Choosing a statistical model for analysis of perioperative data: A balance between statistical rigor and usability for surgeons

摘要

For a better understanding of medical statistics and delivering a high-quality report endorsed by an adequate statistical method, academic surgeons would be encouraged to read an article by Dr Fabri in the recent issue of the Journal of the American College of Surgeons. In the article, the pitfalls implementing and interpreting a logistic regression model for big data, such as those from American College of Surgeons (ACS) NSQIP, are mentioned repeatedly: "logistic regression by itself does nothing to make the inputs actually independent of each other" and "including statistically significant variables does not mean that they are important."