Investigators reviewed adult admissions for diabetic ketoacidosis at a Norwegian university hospital across a decade. They included 178 unique patients and 306 admissions after detailed chart review. Mean monthly admissions fell from 3.20 in 2012 to 1.875 in 2022. Poisson regression estimated an annual decline of 6.9% (rate ratio 0.931; 95% CI 0.896–0.968; p < 0.001), which corresponds to an approximate 51% reduction in modelled admission counts over ten years. The study team observed this decline despite growth in the hospital catchment population, suggesting changes in clinical care or population-level factors during the study interval.
Median age at admission was 34 years (IQR 22–51). The cohort had poor glycaemic control at presentation: median HbA1c was 88 mmol/mol (IQR 72–105). Only 10 admissions (3.3%) had an HbA1c ≤ 53 mmol/mol. Investigators classified 58% of ketoacidoses as severe. Using a 300 mOsm/kg threshold, concomitant hyperosmolality was present in 44% of admissions. Ketone measurements were not available for 111 admissions; five additional admissions had ketone values below the diagnostic threshold but were retained when clinical assessment and response to therapy supported a DKA diagnosis. The most common precipitant documented was forgotten or intentionally omitted insulin doses; in admissions related to missed or omitted insulin, clinicians noted a psychiatric illness history in 40% of those cases. Insulin pump users accounted for 83 admissions (30.1% of pre-existing type 1 diabetes admissions), and 24 admissions were attributed to pump failure, including three on automated insulin delivery devices.
Emergency department recognition and prompt initiation of guideline-based therapy were common. Isotonic fluids were started in 98% of admissions and insulin in 96% according to local guidance. Hospital protocol required potassium infusion prior to insulin when serum potassium was ≤ 5 mmol/L; however, potassium infusion was documented in only 55% of eligible admissions. The study team used time from first arterial blood gas to a bicarbonate ≥ 18 mmol/L to define biochemical normalisation. They found that each successive calendar year was associated with a 2.8% shorter time to biochemical normalisation (p = 0.025), amounting to a 24% cumulative relative decrease over the decade. Length of stay also fell over time, decreasing by an estimated 3.7% per calendar year. New-onset diabetes and macrovascular disease were linked with substantially longer admissions (151% and 87% longer stays, respectively, in adjusted analyses reported in supplemental tables).
Repeated admissions were frequent and concentrated. Twenty-two percent of participants experienced more than one admission during the study period; that subgroup accounted for 55% of all admissions. A very small group had extreme recurrence: five individuals (2.8%) had six or more episodes and one individual had 24 admissions. Among participants with previously known diabetes who survived their index admission (n = 143), 36 had at least one readmission in the study period, together accounting for 122 readmissions. Multivariable analyses identified psychiatric illness (adjusted OR 2.827; 95% CI 1.070–7.468; p = 0.036) and receipt of disability benefits (adjusted OR 4.941; 95% CI 1.576–15.490; p = 0.006) as predictors of having more than one documented lifetime DKA admission. Higher age at index admission was associated with lower odds of multiple lifetime admissions (aOR 0.727 per 10 years). In recurrence-specific models, documented prior DKA increased hazard of first readmission (HR 3.058; 95% CI 1.469–6.365), while suspected treatment non-adherence (rate ratio 4.402) and psychiatric illness (rate ratio 2.643) were associated with higher rates of recurrent admissions across the follow-up window. The authors reported no clear association between index HbA1c and recurrence in their data.
Short-term mortality during admissions was low: five admissions ended in death, equivalent to an in-hospital mortality of 1.6%. The study used the first observed admission to estimate longer-term survival and reported a five-year cumulative mortality of 14.7% from that index admission. In Cox regression analyses through 26 November 2024, two variables were associated with increased all-cause mortality: alcohol or substance misuse (hazard ratio 2.499; 95% CI 1.191–5.245; p = 0.015) and higher age at index admission (hazard ratio 1.874 per 10 years; 95% CI 1.511–2.324; p < 0.001). The electronic records provided date but not cause of death. The authors also modelled recurrence as a time-varying covariate in additional analyses (details and diagnostics are reported in the supplementary appendix).
The study team reported several important limitations. It was a single-centre, retrospective cohort reliant on electronic medical record documentation; such designs are vulnerable to misclassification and unmeasured confounding. Ketone values were missing in 111 admissions, and five admissions with ketone values below diagnostic thresholds were retained based on clinical adjudication—criteria the authors explicitly described. C-peptide and antibody testing were not performed as part of the study protocol, so diabetes classification relied on the most recent record. Data linkage provided dates but not causes of death. Individual-level data cannot be made publicly available under applicable data protection requirements and the approved study arrangements; access inquiries require ethical and institutional review as detailed by the study authors. These limitations frame the findings: the declining admission trend and improved in-hospital course are robust within the reviewed records, but generalisability beyond this setting and causal interpretation of associations with psychosocial factors are constrained. The article ends with the specified need for targeted follow-up after DKA episodes and the data-access and ethical constraints noted in the source.