Abstract 4368940: Contemporary Demographics, Management, and Outcomes in Acute Pulmonary Embolism: A Pre- and Post-AI Era Comparison

The paradigm for the treatment of acute pulmonary embolism (PE) has been changing with the introduction of new interventional devices and techniques. Concurrently, computer-aided smart applications for automatic PE identification on computed tomography angiography (CTA) have been recently integrated in clinical practice to help triage acute PE and alert the treatment teams. We hypothesize that artificial intelligence (AI) integration changed PE diagnosis, treatment, and outcomes compared to the pre-AI period. The current study compares the presentation, risk stratification, treatment patterns and outcome of acute PE patients before and after the integration of AI-assisted diagnostic tools. For the pre-AI period, PE patients were identified retrospectively through chart reviews of cases flagged with PE during 2022 and 2023 across five hospitals in an urban healthcare system. For the post-AI period, an FDA-approved AI application was used to prospectively identify all PE cases automatically detected on CTA over a 6-month period in 2024 at the same hospitals. All CTA scans were reviewed, and true-positive cases were included. Of the 2,013 screened, 1,489 patients were included: 501 pre-AI patients and 988 post-AI patients. The subjects had a mean age of 60.8 ± 17.4 and 62.8 ± 16.1, respectively (p= 0.03). The majority in both groups were female and African American / Black. In the pre-AI, the majority of patients were classified as low-risk (49.3%) while most patients were classified as intermediate-risk (57.5%) post-AI integration (p<0.001). Advanced therapy (mechanical thrombectomy, catheter directed lysis, thrombolysis, mechanical support or surgery) was used in a similar pattern during the pre-AI period and post-AI period (8.6% vs 9.4% ; p=0.63). In-hospital mortality (6.8% vs. 9.4%; p=0.09) and PE- related deaths (2.8% vs. 3.4%; p=0.5). The AI application had a 12.7% false positive rate. Smart-phone based application helped identify a larger number of PE patients in clinical practice, changed risk stratification, led to more interventions, but did not change the treatment pattern or outcome.