# Impact of an Artificial Intelligence-Guided Pulmonary Embolism Response Team (AIPERT) on Patient Transfers, Diagnosis, and Management: A Healthcare System Experience

## TL;DR

An AI tool helped improve the care of pulmonary embolism patients by increasing transfers and advanced treatments across a healthcare system.

## Contribution

Demonstrates real-world impact of AI in enhancing acute PE care through improved triage and intervention rates.

## Key findings

- AI implementation led to a 100% increase in PE transfers in Year 1 and 108% in Year 2.
- Catheter-based thrombectomy rates increased significantly after AI deployment.
- Hospital length of stay decreased following AI-assisted triage implementation.

## Abstract

Background: Pulmonary embolism (PE) is a time-sensitive condition with variable clinical presentations and outcomes. Rapid risk stratification and appropriate triage are essential for optimizing treatment and patient outcomes. Artificial intelligence (AI) offers an opportunity to enhance clinical decision-making, yet its real-world applications remain limited. Objective: The objective of this study was to describe a single healthcare system’s implementation and early experience with an AI-enabled triage tool for pulmonary embolism patients across a multi-hospital network. Methods: This retrospective observational study evaluated the deployment of an AI-based clinical decision support system within a healthcare network. The AI tool detected PE and right ventricular (RV) strain and alerted the PE response team (PERT) to facilitate timely transfer and intervention. Three cohorts were evaluated: pre-AI, Year 1 post-AI, and Year 2 post-AI. Outcomes included transfer volumes, advanced therapy rates, and hospital length of stay (LOS). Results: A total of 183 PE transfer patients were analyzed: 36 pre-AI, 72 in Year 1 post-AI, and 75 in Year 2 post-AI. Transfers increased by 100% in Year 1 (p = 0.0005) and 108% in Year 2 (p = 0.011) compared to pre-AI. Catheter-based thrombectomy increased from 10 pre-AI to 18 in Year 1 (+80%, p < 0.0001) and 28 in Year 2 (+180%, p = 0.0006). After-hours diagnosis rose from 69.4% pre-AI to 70.8% in Year 1 (p = 0.027) and 77.3% in Year 2 (p = 0.088). Surgical embolectomy showed a borderline increase in Year 2 (p = 0.04), though case numbers were small. Conclusions: Implementation of an AI-assisted triage platform for PE was associated with sustained increases in interhospital transfers and advanced interventions, and a reduction in hospital length of stay. These findings support the potential for AI to standardize and expedite acute PE care in a multi-hospital health system.