Mount Sinai researchers develop an institution-specific machine learning model to improve mortality risk prediction in cardiac surgery patients.
A recent research breakthrough has revealed a machine learning-based model that dramatically enhances the ability to predict mortality risk for individual patients undergoing cardiac surgery. Developed by a team of researchers at Mount Sinai, this pioneering model offers significant advantages over existing population-derived models, thereby reshaping the landscape of healthcare delivery.
A New Benchmark in Healthcare Delivery
The revolutionary model, based on a wealth of electronic health records (EHR), is the first institution-specific algorithm for assessing a cardiac patient’s risk ahead of surgery. The study was published in The Journal of Thoracic and Cardiovascular Surgery (JTCVS) Open.
Dr. Ravi Iyengar, the senior author of the study, noted, “The standard-of-care risk models used today are limited by their applicability to specific types of surgeries, leaving out significant numbers of patients undergoing complex or combination procedures for which no models exist.”
He further added, “Our team rigorously combined electronic health record data and machine learning methods to demonstrate for the first time how individual institutions can build their own risk models for post-cardiac surgery mortality.
Machine Learning Models in Medical Field
Machine learning has transformed a wide array of medical fields, with prediction models often surpassing their standard-of-care counterparts. The Society of Thoracic Surgeons (STS) risk scores, often considered the gold standard for evaluating a cardiac surgery patient’s procedural risk, are derived from population-level data. However, these models may fail to accurately predict risk for specific patients with complicated pathologies requiring tailored preoperative evaluations and complex surgeries.
A Paradigm Shift in Risk Assessment
Cardiovascular surgeons and data science specialists at The Mount Sinai Hospital, under the supervision of Dr. Gaurav Pandey, hypothesized that machine learning-based models could offer a robust solution. Utilizing routinely collected EHR data from their own institution, they developed a risk prediction model that is personalized for the patient and specific to the hospital. This approach allowed for the incorporation of crucial data about Mount Sinai’s patient population, including demographics, socioeconomic factors, and health characteristics.
Utilising XGBoost for Predictive Analysis
Central to the performance of this methodology was an open-source prediction algorithm known as XGBoost, which builds an ensemble of decision trees by progressively focusing on harder-to-predict subsets of training data. The researchers used XGBoost to model 6,392 cardiac surgeries performed at The Mount Sinai Hospital from 2011 to 2016.
Outperforming Conventional Models
The study demonstrated that the XGBoost model outperformed STS risk scores for mortality across all commonly conducted categories of cardiac surgery. The XGBoost model’s performance was consistently high across all types of surgery, highlighting the potential of machine learning and EHR data in developing effective institution-specific models.
Dr. Pandey emphasized, “Accurate prediction of postsurgical mortality is critical to ensure the best outcomes for cardiac surgery patients, and our study shows that institution-specific models may be preferable to the clinical standard based on population data.”
Implications and Future Directions
The success of this study underscores the practicality of healthcare institutions developing their own predictive models through sophisticated machine learning algorithms. These models can potentially replace or complement the established STS template, thus opening up a new frontier in personalised patient care.
This transformative research was generously funded by grants from the National Institutes of Health, underscoring the importance of such efforts in the pursuit of improving patient outcomes in cardiac surgery and beyond.
As the medical community continues to leverage machine learning and AI technologies, the future of healthcare seems bright, with the promise of more personalised, efficient, and effective treatments. Indeed, the adage ‘one size fits all’ may soon become a thing of the past in medicine, replaced by a more nuanced, individualised approach to patient care.