sex therapeutic groups [61].

In the current study, mood disorders were not associated with mortality, which is surprising since CCHS respondents (linking data from four surveys: CCHS 1.1 (2000/2001), CCHS 1.2 (2002), CCHS 2.1 (2003/2004), and CCHS 3.1 (2005/2006)) with symptoms of major depressive episodes (according to versions of Composite International Diagnostic Interview, a structured interview that does not depend on help-seeking) do have elevated all-cause mortality [29]. As the variable included in this analysis was self-reported diagnoses of mood disorders, it is possible that help-seeking and potentially treated individuals with major depression do not have elevated mortality, in contrast to what was reported here for EDs.
