BACKGROUND: The National Healthcare Safety Network surgical site infections risk models for hip (HPRO) and knee (KPRO) replacement are intended for case-mix adjustment when reporting surgical site infection rates across institutions, but they are not validated in external data sets. OBJECTIVE: To evaluate the validity of HPRO and KPRO risk models and improvement in risk prediction with inclusion of information on morbid obesity and diabetes mellitus. DESIGN: Retrospective cohort study. PATIENTS: A single-center cohort of 21,941 hip and knee replacement procedures performed between 2002 and 2009. METHODS: Discriminative ability was assessed using the concordance statistic (C statistic). Calibration was assessed using the Hosmer-Lemeshow goodness-of-fit tests. RESULTS: The discrimination of HPRO was good, with a C statistic of 0.695 for surgical site infections and 0.749 for prosthetic joint infections. The discrimination of KPRO was worse than that of HPRO, with a C statistic of 0.592 for surgical site infections and 0.675 for prosthetic joint infections. Adding morbid obesity and diabetes mellitus to the HPRO and KPRO risk models modestly improved discrimination. There was no significant evidence of miscalibration based on the Hosmer-Lemeshow tests, but calibration of HPRO models appeared to be better than that of the KPRO models. CONCLUSION: HPRO performed better than the KPRO in predicting surgical site infections after hip and knee replacements. Both fared well in predicting prosthetic joint infections.
BACKGROUND: The National Healthcare Safety Network surgical site infections risk models for hip (HPRO) and knee (KPRO) replacement are intended for case-mix adjustment when reporting surgical site infection rates across institutions, but they are not validated in external data sets. OBJECTIVE: To evaluate the validity of HPRO and KPRO risk models and improvement in risk prediction with inclusion of information on morbid obesity and diabetes mellitus. DESIGN: Retrospective cohort study. PATIENTS: A single-center cohort of 21,941 hip and knee replacement procedures performed between 2002 and 2009. METHODS: Discriminative ability was assessed using the concordance statistic (C statistic). Calibration was assessed using the Hosmer-Lemeshow goodness-of-fit tests. RESULTS: The discrimination of HPRO was good, with a C statistic of 0.695 for surgical site infections and 0.749 for prosthetic joint infections. The discrimination of KPRO was worse than that of HPRO, with a C statistic of 0.592 for surgical site infections and 0.675 for prosthetic joint infections. Adding morbid obesity and diabetes mellitus to the HPRO and KPRO risk models modestly improved discrimination. There was no significant evidence of miscalibration based on the Hosmer-Lemeshow tests, but calibration of HPRO models appeared to be better than that of the KPRO models. CONCLUSION: HPRO performed better than the KPRO in predicting surgical site infections after hip and knee replacements. Both fared well in predicting prosthetic joint infections.