Abstract Title: Performance of Two Surgical Site Infection Risk Stratification Models for Predicting Infection Risk in Publicly Reported Data from a Safety Net Hospital
Introduction: Since 2007, the Colorado Hospital-Acquired Infections Disclosure Act has required hospitals and ambulatory surgery centers to report selected surgical site infections (SSI) to the CDC's National Healthcare Surveillance Network (NHSN) to be analyzed and risk adjusted. The Colorado Department of Public Health and Environment uses that information to issue comparisons of facility-specific, risk-stratified infection rates to the public.
Background: In order to provide reliable information to the public, risk adjustment models must be reliable for accurately comparing a facility's performance against that of other procedures with similar risk. However, risk adjustment models currently used for public reporting have not been widely validated. We sought to verify the NHSN basic risk index and standardized infection ratio (SIR) for their abilities to predict SSI in publicly reported herniorrhaphy, hysterectomy, and arthroplasty procedures from an academic public health safety net institution in Denver, CO.
Methods: Cases of arthroplasty (hip and knee) between January 2007 and October 2010; herniorrhaphy and hysterectomy (vaginal and abdominal) between January 2009 and October 2010 were entered into NHSN as mandated by State law. These were retrieved from the NHSN system database and duplicate patients were excluded. Cases were defined as SSI according to the NHSN definitions. The NHSN basic risk index and SIR was calculated for each procedure based on the following variables: wound class, age, anesthesia, ASA, duration of surgery, arthroplasty type, hospital beds, trauma, endoscope, and gender. Association between NHSN basic risk index and SSI were identified using the Cochran-Mantel-Haenszel test for categorical variables. Association between NHSN SIR and SSI were identified using the Wilcoxon two-sample test for continuous variables.
Results: A total of 1337 publicly reported cases were included (488 arthroplasty; 616 herniorrhaphy; 233 hysterectomy), of which 31 were defined as SSI (13 arthroplasty; 8 herniorrhaphy; 10 hysterectomy). The NHSN basic risk index did not correlate with SSI when assessed in aggregate (p=0.19), after arthroplasty (p=0.66), or after hysterectomy (p=0.74); only SSI after herniorrhaphy showed a correlation with the basic risk index (p=0.02). In contrast, the SIR correlated with SSI when assessed in aggregate (p<0.001) and after herniorrhaphy (p=0.001) and hysterectomy (p=0.03); SIR did not correlate with SSI after arthroplasty (p=0.09).
Conclusion: Neither basic risk index nor SIR currently used for risk stratification in publicly reported procedures in CO adequately predicted SSI at this institution. SIR predicted infection risk better than the basic risk index for the procedures that we examined. The SIR uses logistic regression modeling and takes into account more variables and procedure-specific risk factors. We hypothesize that improved risk adjustment is due to consideration of these extra risk factors. Given its inferior risk stratification performance, the basic risk index should no longer be used when analyzing SSI data for public reporting. Although superior, the SIR still is not broadly applicable to all procedures and settings, and further refinement is needed before it can be promoted as a tool to compare risk stratified rates of SSI from arthroplasty procedures at safety net institutions.