Differences between single and multiple DKA episodes
Lifetime national DKA admission data were available for all patients diagnosed with diabetes from 1981 onwards (271 patients from the complete cohort of 298). These patients were stratified into three groups based on the total number of DKA admissions: single admission (n = 96), two to five admissions (n = 111) and more than five admissions (n = 64). Overall, 55% (n = 164) of individuals presenting with DKA were men and no difference in the sexes was noted between the three categories (26.0% of women had more than five DKA presentations compared with 21.7% of men, p = 0.186). Compared with patients with a single admission, those with multiple presentations had a longer duration of diabetes (median 12.8 [IQR 10.0–17.5] vs 7.6 [2.3–13.6] years, p < 0.001) and were diagnosed with diabetes at a younger age (14 [9–23] vs 24 [16–34] years, p < 0.001). Multiple DKA admissions were more common in patients with higher levels of social deprivation (SIMD rank 1825 [813–3346] vs 2723 [1559–4310], p = 0.005) and higher HbA1c levels (103 [89–108] vs 79 [66–96] mmol/mol; 11.6% [10.3–12.0%] vs 9.4% [8.2–10.9%], p < 0.001). Comparisons of other clinical variables can be seen in electronic supplementary material (ESM) Table 1.
Overall, 13.1% (8/61) of patients with more than five DKA admissions had experienced an inpatient admission for psychiatric care, compared with 5.6% (6/108) and 4.3% (4/92) of those with two to five and a single DKA admission, respectively (p = 0.092). Furthermore, 47.5% (29/61) of those with more than five DKA admissions had received antidepressants, compared with 27.8% in the intermediate (30/108) and 12.6% (12/95) in the single-attender groups (p = 0.001). No significant relationships were observed with DKA admission frequency and the use of other classes of medication (ESM Table 2).
Inpatient DKA mortality rate
Over 6 years, 628 DKA admissions were identified, involving 298 individuals with type 1 diabetes. The median age at presentation was 28 years (IQR 22–40 years). No deaths were identified during inpatient admissions for the management of DKA. To ensure our data capture was complete, every death of a patient with type 1 diabetes registered to our outpatient clinic (153 deaths) was reviewed across the study time period using our comprehensive diabetes register. This revealed a single inpatient death where DKA was reported as a contributory factor; the inpatient DKA mortality rate was therefore, at most, 0.16%.
Subsequent mortality following hospital discharge (total cohort)
In patients with a prior DKA presentation, 44 deaths (14.8%) were observed during follow-up (median 4.9 [IQR 3.3–6.7] years) in 298 individuals. The median age of death was 45.9 years (IQR 30.8–58.0 years). Mortality rates were significantly associated with the number of DKA presentations during the 6-year study period, and were highest in those with more than four admissions (29.6% [8/27]), intermediate in those with two to four admissions (18.3% [15/82]) and lowest in those with a single admission (10.6% [20/189]) (p = 0.016).
A total of 19 of these 44 deaths were of uncertain cause, and were often unexpected and potentially attributable to acute metabolic decompensation. The median age of these patients was 31 years (range 20–63 years) (full cause-of-death data are presented in ESM Fig. 1). Overall, 52.3% (23/44) deaths occurred at home at a median age of 38 years (IQR 27.7–52.3 years), compared with a median age of 57.7 years (IQR 40.5–61.4 years) in those dying in hospital (p = 0.01). Ten deaths (eight at home) occurred within 2 months of the final DKA admission; all of these deaths were unanticipated.
A range of clinical factors were associated with increased mortality risk, including older age, longer diabetes duration, prior DKA requiring intensive care admission, psychological issues, neuropathy, previous cardiovascular disease, excess alcohol intake and a longer length of stay during the last hospital admission (Tables 1, 2). Prescription of an antidepressant, at any time since 2009 (when national data collection began, n = 264), was associated with a trend towards increased likelihood of death (HR 2.24, 95% CI 0.99, 5.12, p = 0.055). Social deprivation was not associated with mortality risk in this cohort (Table 1).
Influence of lifetime DKA admissions
Overall, 23.4% (15/64) of those with more than five lifetime DKA admissions died over a median 2.4 year (IQR 2.0–3.8 years) follow-up period (HR 6.18 [95% CI 2.1, 18.3], p = 0.001; Fig. 1). Death occurred in 13.5% (15/111) of those with two to five admissions (HR 3.02 [95% CI 1.1, 8.4; reference is single DKA], p = 0.035) over a median of 3.7 years (IQR 2.4–5.5 years). A single lifetime DKA admission was associated with a 5.2% (5/96) risk of death during the follow-up period (median 4.1 [IQR 2.8–6.0] years). Median age at death was significantly lower in those with more than five DKA admissions (32 [IQR 23–39] years) than in those with two to five admissions (median 53 [IQR 40–58] years) or a single admission (median 53 [IQR 38–66] years) (p = 0.014). The prevalence of cardiovascular disease was not independently associated with mortality. Multivariate analysis identified greater number of DKA admissions, longer diabetes duration, previous psychiatric admissions and older age at diagnosis as independent predictors of death (Table 3).