Glucose Management in the ICU: The Bi-Centre GEM-ICU Cohort Study

Dysglycaemia is common in intensive care unit (ICU) patients and has been associated with worse outcomes. Data on glucose management and insulin use from contemporary intensive care units (ICUs) are limited.

We conducted a bi-centre, retrospective cohort study of acutely admitted, adult patients at two Danish ICUs. Patients were included consecutively by backtracking admissions from December 1, 2023, until the target sample size (n = 300) was reached. Data were collected from electronic health records following a predefined protocol. The primary outcome was the occurrence of hypoglycaemia (< 4 mmol/L) during ICU stay. Secondary outcomes included severe hypoglycaemia (< 2.3 mmol/L), hyperglycaemia (> 10 mmol/L), time below blood glucose target range (< 6 mmol/L), time above target range (> 10 mmol/L), mortality at 30 days, days alive without life support at 30 days (DAWOLS) and days alive out of hospital at 30 days (DAOH). Process variables were also assessed.

We screened 311 ICU patients and included 300, of whom 285 had at least one blood glucose measurement. Hypoglycaemia was observed in 21 patients (7.4%; 95% CI 4.6-11.0) and severe hypoglycaemia in 3 patients (1.1%; 95% CI 0.2-3.0). Hyperglycaemia was observed in 186 patients (65.3%; 95% CI 59.4-70.8). Mortality at 30 days was 24.3% (95% CI 19.6-29.6). Median DAWOLS at 30 days was 27 days (95% CI 26-28; IQR 8-30) and median DAOH at 30 days was 9 days (95% CI 3-14; IQR 0-21).

In this Danish bi-centre cohort of acutely admitted ICU patients, hypoglycaemia was uncommon and severe hypoglycaemia was rare, whereas hyperglycaemia occurred in almost every second patient. Insulin was frequently administered, mostly intravenously, and glucose was monitored often. Patients receiving insulin spent nearly half of the observed time above the target range.

In this retrospective analysis from 2 Danish centres, glucose management and insulin use in the ICU is presented. Rare hypoglycemia and more common hyperglycemia measurement events were counted, along with insulin administration, which was common.