548 - Characterization of ECMO Cannula Performance Using Electronic Health Record-Derived Data

Publication Number: 1527.548

Aditya S. Kalluri, Boston Children's Hospital, Boston, MA, United States; Alireza Akhondi-Asl, Boston Children's Hospital, Boston, MA, United States; Alon Geva, Harvard Medical School, Boston, MA, United States; Dennis Daniel, Boston Children's Hospital, Boston, MA, United States

Background: ECMO cannula performance is dependent on both patient characteristics and mechanical factors, including cannula size, type, and location. Prior efforts to characterize circuit performance have utilized simulated benchtop circuits or computational fluid dynamics models.

Objective: To determine if automated electronic health record data can recapitulate ECMO cannula performance characteristics and inform understanding of cannula performance in diverse clinical scenarios at a high-volume ECMO center.

Design/Methods: We performed a descriptive analysis of anonymized single-center data from 2012-2022, including patient characteristics, ECMO cannulation strategies, and recorded circuit flows and pressures. For each cannula type and location, drainage and reinfusion pressures were plotted against delivered ECMO blood flow and summarized using linear regression. Observed performance curves were compared to manufacturer-reported pressure drop water curves and stratified by patient characteristics.

Results: The dataset contained 633 total ECMO runs, with 62% occurring in the cardiac ICU and the remainder in the medical/surgical pediatric ICU. 67.8% of runs utilized peripheral cannulation. Circuit performance curves paralleled pressure-flow relationships of manufacturer performance curves (Figure 1). Patient-stratified data enabled identification of distinct ECMO performance patterns across patients with the same cannulation strategy, with certain patients demonstrating divergent performance patterns (Figure 2). ECMO performance between different cannulation sites using a given cannula were also compared, with differences in site variability between Biomed cannula sizes in this dataset (Figure 3).

Conclusion(s): Electronic health record-derived ECMO data allows for quantification of real-world circuit performance and patient-circuit interactions, comparison to manufacturer standards, and comparison among patients. This approach contributes to our ability to assess ECMO support in real-world scenarios and may help identify patients with abnormal circuit-patient interaction. Future analyses will examine the association between these and other clinical features beyond ECMO circuit characteristics.

Figure 1: Recapitulation of manufacturer cannula performance characteristics
Linear regression of electronic health record-derived measurements of circuit pressures and ECMO flow was utilized to recapitulate benchtop manufacturer-provided cannula performance curves.

Figure 2: Patient-specific distinctions in cannula performance
Patients with the same cannulation strategy generally demonstrated consistent performance characteristics, although individual patients were identified with distinct cannula performance.

Figure 3: ECMO performance between different cannulation sites
For a given cannula, electronic health record-derived data enabled comparison of ECMO performance between different cannulation sites.