In pediatric cardiac intensive care, some of the most challenging moments occur immediately after surgery, when patients may appear stable despite carrying very different physiological burdens.
Two children can arrive in the intensive care unit after similar congenital heart procedures with comparable vital signs and similar surgical complexity, yet their clinical trajectories may diverge within hours. One recovers quickly, while the other develops prolonged ventilation, circulatory instability, or unexpected complications.
That repeated clinical observation motivated this study.
Existing systems such as RACHS and STAT are highly valuable because they estimate procedural complexity before surgery. However, they do not fully capture how each child physiologically responds to cardiopulmonary bypass and early postoperative stress.
We therefore explored whether early metabolic signals could help identify hidden risk earlier.
The original Glycemic Stress Index (GSI) had been proposed in pediatric neurosurgical care, but congenital cardiac surgery introduces additional physiological factors that glucose alone may not fully represent.
To address this, we modified the score step by step.
First, operative duration was replaced by cardiopulmonary bypass time, which more directly reflects extracorporeal physiological stress. Second, the vasoactive-inotropic score (VIS) was incorporated to account for early postoperative cardiovascular support requirements. Finally, age adjustment was added because younger children respond differently to metabolic stress than older pediatric patients.
Each modification was designed to reflect bedside physiology rather than purely mathematical optimization.
The final Modified Stress Score demonstrated improved prediction of mortality, postoperative complications, ventilation duration, and intensive care stay compared with the original GSI.
Another clinically interesting finding involved patients with simultaneous hyperglycemia and elevated lactate, a pattern we described as metabolic uncoupling. These children often represented the subgroup whose early postoperative appearance could be misleading before later deterioration developed.
Because all variables used in the score are routinely available in daily practice, the model may be clinically applicable without additional cost or delayed testing.
This work also opens the next step: evaluating whether dynamic markers such as lactate clearance, early creatinine change, or vasoactive trends can further improve postoperative prediction.
Like many clinically useful studies, this project began not with statistics, but with a repeated bedside question: why do some children deteriorate when early signs seem reassuring?