Background: Diabetic ketoacidosis (DKA) is a serious, acute complication of diabetes mellitus caused by a state of insulin deficiency requiring exogenous insulin administration. Significant variation in intensive care unit (ICU) admission rates for DKA exists among hospitals, due to patient- and institution-specific factors, with rates ranging from 16-81%. Patients with DKA admitted to an ICU appear to be less critically ill, have lower illness severity scores, shorter ICU length of stay (LOS), shorter hospital LOS, and experience less mortality compared to ICU patients without DKA. Using our institutionally-derived, value-driven outcomes tool, we aimed to compare how the treatment location impacts direct hospital cost and clinical and safety outcomes for patients with severe and non-severe DKA.
Methods: This is a retrospective cohort analysis of 284 consecutive patients admitted with a diagnosis of DKA to the University of Utah Health Care (UUHC) system between October 1, 2014 and March 31, 2016. American Diabetes Association (ADA) diagnostic criteria were adapted to categorize patients as severe or non-severe DKA, with severe DKA defined as: pH <7.0, bicarbonate < 10 mEq/L, and/or Glasgow coma score (GCS) <14. All other patients were categorized as non-severe DKA. Using an electronic database query, demographic, clinical, and cost data were obtained to compare clinical and safety outcomes as well as direct hospital costs in patients with severe and non-severe DKA treated in the ICU versus an acute care floor.
Results: Of a total of 284 hospital admissions for DKA, 204 (72%) were categorized as non-severe DKA and 80 (28%) as severe DKA. Patients with severe DKA were younger, and more often admitted to an ICU or treated in an ICU during their hospitalization, in comparison to those with non-severe DKA. Among patients with non-severe DKA, the time to resolution of hyperglycemia was marginally shorter in patients initially admitted to an ICU, while normalization of the anion gap and time to initiation of basal insulin were longer when compared to floor admissions. Hospital LOS in non-severe DKA after controlling for other ICU needs, was non-significant. Treatment-induced hypokalemia, hypoglycemia or 30-day readmission rates did not differ between treatment location, for either severe or non-severe DKA. In patients with non-severe DKA and without other ICU needs, the total cost of care increased 88.5% (p=0.066) when care was administered in an ICU setting rather than the floor. Patients with severe DKA accrued similar total cost regardless of treatment location.
Conclusions: Treatment of non-severe DKA on an acute care floor compared to an ICU setting was associated with improved clinical outcomes, without increased adverse events. Significant reductions in cost were simultaneously noted. Patients with severe DKA demonstrated similar cost and clinical outcomes regardless of treatment location. We feel there is an opportunity to improve care and decrease cost for patients with non-severe DKA at our institution by preferentially admitting them to an acute care floor rather than an ICU, without compromising patient safety. We are in the process of utilizing this information to develop a standardized care pathway that builds on previously studied subcutaneous insulin protocols, to allow us to increase the proportion of patients with DKA admitted to an acute care floor.