The epidemiology of readmissions to the intensive care unit and its use as a quality indicator

Background: Over recent years, there has been increased attention to improving quality of care in Intensive Care Units (ICUs). Various organizations have promoted ICU readmission as a quality indicator because of its association with increased mortality, length of stay, and costs, though there is little evidence that ICU readmissions are causally related to these downstream effects. The overall goal of this dissertation was to determine whether ICU readmissions are good quality indicators for ICUs. Methods: We used the Project IMPACT database, which contains detailed information on over 500,000 ICU admissions in a sample of representative U.S. ICUs to examine three questions: (1) What is the incidence of ICU readmission in the U.S.; (2) what time interval between ICU discharge and readmission best captures readmissions attributable to ICU activities; and (3) do ICU readmissions or other potential quality indicators possess characteristics associated with being good quality indicators. We used a retrospective cohort design for each aim. Results: Median 48 hour ICU readmission rates were 1.8% (IQR: 1.1%, 2.4%, max 5.4%). The median time to readmission was 3 days. The effect of many exposures depended on the duration used to define ICU readmission. Variables reflecting admission severity of illness became more influential as time to readmission increased; variables reflecting ICU discharge circumstances became less influential as time to readmission increased; both had inflection points between one and three days post-discharge. The effect of night discharge varied diurnally with increasing durations of time to readmission. ICU readmissions and ICU process measures were primarily determined by ICU characteristics and were insensitive to severity adjustment. However ICU readmissions were rare, had less variability, and may induce perverse incentives. Conclusions: ICU Readmissions are rare events, occurring for approximately 2% of patients discharged from the ICU. The time interval that best captures readmissions due to ICU activities, and is least biased by daily patterns of bed availability is two calendar days. ICU readmissions have some characteristics associated with good quality indicators; however other measures, like ICU process measures may better capture ICU quality without misattribution of blame