Unscheduled readmissions to the PICU: epidemiology, risk factors, and variation among centers.

OBJECTIVES To determine the rate of, and potential risk factors for, unscheduled PICU readmission and assess for variability among PICUs within the United States. DESIGN AND DATA SOURCE This retrospective cohort study used 2005-2008 data from 73 PICUs in the Virtual PICU Systems database. METHODS AND MEASUREMENTS Early (within 48 hr of PICU discharge) and late (later than 48 hr) unscheduled readmission rates were calculated. Hierarchical logistic regression, with a random intercept for site, was used to identify factors independently associated with early readmission. Significant random effects identified sites with an outlying risk of readmission, adjusting for patient and admission characteristics. MAIN RESULTS For 117,923 children meeting inclusion criteria, the unscheduled readmission rate was 3.7% with 38% (1.4%) occurring early. Half of early readmissions had the same primary diagnosis as the first admission. Patients with late readmissions had a higher mortality (6.6% vs 3.3%, p < 0.001) and longer median total PICU length of stay (11 d vs 6 d, p < 0.0001) than those with early readmission. Patient characteristics strongly associated with increased risk of early readmission included the following: age < 6 months, acute respiratory and renal disease, and several underlying chronic conditions such as liver disease, bone marrow transplant, airway stenosis, and abnormal antidiuretic hormone balances. An initial PICU admission that was unscheduled, originated from the general floor, or with a discharge time between 4 PM and 8 AM was associated with higher risk of readmission. A quarter of sites were identified as potential high (16%) or low (8%) outliers. CONCLUSIONS The rate of unscheduled PICU readmission was low but associated with worse outcomes. Patient and admission/discharge characteristics associated with increased risk of readmissions could be used to target high-risk populations or modifiable factors to improve outcome. Variation of risk among centers suggests room for improvement.