Mount Sinai Research Team Developed Machine Learning-Based Model to Help Healthcare Organizations Predict Mortality Risk for Individual Cardiac Surgery Patients, Offering Significant Performance Benefits Compared to Current Population-Derived Models .
A new data-driven algorithm built on a repository of electronic health records (EHRs) is the first institution-specific model for assessing risk in heart patients preoperatively, enabling providers to You will be able to pursue the optimal course of action. . The team’s research Open Journal of Thoracic and Cardiovascular Surgery (JTCVS).
Currently used standard-of-care risk models have limited applicability to certain types of surgery, excluding a significant number of patients undergoing complex or combination surgeries for which models do not exist. Using a rigorous combination of electronic medical record data and machine learning techniques, our team is the first to demonstrate how individual hospitals can build their own risk models for mortality after cardiac surgery. ”
Dr. Ravi Iyengar, Senior Author, Dorothy H. and Louis Rosenstill Professor of Pharmacology, Icahn School of Medicine, Mount Sinai, Director, Mount Sinai Systems Biomedical Institute
Predictive models based on machine learning algorithms have been generated across different areas of medicine, and some predictive models have shown improved results over their standard-of-care counterparts. In cardiac surgery, the Society of Thoracic Surgeons (STS) risk score is considered the gold standard and is routinely used to assess surgical risk in cardiac surgery patients. While these continue to provide important benchmarks for hospitals to assess and improve their performance, they are based on population-level data and therefore require coordinated preoperative assessment and complex surgery. It may not be possible to accurately predict risk for certain patients with complex medical conditions.
Cardiovascular surgeons and data science experts at Mount Sinai Hospital, under the supervision of Dr. Gaurav Pandey, co-senior author and Associate Professor of Genetics and Genomic Sciences at Icahn Mount Sinai. Machine learning-based models using EHR data provide an effective solution. So they used his regularly collected EHR data to create a rigorous machine learning framework to develop a patient-tailored, hospital-specific postoperative mortality risk prediction model. bottom. It implicitly incorporates important information about the Mount Sinai patient population, including demographics. Socioeconomic factors, and health characteristics. This is in contrast to population-derived models like the STS, which are based on data from diverse health systems in different regions of the country. Further driving the performance of this methodology was a highly effective open-source prediction algorithm known as XGBoost. The algorithm builds an ensemble of decision trees by gradually focusing on less predictable subsets of the training data.
The research was led by co-author Aaron J. Weiss, M.D., Ph.D., a former Cardiothoracic Surgical Resident at Mount Sinai Hospital and a Ph.D. in Clinical Research from Icahn Mount Sinai University, currently works at the Cleveland Clinic. Dr. Arjun Yadoh, adjunct assistant professor of pharmacology at Icahn Mount Sinai and now senior data scientist at the National Center for Advancement of Translational Sciences, part of the National Institutes of Health (NIH NCATS), also collaborated on the study. took the lead. These researchers used XGBoost to model 6,392 of his heart surgeries, including heart valve surgeries performed at Sinai Hospital from 2011 to 2016. Coronary artery bypass graft. Aortic resection, replacement, or anastomosis. Revision heart surgery has been shown to significantly increase the risk of death. The team then compared the model’s performance with the STS model on the same patient set.
This study showed that the XGBoost model outperformed the STS risk score in terms of mortality in all commonly performed categories of cardiac surgery for which the STS score was designed. The predictive performance of his XGBoost model across all surgical types is also high, demonstrating the potential of machine learning and his EHR data for building effective site-specific models.
“Accurately predicting postoperative mortality is critical to ensuring the best outcomes for cardiac surgery patients, and our study favors site-specific models rather than clinical standards based on population data.” is likely to be favorable,” emphasizes Dr. Pandey. “Equally important, we have demonstrated that it is practical for healthcare organizations to develop their own predictive models through sophisticated machine learning algorithms to replace or complement established STS templates. .”
This study was funded by a grant from the National Institutes of Health.
sauce:
Mount Sinai Health System
