Abstract
BACKGROUND: Mortality in psychiatric hospitals is both elevated and poorly understood. We aimed to address a knowledge gap about cause-specific mortality among current psychiatric inpatients using a meta-analytic synthesis of primary research reporting on the natural, unnatural, and cause-specific mortalities. METHODS: We conducted a systematic review and meta-analysis according to PRISMA and MOOSE guidelines. Searches for peer reviewed English language papers published between 1 January 1960 and 24 July 2025 and indexed in MEDLINE, PsycINFO, EMBASE and PubMed were supplemented using searches of grey literature and hand searches of reference lists. Data were pooled using random effects meta-analysis. Mortalities were assessed by cause-specific estimates of. i) the percentage of deaths, ii) mortality per 10,000 admissions and iii) mortality rates per 100,000 person-years. Temporal trends were examined using mixed effects meta-regression. A mixed effects model was used to examine subgroups within each mortality according to publication date and the national income of the study setting. This study is registered with PROSPERO (CRD42024572461). FINDINGS: Thirty-eight studies published from diverse geographic settings over the last five decades were included in a meta-analysis of the percentage of mortality according to cause. Of these, twenty-eight studies were included in a meta-analysis of the mortality per 10,000 admissions and twenty-five studies were included in a meta-analysis of mortality rates per 100,000 patient-years. Natural causes accounted for 81.6% (95% confidence interval (CI) 75.7-86.3%, I-square = 98.5) of inpatient deaths. Vascular causes accounted for 31.0% (95% CI, 27.0-35.2%) of deaths, and infection accounted for 17.8% (CI 11.8-26.0%), The pooled estimate of natural deaths per 10,000 admissions was 78.9 (95% CI, 43.1-143.9, I-square = 99.8), falling from 246.6 (95% CI, 132.9-453.1) in pre-2000 studies to 35.2 (95% CI, 18.8-65.8) in more recent studies. The pooled estimate of the natural mortality rate per 100,000 patient-years was 2905.0 (95% CI, 2350.8-3459.2, I-square = 99.4), with rates falling from 5445.4 (95% CI, 4460.9-6429.9) in pre-2000 studies to 1374.3 (95% CI, 968.3-1780.2) in more recent studies. Suicides and accidents respectively accounted for 4.8% (95% CI, 3.0-7.7%) and 3.3% (2.3-4.7%) of deaths. Unnatural mortality rates have been stable over time. All mortality measures had high between study heterogeneity that was not well explained by the available moderators of year of publication and national income of the study setting. INTERPRETATION: Mortality rates in psychiatric hospitals are highly heterogeneous but have been falling over time. Most inpatient deaths have a natural cause. Our findings suggest a need for primary research examining the demographic and diagnostic associations with inpatient mortality, while underscoring the opportunities for better integration of medical and mental health care. FUNDING: None.