Document resource
Introduction In-hospital mortality after acute pulmonary embolism (PE) remains clinically important. Real-world risk determinants may differ from those in trial cohorts and reflect physiological instability, thrombotic burden, and comorbidity. We examined predictors of in-hospital death in an NHS cohort of patients admitted with confirmed PE.Methods Retrospective observational cohort study using routinely collected electronic health record data from an acute hospital (July 2020–February 2025). Adults with confirmed PE and recorded PE risk category were included. The primary outcome was in-hospital death. Candidate predictors comprised demographics; PE severity (haemodynamic instability defined as SBP <90 mmHg); CTPA clot phenotype (level and laterality); escalation of care; biomarkers (including D-dimer, troponin, CRP, full blood count, renal function and albumin); and echocardiographic RV strain where available. D-dimer and troponin were log-transformed (log1p). Univariable associations were assessed by logistic regression. To mitigate sparse data/separation and produce biologically plausible estimates, multivariable analysis used bias-reduced logistic regression (brglm2) with adjusted odds ratios (aOR) and 95% confidence intervals (CI).Results Among 31086 cardiac referrals between July 2020 and February 2025, 188 had confirmed PE. Of these 188 patients, 22 (11.7%) died during the index hospital admission. Death was associated with a higher PE risk category (risk category 3: 31.8% vs 9.6%; p=0.005), haemodynamic instability (13.6% vs 2.4%; p=0.044), cardiac arrest (18.2% vs 0.6%; p<0.001) and ITU admission (27.3% vs 1.8%; p<0.001). D-dimer was higher in non-survivors (17,848.9±12,342.5 vs 7,744.9±7,649.1; p<0.001). In the bias-reduced multivariable model, haemodynamic instability (aOR 56.4, 95% CI 2.89–1,100; p=0.008), higher D-dimer (log1p) (aOR 4.39, 95% CI 1.90–10.1; p<0.001) and type 2 diabetes mellitus (aOR 5.49, 95% CI 1.13–26.8; p=0.035) were independently associated with mortality. Troponin, CRP, renal indices, laterality and echocardiographic RV strain were not independently associated after adjustment. A CTPA saddle phenotype showed an inverse association (aOR 0.02, p=0.009), which should be interpreted cautiously given the low counts and potential confounding.Conclusion In this real-world NHS PE cohort, in-hospital mortality was independently associated with haemodynamic compromise and thrombotic burden (D-dimer), with additional risk from diabetes. These findings support integrated early risk assessment, emphasising physiological instability and biomarker burden to identify patients at the highest risk of deterioration and death.Abstract 460 Figure 1Predictors of Inpatient Mortality in the Context of Acute PE