Automated ascertainment of bleeding and target lesion revascularization after percutaneous coronary intervention (PCI) using electronic health record (EHR) data

PROJECT SUMMARY Percutaneous coronary intervention (PCI) is the most common cardiac procedure with over 650,000 PCI performed annually in the U.S. Post-PCI complications which occur in a significant proportion of patients are associated with an increased risk of morbidity and mortality. Reliable ascertainment of post-PCI events is

important for performance measurement, submission to disease registries, clinical trials, and for cardiac catheterization laboratory (CCL) safety monitoring. Claims based detection of PCI complications is inadequate. Assessing post-PCI events reliably requires an in-depth manual chart review, which incurs a significant

provider and administrative burden. However, with advances in health information technology and nationwide adoption of electronic health record (EHR) systems, it possible to utilize EHR for the automatic derivation of clinical events. Dr. Murugiah proposes to create and validate automated algorithms which can be applied to

EHR data to detect two important post-PCI events which are a common focus of clinical trials and quality improvement efforts – in-hospital bleeding and 1-year target lesion revascularization (TLR). Using EHR data at a large health system, Dr. Murugiah will develop a hybrid algorithm to detect major bleeding post-PCI by

leveraging structured data fields such as laboratory values, as well as unstructured data such as imaging reports, cardiac catheterization reports, and progress notes incorporating Natural Language Processing (NLP) techniques (Aim 1). Similarly, using cardiac catheterization reports for patients undergoing repeat

revascularization within 1-year, an algorithm will be developed to detect TLR (Aim 2). Both algorithms will be externally validated using EHR data from another large institution. The final algorithm will be implemented into a tool generating scheduled reports of bleeding and TLR, to be fed back to the quality assurance team for the

CCL and to individual operators. Individual operators will be surveyed to obtain feedback about the algorithm, reporting process, and their perceived benefit. The final tools will be made open source (Aim 3). An automated algorithm for the detection of post-PCI events within EHR can reduce administrative burden, enable the

generation of new knowledge from EHR based observational studies, and enable pragmatic clinical trials. Further, this project can serve as a proof of concept of the utility of hybrid tools leveraging both structured data and clinical text for surveillance and quality measurement. Dr. Murugiah has a career interest in studying and

improving the treatment for ischemic heart disease using multidimensional datasets and EHR data to develop real time risk prediction models and decision support tools, and conduct EHR based comparative effectiveness studies and clinical trials. During the award period he will leverage the experience of his mentorship team

which includes national experts in cardiovascular outcomes research, clinical informatics, and computational linguistics. He will also acquire formal training in clinical informatics by completing a Master of Health Science degree which will provide him the necessary platform to make the transition into an independent investigator.