Infection Control and Hospital Epidemiology

In this issue of Infection Control and Hospital Epidemiology, three articles report various aspects of the costs associated with nosocomial infections. Hollenbeak et al. use four different statistical methods to determine the impact of nonrandom selection on the attributable cost of surgical-site infection (SSI) following coronary artery bypass graft (CABG) surgery. In this study, they analyzed data from 41 cases of deep chest infections following CABG surgery at a Midwestern community medical center and compared them with data from 160 randomly selected uninfected controls. Using the hospital's cost accounting system, they compared the costs of infected and uninfected patients by a t test in a matched and unmatched comparison, by regression analysis, and using Heckman's twostage method. Unmatched comparison demonstrated $20,012 in excess cost for the infected patients and after 1:1 matching on age, gender, diabetes, renal insufficiency, and length of surgical procedure, the excess cost for the infected patients was $19,579, which was $433 lower than the unmatched estimate. Forward stepwise logistic regression was then used to estimate the attributable cost of SSI by controlling for other patientand process-related risk factors that may affect the cost of care, including severity of illness, duration of surgery, gender, obesity, congestive heart failure, diabetes, reexploration for bleeding, and use of an intra-aortic balloon pump. Use of an intra-aortic balloon pump was the only variable besides SSI that significantly affected the cost of CABG surgery. After controlling for the use of an intra-aortic balloon pump, the authors found that deep chest SSI increased cost by $19,311. The authors then used Heckman's two-stage procedure to control for the impact of nonrandom selection, which may affect the cost of SSIs. For example, many of the factors that increase the risk of SSIs may also independently increase the cost of care, such as diabetes, heart failure, severity of illness, renal failure, and duration of surgery. The two-stage method corrects for the possibility of nonrandom selection when assessing cost impact. Using this method, the authors found that the economic impact of deep chest SSI was $14,211, approximately $6,000 (30%) lower than the cost estimates defined by the other methods. However, the coefficient on the hazard function was not statistically significant, implying that the cost estimates from the three other methods did not suffer from selection bias. Thus, the cost estimates produced by unmatched comparison, matched comparison, and linear regression are reasonable to use.