This study reveals that combining stress hyperglycemia ratio and glucose variability significantly improves mortality prediction in sepsis, with risk patterns varying significantly based on a patient's baseline metabolic status (Normal, Pre-DM, or DM).