Background: Length of stay (LOS) is an important outcome measure in the evaluation of the quality of care for hospitalized patients. LOS in adults with diabetic ketoacidosis (DKA) is reported to be about 3 days, but detailed information about LOS in pediatric patients with DKA is lacking. Knowledge of factors that impact LOS is essential for developing strategies to mitigate them. We sought to determine factors associated with prolonged LOS in pediatric patients with DKA. We hypothesized that there will be patient-related, care-related, and laboratory-related factors that impact LOS.

Methods: This retrospective study included children 24 months through 17 years with DKA admitted through our freestanding children’s hospital emergency department (ED). Patients were identified in our electronic health record using ICD-9 & 10 codes for DKA. Data collected included demographics, initial values of serum glucose, bicarbonate (HCO3), pH, anion gap, known or new diagnosis of diabetes mellitus, previous history of DKA, receipt of any IV insulin prior to arrival to our ED, and inpatient admission unit. Date and time were recorded for: ED arrival, hospital admission, hospital discharge, normalization of HCO3 (>15 mmol/L), closure of anion gap (≤ 15 mmol/L), start of insulin drip, first dose of glargine, and start of IV fluids. Inpatient LOS was calculated from hospital admission to hospital discharge. Prolonged inpatient LOS was defined as 2.37 days, which is greater than the 75th percentile LOS for our sample. We performed multivariate binary logistic regression to determine factors associated with prolonged length of inpatient stay and report odds ratio (OR) and 95% confidence interval (CI).

Results: Of the 500 cases of DKA reviewed, 125 had prolonged LOS. In the regression model, black race (OR 1.92, 95% CI 1.10-3.45), new diagnosis of diabetes mellitus (DM) (OR 25.18, 95% CI 8.86-71.61), prior receipt of IV insulin (OR -3.68, 95% CI 1.79-7.56), previous history of DKA (OR 2.95, 95% CI 1.12-7.78), and time to normalization of HCO3 (OR 1.07, 95% CI 1.01-1.13) were statistically significantly associated with prolonged length of inpatient stay (Figure). Notably, age, day of admission, admission to pediatric intensive care unit, time to normalization of anion gap, time to start insulin infusion, time to first dose of glargine, and initial values of HCO3, glucose, and anion gap were not associated with prolonged length of inpatient stay.

Conclusions: Patient factors, treatment, and evolution of laboratory values impact length of inpatient stay in pediatric patients with DKA. Though many factors were identified, new diagnosis of DM was the most strongly associated with prolonged LOS. Further investigation into the reasons for this reveals clear areas for potential intervention. Understanding the factors that impact LOS allows for development of targeted interventions and systems changes to improve hospital care.

IMAGE 1: Adjusted Odds Ratio and 95% CI of Factors Associated with Prolonged Length of Inpatient Stay