Infectious Complications Following Radical Cystectomy: A Structural Equation Model
Objective: To use factor analysis to structure items from the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) into latent variables associated with infectious complications, and then to use structural equation modeling (SEM) to organize those latent variables into a predictive model of POIC. Predictive models of post-operative infectious complications (POIC) have traditionally relied upon logistic regression and inconsistent variable groupings. A more standardized approach to a valid construct would allow for more unitary research and improved clinical decision making. Materials and Methods: The study evaluated data from 1580 recipients of radical cystectomies in the ACS NSQIP PUF 2013 database. Pre-operative, operative, and post-operative data were analyzed. Exploratory Factor Analysis (EFA) and theory-based selection were used to create latent variables for a predictive model of POIC which was analyzed with structural equation modeling.Results: After reducing unrelated variables using EFA, two latent variables successfully predicted POIC, a Global Health Variable (Dyspnea, COPD, Diabetes, and Hypertension) and a Proximal Pre-operative Infectious Comorbidity Variable (pre-operative transfusion, pre-operative wound infection, and pre-operative sepsis). The final model produced was well-fit and suggest two unique pathway indicators for understanding which patients are at higher risk for POIC.Conclusion: Discerning the most significant items and their role in the POIC model offer clinical insight into adverse events and new considerations into the prevention of such events. Patients endorsing multiple items in the model may benefit from pre-operative optimization of modifiable conditions and closer post-operative surveillance.