Mount Sinai researchers have developed a machine learning algorithm that allows healthcare facilities to forecast the chance of death for specific patients undergoing cardiac surgery.
The Need for Accurate Mortality Risk Prediction
Accurately assessing the mortality risk of patients undergoing cardiac surgery is of utmost importance. It will allow healthcare professionals to make informed decisions, tailor treatment plans, and allocate resources effectively. Traditional risk prediction models have shown limitations, leading researchers to explore alternative approaches to enhance accuracy and improve patient care.
Introducing the Machine Learning Model
A groundbreaking algorithm tailored to individual institutions has been developed to evaluate the risk of cardiac patients before surgery. By analyzing extensive electronic health records (EHR), healthcare professionals can now make more informed decisions and provide optimal care. The research outcomes were reported this week in The Journal of Thoracic and Cardiovascular Surgery (JTCVS).
Current risk models used in healthcare are limited to certain types of surgeries, disregarding a considerable portion of patients who undergo complex or combined procedures lacking appropriate models. In a groundbreaking study, researchers successfully combined electronic health record data with machine learning techniques to demonstrate how institutions can create their own risk models for predicting post-cardiac surgery mortality.
Machine learning algorithms have been employed across different areas of medicine to create prediction models. Some of these models have surpassed the already employed traditional techniques in terms of effectiveness. The Society of Thoracic Surgeons (STS) risk scores are regarded as the gold standard in the area of cardiac surgery for evaluating patient procedural risk. These scores serve as valuable benchmarks for hospitals to assess and enhance their performance. However, due to their reliance on data from the general population, these scores may not accurately predict the risk for individual patients with intricate conditions who necessitate customized preoperative evaluations and complex surgeries.
Achieving Improved Mortality Risk Prediction
The researchers hypothesized that using machine learning models with their own hospital’s EHR data could be a good solution. To test this, they developed a robust machine-learning framework using regularly collected EHR data. This framework created a risk prediction model for post-surgical mortality that was personalized to each patient and specific to their hospital. They were able to do this by including crucial data about their patient population, including demographics, socioeconomic factors, and health traits. This approach differed from population-based models like STS, which use data from multiple health systems across the country. The researchers also used an effective open-source prediction technique called XGBoost, which improves accuracy by focusing on challenging subsets of training data using a group of decision trees. This technique further enhanced the effectiveness of the technology they developed.
The scientists employed XGBoost to replicate 6,392 heart surgeries conducted at Mount Sinai Hospital from 2011 to 2016. The procedures that are associated with the increased death risk included heart valve replacements, coronary artery bypass grafts, aortic resection, replacement, or anastomosis, and reoperative cardiac surgery.
Subsequently, the researchers assessed the efficacy of their model by comparing it to the STS models using the same group of patients.
According to the study, the XGBoost model outperformed STS risk scores for mortality in all frequently performed categories of cardiac surgery for which STS scores were designed. The prediction performance of the XGBoost model across all surgical kinds was also strong, suggesting the potential of machine learning with EHR data for constructing successful institution-specific models.
Implications for Patient Care
Precise forecasting of postoperative mortality is vital for optimal outcomes in cardiac surgery patients. This research indicates that personalized models developed by individual institutions could outperform the conventional clinical approach relying on general population data. Additionally, the researchers demonstrated that healthcare providers can feasibly adopt robust machine learning algorithms to develop their own predictive models, either as substitutes for or enhancements to the standard STS template.
The Path to Future Innovations
The success of the machine learning model in mortality risk prediction for cardiac surgery patients highlights the extensive potential of artificial intelligence in healthcare. From preoperative planning to postoperative monitoring, machine learning algorithms can assist healthcare professionals in making informed decisions and delivering optimal care to patients.
The research conducted by Mount Sinai researchers unveils a promising future for the integration of machine learning in the field of cardiac surgery. By improving mortality risk prediction through the use of advanced algorithms, healthcare professionals can enhance patient outcomes and revolutionize the way cardiac surgeries are performed. As the technology evolves, it is crucial to ensure ethical considerations, data privacy, and ongoing validation to harness its full potential and maximize the benefits for patients worldwide.
Dr. Tamanna Anwar is a Scientist and Co-founder of the Centre of Bioinformatics Research and Technology (CBIRT). She is a passionate bioinformatics scientist and a visionary entrepreneur. Dr. Tamanna has worked as a Young Scientist at Jawaharlal Nehru University, New Delhi. She has also worked as a Postdoctoral Fellow at the University of Saskatchewan, Canada. She has several scientific research publications in high-impact research journals. Her latest endeavor is the development of a platform that acts as a one-stop solution for all bioinformatics related information as well as developing a bioinformatics news portal to report cutting-edge bioinformatics breakthroughs.