SP. Blank and RM. Blank (2025). Optimising risk stratification in diabetic ketoacidosis: a re-evaluation of acid-base status and hyperosmolarity using observational data. Internal and emergency medicine, 20(6), 1933-1940. https://doi.org/10.1007/s11739-024-03818-x

PubMed 39643802

[Abstract]
Accurate assessment of severity in diabetic ketoacidosis (DKA) can optimise early management and facilitate prioritisation for high acuity care. The primary aim was to evaluate the relationship between severity of acidosis (considering pH, bicarbonate, and anion gap) and hyperosmolarity with hospital mortality. Secondary outcomes included intensive care mortality, mechanical ventilation, vasopressor/inotrope use, and dialysis. A retrospective cohort study was conducted of adults (≥ 16 yr) with DKA admitted to US intensive care units. Data were extracted from the Medical Information Mart for Intensive Care (MIMIC)-IV dataset and eICU Collaborative Research Database. Univariable and multivariable logistic regression analyses were used to evaluate biochemistry obtained within 4 h of admission and the primary and secondary outcomes. We identified 4071 eligible admissions. There was no clear relationship between serum bicarbonate or anion gap and any outcome. Almost half the population did not have blood gas analysis within 4 h of admission; for 2292 patients with blood gases available, pH < 7 and inappropriately high PCO2 were associated with significant increases in mortality and all secondary outcomes. Osmolarity ≥ 320mosm/L was associated with fourfold increased mortality and higher rates of mechanical ventilation, use of vasopressors/inotropes, and dialysis. Failure of adaptive mechanisms (thirst and hyperventilation) indicating physiological decompensation may be more important for risk stratification in DKA than the degree of acidosis, which was only associated with outcome when severely abnormal. Blood gas analysis is essential to adequately assess disease severity as bicarbonate and anion gap were not predictive of outcome.