A study found that a machine learning model, trained and evaluated with data from 1,477,561 patients, effectively identified individuals at high risk of mortality within 30 days after surgery. The model outperformed the current popular pre-surgical risk calculator tool. The study reported the 30-day mortality, and major adverse cardiac or cerebrovascular events (MACCEs). The Area Under the Receiver Operating Characteristic curve (AUROC) scores were extremely high for the training and test sets. In comparison to the National Surgical Quality Improvement Program (NSQIP) surgical risk calculator tool, the new machine learning model exhibited higher AUROC scores, which indicates superior performance. The model offered improved risk interpretation by identifying key features contributing to the prediction of adverse events.
Age and recent albumin levels were found to be significant factors. The study highlighted the scarcity of predictive tools to identify high-risk patients, emphasizing the model’s accuracy and the advantages it offers over existing risk prediction tools. The model’s robustness was attributed to its utilization of a large and diverse patient population, incorporating various social determinants of health, too. The application of the model in clinical practice provided automated and up-to-date risk predictions, reducing the need for manual data extraction.
The study’s findings were complemented by the endorsement of other experts in the field who recognized the model’s clinical utility. The study employed a substantial amount of data from electronic health records (EHRs) and selected patients from the University of Pittsburgh Medical Center (UPMC) health system. However, limitations of the study included its reliance on existing EHR data and the lack of validation using test sets from other institutions.
The whytry.ai article you just read is a brief synopsis; the original article can be found here: Read the Full Article…
