Sequential Pattern Mining of Longitudinal Adverse Events After Left Ventricular Assist Device Implant

Area of Science:

• Cardiology
• Medical Informatics
• Biostatistics

Background:

• Left ventricular assist devices (LVADs) are crucial for advanced heart failure but carry risks of adverse events (AEs).
• Previous research often studied AEs in isolation, overlooking their sequential relationships and impact on clinical outcomes.
• Understanding AE sequences is vital for improving patient management after LVAD implantation.

Purpose of the Study:

• To explore and identify common sequential chains of adverse events (AEs) following continuous-flow LVAD implantation.
• To investigate the interrelations between AEs and their correlation with clinical outcomes.
• To lay the groundwork for predicting and preventing subsequent AEs in LVAD patients.

Main Methods:

• Utilized a large dataset of 58,575 AEs from 13,192 patients in the INTERMACS registry (2006-2015).
• Employed pattern mining, including sequence creation, hierarchical clustering of AE patterns, and Markov modeling to extract temporal AE chains.
• Analyzed AE sequences to identify distinct groups with common temporal progressions.

Main Results:

• Identified seven distinct groups of sequential AE patterns following LVAD implantation.
• These groups represent common trajectories such as recurrent bleeding, device malfunction, infection, transplant pathways, arrhythmias, neurological decline, and multi-organ failure.
• The findings highlight the interconnectedness of various AEs in the post-LVAD patient journey.

Conclusions:

• Sequential AE patterns exist and are associated with specific clinical outcomes in LVAD recipients.
• Recognizing these temporal AE chains can aid in the prediction and prevention of adverse events.
• This research provides a novel framework for understanding and managing complex AE profiles in advanced heart failure patients with LVADs.