High insulin sensitivity rules out sepsis for the majority of patient hours and provides a negative predictor to obviate pre-emptive antibiotic use. The model-based SI value may be determined in real-time from glycaemic control data. The simple clinical ratio at stable euglycemia can be readily implemented at the bedside. Low insulin sensitivity is not an effective positive predictor, as it can equally mark the presence of sepsis or other conditions. These results indicate the potentially significant diagnostic data available from metabolic markers derived from glycaemic control data in critical care.