Individuals with eating disorders have a high mortality risk. Few population-based studies have estimated this risk in eating disorders other than anorexia nervosa.

To investigate all-cause mortality in a population-based cohort of individuals who received hospital-based care for an eating disorder (anorexia nervosa, bulimia nervosa or eating disorder not otherwise specified) in Ontario, Canada.

We conducted a retrospective cohort study of 19 041 individuals with an eating disorder from 1 January 1990 to 31 December 2013 using administrative healthcare data. The outcome of interest was death. Excess mortality was assessed using standardised mortality ratios (SMRs) and potential years of life lost (PYLL). Cox proportional hazards regression models were used to examine sociodemographic and medical comorbidities associated with greater mortality risk.

The cohort had 17 108 females (89.9%) and 1933 males (10.1%). The all-cause mortality for the entire cohort was five times higher than expected compared with the Ontario population (SMR = 5.06; 95% CI 4.82–5.30). SMRs were higher for males (SMR = 7.24; 95% CI 6.58–7.96) relative to females (SMR = 4.59; 95% CI 4.34–4.85) overall, and in all age groups in the cohort. For both genders, the cohort PYLL was more than six times higher than the expected PYLL in the Ontario population.

Patients with eating disorders diagnosed in hospital settings experience five to seven times higher mortality rates compared with the overall population. There is an urgent need to understand the mortality risk factors to improve health outcomes among individuals with eating disorders.

As life expectancy increases, more people have chronic psychiatric and medical health disorders. Comorbidity may increase the risk of premature mortality, an important challenge for health service delivery.

Population-based cohort study in Ontario, Canada of all 11 246 910 residents aged ⩾16 and <105 on 1 April 2012 and alive on 31 March 2014. Secondary analyses included subjects having common medical disorders in 10 separate cohorts. Exposures were psychiatric morbidity categories identified using aggregated diagnosis groups (ADGs) from Johns Hopkins Adjusted Clinical Groups software® (v10.0); ADG 25: Persistent/Recurrent unstable conditions; e.g. acute schizophrenic episode, major depressive disorder (recurrent episode), ADG 24: Persistent/Recurrent stable conditions; e.g. depressive disorder, paranoid personality disorder, ADG 23: Time-limited/minor conditions; e.g. adjustment reaction with brief depressive reaction. The outcome was all-cause mortality (April 2014–March 2016).

Over 2 years' follow-up, there were 188 014 deaths (1.7%). ADG 25 conferred an almost threefold excess mortality after adjustment compared to having no psychiatric morbidity [adjusted hazard ratio 2.94 (95% CI 2.91–2.98, p < 0.0001)]. Adjusted hazard ratios for ADG 24 and ADG 23 were 1.12 (95% CI 1.11–1.14, p < 0.0001) and 1.31 (95% CI 1.26–1.36, p < 0.0001). In all 10 medical disorder cohorts, ADG 25 carried significantly greater mortality risk compared to no psychiatric comorbidity.

Psychiatric disorders, particularly those graded persistent/recurrent and unstable, were associated with excess mortality in the whole population, and in the medical disorder cohorts examined. Future research should examine whether service design accounting for psychiatric disorder comorbidity improves outcomes across the spectrum of medical disorders.