**1. Introduction**

Eating disorders (EDs) are serious and persistent psychiatric disorders characterized by severe disturbance in body weight and eating behavior [1]. Based on the Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition (DSM-5) classification, EDs include anorexia nervosa, bulimia nervosa, binge eating disorder, other specified feeding or eating

**Citation:** Pedram, P.; Patten, S.B.; Bulloch, A.G.M.; Williams, J.V.A.; Dimitropoulos, G. Self-Reported Lifetime History of Eating Disorders and Mortality in the General Population: A Canadian Population Survey with Record Linkage. *Nutrients* **2021**, *13*, 3333. https:// doi.org/10.3390/nu13103333

Academic Editor: Lourdes Varela

Received: 31 July 2021 Accepted: 19 September 2021 Published: 23 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

disorders, and avoidant/restrictive food intake disorder [2]. Based on the systematic review of 33 studies, the lifetime prevalence of EDs (total) is 8.4% for women and 2.2% for men [1], that of anorexia nervosa is 1.4% for women and 0.2% for men, that of bulimia nervosa is 1.9% for women and 0.6% for men, and finally, that of binge eating disorder is 2.8% for women and 1.0% for men [1]. EDs are associated with significantly impaired health-related quality of life compared with the healthy population and even with those with other psychiatric conditions [3]. These disorders encompass a range of problematic behaviors, including starvation, binge eating, and purging, leading to an increased risk of premature death [4]. Many factors in these patients have also been identified as predictors of mortality, such as type of ED diagnosis, low body mass index (BMI), suicide behaviors, alcohol abuse, and comorbidities [5–7]. In addition, the age of onset and age of treatment are also two important death-predictive factors in patients with EDs, as evidence has shown that older individuals (25–44 age group) have an elevated risk of mortality for all types of EDs compared to youth (15–24 age group) [8]. Mortality data on EDs are important and considered an indicator of illness severity [9]. Most mortality studies in EDs have focused on anorexia nervosa; however, a few studies of bulimia nervosa and other specified feeding or eating disorders, and even fewer for binge eating disorder, have been published [4]. The standard measures for mortality are the crude mortality rate (CMR) (CMR is the proportion of death within the study population over a specific period) [10], the standardized mortality rate (SMR) (SMR is calculated using the number of observed deaths in a targeted population at a certain point of time divided by the number of expected deaths in the general population while taking into account certain demographic variables) [11], and hazard ratios (HR) (the HR in survivorship curves is the temporal progression of death within a group and defined as the hazard in the groups with EDs divided by the hazard in the control groups) [12,13].

The ED mortality studies are largely based on cohorts from inpatient settings or case registers covering a circumscribed geographical area, such as the catchment area of a hospital, while relatively less is known about the general population [9,14,15]. For example, a retrospective Canadian cohort study, using administrative healthcare data of 19,041 individuals with ED from 1990–2013, showed that individuals with EDs identified in hospital settings had roughly a five-fold higher mortality rate relative to the general population [16]. The age, sex, and place of residence-adjusted HR for all-cause mortality of EDs in a longitudinal study in Finland among 2450 adults referred to a tertiary care-level ED unit was 3.54 (95%CI 2.52–4.96) [14]. The mortality associated with different types of mental disorders, including EDs, schizophrenia, mood disorders, personality disorders, and behavioral disorders, was also investigated in a recent population-based cohort study in Denmark on 7,369,926 people (23,196 persons with ED) younger than 95 years of age from 1995–2015. Similar to the previous studies, diagnosed individuals in this study also only included those with mental disorders registered in psychiatric inpatient, outpatient, and emergency settings. Therefore, one of the limitations of this cohort was a vulnerability to selection bias arising from a lack of representation of patients who are only treated by a general practitioner or who do not seek specialized help for their mental health. The CMR per 1000 person-year for EDs in this study was 3.0; however, those for schizophrenia and mood disorders were higher (28.1 and 31.2, respectively) [17]. Nevertheless, these aforementioned findings on mortality of EDs may not be representative of a general population and may not be generalizable to a greater range of cases of EDs. They may represent a very small and distinct proportion of the wider EDs population [18–20]. Since it has been revealed that despite the elevated contact with health care services among people with EDs, somewhere between 67% and 83% of cases fail to engage with treatment after referral [21,22]. A wide range of factors lead to this unmet need for treatment, such as difficulty accessing specialist services, the financial cost of treatment, perceived shame, and stigma attached to EDs [19,21]. Therefore, an important question to be answered is whether EDs are associated with higher mortality in the general population.

For epidemiological surveys of EDs in the general population, a self-reported current and lifetime screening tool using single-item questions, such as "Have you ever had anorexia", has been shown to have a reliable specificity and sensitivity [23]. Moreover, the validity of self-reported health in population-based studies has repeatedly been confirmed and even found to be a stronger predictor of associated mortality than instruments explicitly designed for this purpose [24]. However, to our knowledge, there is no empirical research available to assess the mortality of EDs based on the self-reported current and lifetime history of EDs in the general population. Nevertheless, these findings will assist health care providers and policymakers in public health messages about the serious consequences of eating disorders. While this study was conducted in Canada, the results are likely to apply to the general populations of other developed countries. However, to the extent that cultural, health system, or other between-country differences may affect the association of EDs with mortality, the results are most directly supportive of a need to address the issue of mortality in people with EDs in Canada. Hence, the main objective of the current study was to investigate the mortality of self-reported lifetime history of EDs among participants in the population-representative Canadian Community Health Survey linked to national mortality data, which covers all of the general Canadian population.

