RESUMO
PURPOSE: To move beyond the correlation analysis between glucose statistics and patient outcome and to benchmark summary statistics calculated on all glucose values taken during an entire individual patient-stay. METHOD: The sample consisted of 141 patient-stays with multiple finger-stick blood glucose (BG) measurements resulting in 5,428 total BGs. Glucose control assessments of the BG profiles for each patient-stay were independently conducted by two endocrinologists. Summary statistics (mean, standard deviation, mean absolute glucose [MAG], percentage of hours of patient-stay with BG over threshold) were calculated for BG using patient-stay as the unit of analysis. RESULTS: Produced numerical benchmarks for the mean (148 mg/dL), SD (40 mg/dL), MAG (6.2), and the percentage of patient-stay hours over 200 mg/dL (10%). Specificity and sensitivity were high, ranging from 91% to 89% for specificity and from 90% to 81% for sensitivity. CONCLUSION: This study demonstrated that benchmarks developed on specific patient populations can be used to evaluate glycemic control in a straightforward, computationally simple system.
