Abstract
OBJECTIVES: Although children who survive sepsis are at risk for readmission, identification of patient-level variables associated with readmission has been limited by administrative datasets. We determined frequency and cause of readmission within 90 days of discharge and identified patient-level variables associated with readmission using a large, electronic health record-based registry. METHODS: This retrospective observational study included 3464 patients treated for sepsis or septic shock between January 2011 and December 2018 who survived to discharge at a single academic children's hospital. We determined frequency and cause of readmission through 90 days post-discharge and identified patient-level variables associated with readmission. Readmission was defined as inpatient treatment within 90 days post-discharge from a prior sepsis hospitalization. Outcomes were frequency of and reasons for 7-, 30-, and 90-day (primary) readmission. Patient variables were tested for independent associations with readmission using multivariable logistic regression. RESULTS: Following index sepsis hospitalization, frequency of readmission at 7, 30, and 90 days was 7% (95% confidence interval 6%-8%), 20% (18%-21%), and 33% (31%-34%). Variables independently associated with 90-day readmission were age ≤ 1 year, chronic comorbid conditions, lower hemoglobin and higher blood urea nitrogen at sepsis recognition, and persistently low white blood cell count ≤ 2 thous/µL. These variables explained only a small proportion of overall risk (pseudo-R2 range 0.05-0.13) and had moderate predictive validity (area under the receiver operating curve range 0.67-0.72) for readmission. CONCLUSIONS: Children who survive sepsis were frequently readmitted, most often for infections. Risk for readmission was only partly indicated by patient-level variables.