Researchers analyzed data from over 15,000 patients who underwent surgery for gastric cancer. The study looked at different ways to predict a common complication called an anastomotic leak, which can occur after surgery. They compared traditional regression models against modern machine learning models.
The results showed that machine learning models had a higher accuracy score than standard regression models. Additionally, models that included information gathered after surgery performed better than those using only data from before or during the operation. These findings suggest that certain types of technology and timing of data can improve how doctors predict risks.
However, there are important reasons to be cautious with these results. The study was a meta-analysis involving many different reports, and some had high risks of bias. There were also concerns about how well the models were calibrated for real-world use. Because only a few studies compared specific methods, more large-scale testing is needed before these tools can change standard medical practice.
