Development of the Sternal Wound Infection Prediction Scale

OBJECTIVE: To develop an instrument to reliably predict the occurrence of sternal wound infection (SWI). DESIGN: Instrument development with retrospective chart review. SETTING: Southwestern university-affiliated medical center. PATIENTS: one hundred eighty-five adults who underwent cardiac surgery. SWI developed in 94 patients, and 91 remained infection free. In phase I, there were 82 subjects: SWI developed in 41, and 41 remained infection free. In phase 2, there were 103 subjects: SWI developed in 53, and 50 remained infection free. OUTCOME MEASURE: Development of a postoperative SWI. INTERVENTION: The Sternal Wound Infection Prediction Scale (SWIPS) was developed in 2 phases. Phase 1 consisted of designing the SWIPS with use of 19 risk factors most often identified with SWI and revising this scale (SWIPS-R) with use of 21 risk factors. Eighty-two patient records were reviewed. After determining the mean cutscores in phase I, phase 2 provided cross-validation results with use of 103 additional patient records. RESULTS: The SWIPS produced 62.1% correct predictions of infection and noninfection; whereas the SWIPS-R produced 72.8% correct predictions. Phase 2 cross- validation results for the 21 risk factors demonstrated that 62.1% and 72.8% of the patients were correctly diagnosed using the SWIPS and SWIPS-R, respectively. A multivariate logistic regression was also performed on the risk factors to predict infection/noninfection groups for all 185 patients. A reduced logistic regression model included 9 of the risk factors and correctly classified 76.2% of the subjects. CONCLUSIONS: The SWIPS-R, with 21 risk factors, and the logistic regression model, with 12 risk factors, provided excellent classification rates of infection/noninfection. However, more data need to be collected to further strengthen reliability.