Impact of a computer-generated alert system on the quality of tight glycemic control

Abstract

To assess the impact of a computer-generated blood glucose (BG) alert, generated by a Patient Data Management System (PDMS) and superimposed on a paper-based guideline, on tight glycemic control (TGC) in the intensive care unit (ICU). TGC in the Leuven University Hospitals is performed by nurses using a paper-based guideline that allows anticipative, intuitive decision-making. An electronic alert was implemented on 1 August 2007 in which a pop-up appears in the PDMS at the following BG thresholds: >180, >110, 60-80, 40-60 and <40mg/dl. The impact of this electronic alert was assessed by a sequential cohort study: the mean BG, the glycemic penalty index (GPI), the hyperglycemic index (HGI), the number of hypoglycemic events, the standard deviation (SD) of the BG time series and the number of BG measurements were compared in all adults admitted 6months before ('pre-alert group', n=729) and after ('alert group', n=644) the implementation of the electronic alert. The alert resulted in a reduction of mean BG (112 vs. 110mg/dl, p=0.002), GPI (20 vs. 19, p=0.029) and HGI (10 vs. 9mg/dl, p=0.004), without increasing BG sampling (median 25 measurements/patient in both groups, p=0.776). The alert reduced the proportion of patients experiencing an episode of hypoglycemia from 6.5 to 4.0% (p=0.043). The SD of the BG time series was not affected (28mg/dl in both groups, p=0.566). A computer-generated alert was able to statistically significantly improve the quality of TGC in ICU patients without increasing the need for blood sampling.