Bouncing Back With SWIFT (Stability and Workload Index for Transfer Score): Is It Applicable to ICUs in Urban America?

Abstract

PURPOSE: Over the past few decades, the number of patients surviving critical illness has improved. However, ICU readmission rates have remained unchanged. The armamentarium has been relatively empty regarding discharge criteria from the ICU for physicians. In 2008, investigators from the Mayo Clinic developed and validated a new protocol scoring system -The Stability and Workload Index for Transfer (SWIFT) score. A standardized prediction tool is an ideal premise to optimize patient outcomes. The SWIFT scoring system has value in that it is the first predictive tool for intensive care readmission. Is this scoring system to applicable to their ICU? The purpose of this study is to assess the validity of the SWIFT Scoring System at an ICU in an urban public hospital. METHODS: This is a retrospective observational cohort study comprising the medical intensive care units at Parkland Hospital. The cohort for this study consists of consecutive patients discharged alive from the medical ICUs at Parkland Hospital from June 1, 2010 to May 31, 2011. The primary outcome variables paralleled the SWIFT study, measuring unplanned ICU readmission or unexpected death within 7 days of ICU discharge. The performance of the SWIFT Score was assessed amongst our cohort for its accuracy in predicting ICU readmissions. RESULTS: Our cohort included 2,054 patients admitted to the medical ICU at Parkland Hospital over 1 year. Patients were excluded for ICU stay <24hrs or planned admissions (499), those discharged home or transferred to another hospital/ICU within 7 days (358), those discharged to comfort care (14) and those who died during their ICU stay (141). 1,042 patients were discharged alive and evaluated as our baseline patient population. Fifty-two patients (5%) were readmitted to an ICU at our institution and six (0.5%) patients died unexpectedly within 7 days; the combined readmission and unexpected death rate was 5.5%. The sensitivity of the SWIFT score for predicting ICU readmissions was 0.21 and specificity was 0.83. This compares with original data showing sensitivity of 0.56 and specificity of 0.83. CONCLUSIONS: While the specificity of the SWIFT score remained robust in our study, the sensitivity was lacking at predicting ICU readmission.