Objective

This retrospective cohort study compares use of paper-based insulin protocols to a computer-based insulin protocol (EndoTool) for intensive care unit patients.

Setting

576-bed tertiary care teaching hospital and Level 1 trauma center.

Results

• Among non-cardiothoracic surgery patients in a surgical ICU, mean glucose was lower with EndoTool (127.2 vs 142.9 mg/dL), incidence of hyperglycemia was lower (8.5% vs 28.5% of patients with 10% or more of blood glucose measurements >=200 mg/dL) and incidence of severe hypoglycemia was lower (0.00% vs 0.34% of blood glucose measurements <40 mg/dL per 100 person days).
• Among cardiothoracic surgery patients in a surgical ICU, mean glucose was lower with EndoTool (130.9 vs 138.8 mg/dL), incidence of hyperglycemia was lower (6.2% vs 15.5% of patients with 10% or more of blood glucose measurements >=200 mg/dL), and incidence of severe hypoglycemia was the same (0.00% of blood glucose measurements <40 mg/dL per 100 person days).
• Among all other ICU patients, including medical, trauma and neuroscience, mean glucose was lower with EndoTool (141.5 vs 159.9 mg/dL), incidence of hyperglycemia was lower (35.2% vs 64.1% of patients with 10% or more of blood glucose measurements >=200 mg/dL), and incidence of severe hypoglycemia was lower (0.00% vs 0.34% of blood glucose measurements <40 mg/dL per 100 person days).

Conclusions

Use of a computer-based insulin protocol (EndoTool) for intensive care unit patients is associated with lower mean glucose and much higher performance at minimizing hyperglycemia than paper-based protocols; and for the majority of patients, it is associated with significantly decreased glucose variability and lower incidence of severe hypoglycemia.