Budget Impact Analysis of Rapid Screening for Staphylococcus aureus Colonization Among Patients Undergoing Elective Surgery in US Hospitals

Publisher Website
Google Scholar
Abstract
Objective. To evaluate the economic impact of performing rapid testing for Staphylococcus aureus colonization before admission for all inpatients who are scheduled to undergo elective surgery and providing subsequent decolonization therapy for those patients found to be colonized with S. aureus. Methods. A budget impact model that used probabilistic sensitivity analysis to account for the uncertainties in the input variables was developed. Primary input variables included the marginal effect of S. aureus infection on patient outcomes among patients who underwent elective surgery, patient demographic characteristics, the prevalence of nasal carriage of S. aureus, the sensitivity and specificity of the rapid diagnostic test for S. aureus colonization, the efficacy of decolonization therapy for nasal carriage of S. aureus, and cost data. Data sources for the input variables included the 2003 Nationwide Inpatient Sample data and the published literature. Results. In 2003, there were an estimated 7,181,484 patients admitted to US hospitals for elective surgery. Our analysis indicated preadmission testing and subsequent decolonization therapy for patients colonized with S. aureus would have produced a mean annual cost savings to US hospitals of $231,538,400 (95% confidence interval [CI], -$300 million to $1.3 billion). The mean annual number of hospital-days that could have been eliminated was estimated at 364,919 days (95% CI, 67,893-926,983 days), and a mean of 935 in-hospital deaths (95% CI, 88-3,691) could have been avoided per year. Sensitivity analysis indicated a 64.5% probability that there would be cost savings to US hospitals as a result of preadmission testing and subsequent decolonization therapy. Conclusion. The addition of preadmission testing and decolonization therapy to standard care would result in significant cost savings, even after accounting for variations in the model input values.