24-hour risk updating for major ICU complications after severe trauma: a prospective cohort study

Conventional trauma prognostic models rely primarily on admission data and cannot account for subsequent physiological response to treatment. We evaluated whether routinely collected physiological information available at 24 h provides incremental prognostic information for major intensive care unit (ICU) complications in severely injured patients.

We conducted a prospective cohort of adults with severe trauma (Injury Severity Score [ISS] ≥ 15) admitted to the ICU between January 2024 and April 2025. The primary outcome was a composite of major respiratory, hemodynamic, renal, or septic complications during the ICU course. TRISS was compared with an admission ridge model incorporating ISS, Glasgow Coma Scale, hemoglobin, lactate, and international normalized ratio, and a 24-h updated ridge model additionally incorporating 24-h lactate and change in creatinine. Performance was internally validated with 1,000 bootstrap resamples.

Among 109 patients, median ISS was 29 (IQR 20-38), and 73 (67.0%) developed major ICU complications. Optimism-corrected AUCs were 0.788 for TRISS, 0.873 for the admission model, and 0.887 for the 24-h updated model; Brier scores were 0.178, 0.140, and 0.131. Compared with the admission model, the 24-h updated model improved performance (ΔAUC 0.016, 95% bootstrap interval 0.004-0.030; ΔBrier - 0.010, - 0.017 to - 0.005). Twenty-four-hour lactate and creatinine change were the strongest contributors to the update.

In severely injured ICU patients, incorporating physiological evolution during the first 24 h refined admission-based prognostic assessment. Lactate and creatinine trajectories provided incremental information for major ICU complications during the ICU course. External validation is required before clinical implementation.