Author Affiliations: Department of Epidemiology, School of Public Health, University of Alabama at Birmingham, Birmingham, Alabama.
In their study, Andres et al1 report on the development of a score designed to predict morbidity among patients undergoing liver resection. We have several concerns regarding the statistical analyses described in their study. The first relates to the use of a linear regression model to select variables associated with the occurrence of perioperative complications. Even though the study does not describe how a perioperative complication was operationalized, there are 2 possibilities: (1) it is a count variable (ie, number of complications) or (2) it is a binary variable (ie, 0 vs ≥1 complication). In either case, a linear regression model is not appropriate; rather, a Poisson/negative binomial regression model or a logistic regression model would have been the correct tool, respectively.
Tamhane A, McGwin G. Analytical Concerns Regarding Complications of Elective Liver Resections in a Center With Low Mortality. Arch Surg. 2011;146(12):1455. doi:10.1001/archsurg.146.12.1455