#### **2. Materials and Methods**

### *2.1. Data Source*

The data source was the Canadian Community Health Survey (CCHS) 1.2, also known as the CCHS mental health and well-being survey conducted by Statistics Canada [25]. The CCHS is a cross-sectional survey that collects information related to mental health status, mental health care utilization, and mental health determinants for the Canadian population. This survey was conducted in 2002 (between May and December), and the sample size was 36,984 with a 77% response rate (Figure 1) [26]. The sampling method was a multi-stage stratified cluster design. The inclusion criteria in this national survey consisted of noninstitutionalized people aged 15 years or older living in private dwellings in the 10 Canadian provinces [26,27]. The exclusion criteria contained individuals residing in the three territories, on reserves and other Aboriginal settlements in the provinces, the clientele of institutions, children aged 15–17 that are living in foster care, the full-time members of the Canadian Forces, and residents of some remote areas, groups that, in total, exclude less than 3% of the general population [26,28]. A "share file" that included the participants who provided consent for their data to be linked to other data sources was subsequently linked, by Statistics Canada, to the Canadian Vital Statistics Database (CVSD), allowing confirmation of vital status and, where relevant, and date of death, see Figure 1. A detailed description of data linkage procedures and their quality assessment has been reported elsewhere [29,30]. The linked data are available to researchers through the Canadian Research Data Centres Network. The current analysis took place in the Prairie Regional Data Centre on the University of Calgary Campus [29,31].

**Figure 1.** Flow diagrams for data linkage (estimates are rounded in keeping with Statistics Canada data release guidelines). † CVSD, Canadian Vital Statistics Database.

#### *2.2. Measures*

The demographic information of the entire population, including age, sex, employment status (currently working versus not employed), the highest educational level, and marital status in three groups: "Married/Common-law" (in Canada, common-law and legally married spouses have very similar legal standing [29]), "Single", i.e., never married, and "Widow/Separated/Divorced" were collected using field-tested items. Diagnosis of the EDs, schizophrenia, and mood disorders such as depression, bipolar disorder, mania or dysthymia, and post-traumatic stress disorder (PTSD), were assessed in each survey's "chronic conditions" module. The wording of the item was similar in each survey: "Remember, we're interested in conditions diagnosed by a health professional". For example: "Do you have an eating disorder such as anorexia or bulimia?". The interview questions for these disorders were those of a Canadian adaptation of the World Mental Health version of the Composite International Diagnostic Interview (WMH-CIDI) [26]. The WMH-CIDI is a lay-administered psychiatric interview that generates a lifetime profile of a person with a disorder defined partly according to both the 10th version of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) and the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) [26].

BMI was calculated by dividing the self-reported participants' weight in kilograms by the square of their self-reported height in meters (kg/m2). The low BMI for participants younger than 18 years old was defined as less than the tenth percentile for age and sex [32], and for older participants below 18.5 kg/m2, as recommended by the World Health Organization (WHO) criteria [33,34]. Binge drinking was assessed with a question in the CCHS survey "How often in the past 12 months have you had five or more drinks on one occasion?". Binge drinking was defined if the response was "more than once a month". Smoking to cope with stress was determined based on the question "When dealing with

stress, how often do you try to feel better by smoking more cigarettes than usual?" if the response was "sometimes" or more.

All of the variables mentioned above were assessed using field-tested survey modules developed by Statistics Canada, which were made available in multiple languages. Detailed information on each of the surveys, questionnaires, and user guides are available from the 2002 mental health survey documentation [25,29,35].

#### *2.3. Data Analysis*

The interview date for each participant is recorded in the survey, which is considered the baseline time point. The participants were classified as exposed or not exposed to EDs according to their answers in the survey. Time to death was calculated as the difference between the date of death and the interview date. In the linked data, both the date and underlying cause of death for those who subsequently died were recorded. Time to the event (death) was calculated by subtracting the date of the baseline interview from the date of death [29]. Those surviving up to the linkage date (31 December 2017) were censored at that date. Covariates were derived from the survey dataset and were, therefore, all recorded at baseline. No information on time-varying covariates was available.

The stratified multi-stage sampling procedure used in CCHS leads to unequal selection probabilities. Therefore, the estimates should be analyzed using a set of 500 replicate bootstrap-sampling weights to ensure population representativeness. A procedure for bootstrap-weighting recommended by Statistics Canada was used in this analysis [28].

As in most studies of mortality, a time-to-event method of analysis was considered most appropriate due to its ability to address differing person time contributions by study participants. Therefore, after confirming the proportional hazard assumption, a Cox proportional hazards model was used to estimate hazard ratios from the time to event data [36] using Stata version 16. An unadjusted HR (crude HR) was estimated using a model that contained EDs but not covariates. While the CCHS dataset contains a rich set of covariates, the number of participants with EDs who died was too small to produce models with multiple covariate adjustments; therefore, covariates in this study included age (treated as a continuous variable) and sex. Confounding was observed only by age; however, to be consistent with other EDs mortality studies, the impact of sex and both age and sex were also assessed by adding each variable (along with an interaction term) one at a time to the model. For all analyses, the alpha level was set at 0.05.

#### **3. Results**

This study sample was restricted to respondents age 15 years and older at the time of the survey (consisting of approximately 31,130 respondents and mean age = 43.95 years). As shown in Table 1, 0.5% reported a positive lifetime history of EDs. About 14.4% of these individuals died, 89% were female, 52.5% were single, most had a post-secondary graduate degree (64.4%), and 11.5% reported a low BMI.

A Cox proportional hazard model showed that the unadjusted HR (crude HR) for the lifetime history of an ED was 1.35 (95% CI 0.70–2.58). When the model was adjusted for age and then both age and sex, the HR increased to 4.4 (95% CI 2.2–8.0) (data not shown) and 4.54 (95% CI 2.33–8.84, Table 2), respectively.

