People with severe mental illness (schizophrenia spectrum disorders, bipolar disorder, and severe major depressive disorder) experience a persistent mortality gap compared with the general population. Previous evidence has been dominated by single-country studies and heterogeneous designs, limiting transnational comparability and public health prioritisation. This study applied a harmonised, transdiagnostic framework across five European countries to quantify cause-specific premature mortality before age 75 and to decompose excess mortality into complementary dimensions: relative inequity, absolute excess burden, and severity of premature mortality.
The study used nationwide administrative registers from Denmark, Finland, France, Poland, and Sweden following a pre-registered harmonised protocol. Eligible participants were people aged 15–65 years at first recorded diagnosis of severe mental illness during country-specific study periods (2004–23). Severe mental illness was defined using ICD-10 codes: schizophrenia spectrum disorders (F20–F29), bipolar disorder (F30–F31), and major depressive disorder classified as severe if identified via inpatient contact or receipt of disability pension (F32–F33).
Cohort ascertainment used inpatient and specialised outpatient registers, sickness absence records (>14 days) and disability pensions depending on country availability. Lead-in periods were applied to reduce bias from accumulation of prevalent cases in early observation years. People with missing age or sex were excluded; ethnicity data were not available in the registers.
Cause-specific mortality used ICD-10 groupings: neoplasms (C00–D48), cardiovascular disease (I00–I99), respiratory disease (J00–J99), endocrine and metabolic diseases (E00–E99), gastrointestinal disease (K00–K93), and other natural causes. External causes (V00–Y98) and ill-defined or unknown causes (R96–R99) were excluded from natural-cause analyses; deaths with missing ICD codes were categorised as ill-defined/unknown.
National death registers capture 98–100% of deaths, but cause-of-death data quality varied: Finland and Sweden reported high-quality cause-of-death data; Denmark had some uncertainty in reporting of cause-quality metrics; France had a substantial proportion (12%) of missing causes requiring multiple imputation; Poland had known limitations in death-certificate coding and cause-of-death data were available only for hospital deaths, so Poland was treated separately in cause-specific analyses.
Three complementary indicators were used to characterise excess mortality:
Country-specific estimates were pooled using random-effects meta-analysis (generic inverse-variance method) to account for variability across countries. Poland was included only in pooled all-cause mortality estimates because of non-comparable cause-of-death data.
Across country-specific study periods there were 4,861,795 people with severe mental illness (57% women; mean age 41 years at index diagnosis). Cohort breakdown: 1,460,483 with schizophrenia spectrum disorders (45% women; mean age 38), 669,727 with bipolar disorder (62% women; mean age 41), and 2,731,585 with severe major depressive disorder (62% women; mean age 43). Source of index diagnosis varied by country; overall, inpatient contacts and disability pensions were major contributors to cohort identification. More than half of the combined cohort resided in France.
Health system indicators and avoidable mortality varied between countries: avoidable mortality was highest in Poland and lowest in Sweden and France.
Between Jan 1, 2004, and Dec 31, 2023, the combined cohort accounted for 561,903 deaths from any cause. Pooled sex- and age-standardised all-cause mortality was 2.64-fold higher (SMR 2.64, 95% CI 2.25–3.11) in people with severe mental illness compared with the country-specific general populations, indicating substantial relative excess mortality across countries.
Decomposing excess mortality revealed distinct patterns by cause and diagnosis. Highest absolute excess mortality (SDRe) was observed for cardiovascular disease across diagnoses—for example, schizophrenia spectrum disorders had a reported SDRe of 22.08 excess deaths per 10,000 person-years (95% CI 8.77–35.39), with lower but elevated SDRes for bipolar disorder and severe depression.
By contrast, the largest relative excesses (SMRs) were observed for other disease categories: respiratory diseases showed the highest relative excess in schizophrenia spectrum disorders (SMR 6.49, 95% CI 5.52–7.64), with elevated SMRs also seen for bipolar disorder and severe depression though at lower magnitudes. Endocrine and metabolic diseases and gastrointestinal diseases also showed markedly increased SMRs in schizophrenia and elevated SMRs across other diagnoses.
The analysis identified a dual-burden pattern: some causes contribute most to absolute counts of excess deaths (eg, cardiovascular disease and cancer), while others show the largest relative inequalities (eg, respiratory, endocrine/metabolic, gastrointestinal), and yet others contribute disproportionately to premature years-of-life-lost.
Pre-specified subgroup analyses examined effect modification by sex and age strata (15–34, 35–54, 55–74 years) and conducted sensitivity analyses restricted to narrower diagnostic definitions (eg, schizophrenia-only within F20, bipolar disorder without excluding transient psychoses, depression cohort defined by disability pensions). Cause-specific mortality in Poland was presented separately due to data limitations. The report notes that some subgroup death counts were masked in Denmark when below five.
The mortality gap in severe mental illness is characterised by distinct but interrelated patterns of absolute excess mortality and relative inequality across causes and diagnoses. Considering both absolute and relative dimensions identifies causes that drive the greatest population burden and those that reflect the largest inequities. The authors conclude that reducing premature mortality requires integrated public health approaches combining universal strategies (addressing high-burden causes) with targeted interventions (addressing high-inequity causes), and systemic reforms beyond individual clinical care.
Strengths include a harmonised transnational registry approach covering nearly 5 million people and use of complementary excess-mortality indicators (SMR, SDRe, PYLL) to provide actionable prioritisation. Limitations arise from variability in cause-of-death data quality and completeness across countries—France required multiple imputation for missing cause data, Poland had coding and coverage limitations for cause-specific mortality, and Denmark masked small cell counts—potentially affecting comparability. Ethnicity data were not available. The funder did not influence study conduct or reporting.
The study was funded by the 2024 European Partnership on Transforming Health and Care Systems. The funder had no role in study design, data collection, analysis, interpretation, writing, or the decision to publish.