Objective: Unplanned readmission of hospitalized patients to an intensive care unit (ICU) is associated with a worse outcome, but our ability to identify who is likely to deteriorate after ICU dismissal is limited. The objective of this study is to develop and validate a numerical index, named the Stability and Workload Index for Transfer, to predict ICU readmission.
Design: In this prospective cohort study, risk factors for ICU readmission were identified from a broad range of patients' admission and discharge characteristics, specific ICU interventions, and in-patient workload measurements. The prediction score was validated in two independent ICUs.
Setting: One medical and one mixed medical-surgical ICU in two tertiary centers.
Patients: Consecutive patients requiring >24 hrs of ICU care.
Interventions: None.
Measurements: Unplanned ICU readmission or unexpected death following ICU dismissal.
Results: In a derivation cohort of 1,131 medical ICU patients, 100 patients had unplanned readmissions, and five died unexpectedly in the hospital following ICU discharge. Predictors of readmission/unexpected death identified in a logistic regression analysis were ICU admission source, ICU length of stay, and day of discharge neurologic (Glasgow Coma Scale) and respiratory (hypoxemia, hypercapnia, or nursing requirements for complex respiratory care) impairment. The Stability and Workload Index for Transfer score predicted readmission more precisely (area under the curve [AUC], 0.75; 95% confidence interval [CI], 0.70-0.80) than the day of discharge Acute Physiology and Chronic Health Evaluation III score (AUC, 0.62; 95% CI, 0.56-0.68). In the two validation cohorts, the Stability and Workload Index for Transfer score predicted readmission similarly in a North American medical ICU (AUC, 0.74; 95% CI, 0.67-0.80) and a European medical-surgical ICU (AUC, 0.70; 95% CI, 0.64-0.76), but was less well calibrated in the medical-surgical ICU.
Conclusion: The Stability and Workload Index for Transfer score is derived from information readily available at the time of ICU dismissal and acceptably predicts ICU readmission. It is not known if discharge decisions based on this prediction score will decrease the number of ICU readmissions and/or improve outcome.