Back to 2026 Abstracts
Artificial Intelligence Improves Survival In Operative And Non-operative Pulmonary Embolism
Joshua Meredith, MD, Patrick Muck, MD, Adam Reichard, MD.
Trihealth/Good Samaritan, Cincinnati, OH, USA.
OBJECTIVES: Pulmonary embolism (PE) continues to be the third most common cause of cardiovascular death despite significant technological advancements in treatment. Time to initiation of anticoagulation (AC) has been found to correlate with mortality and pulmonary embolism response teams (PERT) focus on early initiation of AC in these patients. The question remains whether artificial intelligence (AI) systems can help us improve outcomes in these patients. We aimed to evaluate time to anticoagulation therapy (TAC), mortality, hemorrhage, and ECMO utilization before and after implementation of AI PERT alerts.
METHODS: This study is a retrospective cohort study evaluating all patients presenting with pulmonary embolism and undergoing treatment at our tertiary referral center. The two cohorts were established as Pre-AI, and Post-AI. Patient charts were reviewed for demographic information and study outcomes. Sub-analyses evaluated mortality and hemorrhage for systemic tPA patients and High-Risk PE patients. Continuous variables were evaluated with Mann Whitney U tests and categorical variables using Fischer Exact tests.
RESULTS: Patient demographics and study data are listed in Table 1. 124 patients were included in Pre-AI and 135 in Post AI with 218 patients having adequate data to determine TAC. Average TAC was 160 minutes Pre-AI versus 74 minutes Post-AI (p-value 0.41). In-hospital, 30-Day, and 1-Year mortality decreased significantly Post-AI (p values 0.012, 0.009, and 0.004). ECMO utilization did not differ significantly between the groups. Mortality for intermediate-risk patients was 1.2%, for High-Risk was 25.3%, and for all patients 9.3%. Systemic tPA recipients had a mortality of 22.2% compared to 2.4% mortality for CBT patients (p-value <0.0001). Major hemorrhage occurred in 6.9% of patients with 12.2% in the systemic tPA patients and 4.1% in the CBT patients (p-value 0.02). In the High-Risk patient subset, mortality and hemorrhage did not differ significantly by treatment modality (p value 0.064).
CONCLUSIONS: Patient survival improved post initiation of AI PERT alert but TAC did not differ statistically between the Pre-AI and Post-AI patients. Despite the average difference of 86 minutes, there were significant outliers which may be remedied by larger volume studies. Systemic tPA utilization was associated with increased major hemorrhage and mortality compared to CBT.
Back to 2026 Abstracts
