Treating hyperglycemia may improve patient outcome, but is a clinical challenge. Three variations of a computerized insulin protocol were compared with regard to protocol compliance and achievement of glucose target levels. In group 1, the existing protocol was applied, in group 2 the protocol was modified to account for decreasing glucose values; group 3 had a higher threshold for initiating insulin, wider glucose target ranges, and included instructions to regulate glucose around mealtimes. From July 28, 2008 until February 1, 2010, data from 1255 patients admitted to our Intensive Cardiac Care Unit with at least 2 glucose measurements were analyzed. Mean age was 64 +/- 15 years, 66% were male, 21% had diabetes. Groups 1 to 3 included 269, 814, and 142 patients, respectively. Protocol compliance in group 2 was lower with 44% of the glucose measurements performed on time versus 51% in group 1 (P < 0.001), and insulin was dosed correctly in 57% versus 67% (P < 0.001). In group 3, compliance increased, 52% of the measurements were done on time, and insulin was dosed correctly in 71%. Average glucose levels increased in group 3 due to a higher threshold for starting insulin and a wider target range: 70% (group 1), 66% (group 2), and 61% (group 3) had an average glucose of <8 mmol/L (P < 0.001). Also, we observed a decreasing trend in incidence of hypoglycemia and reporting of noncompliance. Further improvements in glucose measurement technology and protocols are needed to optimally treat hyperglycemia in the Intensive Cardiac Care Unit.