Where and how people with schizophrenia die: a population-based, matched cohort study in Manitoba, Canada

To compare place and cause of death for people with and without schizophrenia in Manitoba, Canada.

By using deidentified administrative databases at the Manitoba Centre for Health Policy, a 1:3 matched cohort of decedents aged ≥ 10 years in fiscal years April 1995-March 2008 (n = 3,943 with schizophrenia; n = 11,827 without schizophrenia) was selected and matched on age, sex, geography, and date of death ± 2 months. Schizophrenia was defined as ICD-9-CM code 295 or ICD-10-CA codes F20, F21, F23.2, or F25 in hospital/physician files at least once within 12 years of death.

The median age at death was 77 years. The attributable percentage of deaths was higher for respiratory illnesses (all ages) and suicide (age 10-59 years only), similar for circulatory illnesses, and lower for cancer in decedents with schizophrenia compared to matched controls. For cancer deaths, decedents with schizophrenia were equally likely to die of gastrointestinal, breast, or prostate cancer, but more likely to die of lung cancer at ages 10-59 (32.5% versus 20.6%, P < .004). Place of death was more likely a nursing home (29.7% vs 13.9%) and less likely a hospital (55.5% vs 70.5%) (P < .0001) for decedents with schizophrenia overall and by specific cause, with the exception of suicide deaths showing no difference by place. Except for those who died in nursing homes, decedents with schizophrenia had higher general practitioner but lower specialist rates and inpatient hospital separations.

Generally, patients with schizophrenia were more likely to die in nursing homes but less likely to die in hospitals. Understanding where these patients die is critical for improving access to quality palliative end-of-life care.