SAT-601 Evaluation Of Diabetic Ketoacidosis Outcomes Following The Adoption Of A Computer-based Variable Rate Insulin Infusion Algorithm

Abstract Disclosure: O. Ibrahim: None. M. Wei: None. J.M. Deirmenjian: None. T. Madu: None. G. Simone: None. E.V. Nguyen: None. S. Kim: None. C. Escovedo: None. H. Tran: None. S. Masih: None. N. Ballweg: None. E. Zakis: None. R.Y. Gianchandani: None. Current recommendations suggest a low dose, fixed rate insulin infusion for adult DKA management. Newer guidelines include variable rate infusions which are increasingly being used with computerized algorithms. In this retrospective review, we evaluate Glucommander (GlytecTM), a variable rate computerized insulin infusion algorithm for DKA management within the Cedars-Sinai Health Systems. Methods: Glucommander (GM) was implemented in January 2024 at a tertiary academic center (AC) and a community hospital (CH). Here we compare (PostGM) outcomes to previous standard of care, (PreGM) which included separate paper/provider-adjusted algorithms. Data from electronic medical record of patients age 18 years, between January 2023 - December 2024, with primary admission diagnosis of DKA , treated in the ICU with insulin infusion, and meeting DKA criteria (initial pH < 7.30, bicarbonate (HCO3) < 18 mEq/L, anion gap including potassium (AG) > 14 and beta hydroxybutyrate (BOHB) > 2) was collected. Transition data between January - March 2024 were excluded, as were patients transferred to another hospital, or if drip was continued for non-DKA indication. Primary safety outcomes were the incidences of hypoglycemia (blood glucose, BG <70 mg/dL) and of severe hypoglycemia (BG <40 mg/dL) while receiving insulin infusion. Primary treatment outcomes were insulin drip duration, time to resolution of acidosis (HCO3 >18mEq/L), and ketosis (BOHB <0.6 mmol/L. Secondary outcomes were ICU and hospital LOS. Results: A total of 214 patients were identified PreGM (77 AH and 137 CH) and 324 PostGM (157 AH and 167 CH). There was a decrease in the incidence of hypoglycemia (%) (AH: PreGM 12.7% vs PostGM 10.4%; CH:27.5% vs 5.8%) and severe hypoglycemia (AH: 3.2% vs 0%; CH: 2.4% vs 0%). At the AH, insulin drip duration was mostly unchanged (PreGM 22 vs Post GM 23 hours) while time to resolution of acidosis (22.3 vs 18.3 h), ketosis (42.3 vs 16.2 h), and AG closure (14.1 vs 13.3 h) all decreased PostGM. The ICU LOS was reduced (75.2 vs 63.0 h) as was hospital LOS (8.6 vs 6.0 days). At the CH, insulin drip duration (PreGM 26 vs PostGM 36 h), time to resolution of acidosis (15.4 vs 17.2 h) as well as ketosis (26.5 vs 30.1 h), and AG closure (18.92 vs 3.0 h) all increased, as did ICU LOS (53.0 vs 60.6 h) and hospital LOS (6.8 vs 8.1 days). Discussion: Treatment of DKA using the computerized insulin algorithm was safely achieved with a variable rate computerized insulin infusion (Glucommander). In the CH, time to DKA resolution and LOS went up, while rates of hypoglycemia decreased. This increase in resolution times possibly reflected an aggressive paper protocol to revamp DKA care for PreGM patients. In the teaching AC time to DKA resolution and LOS were both reduced, likely representing standardization of care. Presentation: Saturday, July 12, 2025