Predicting the Trauma and Injury Severity Score (TRISS) in patients admitted to trauma centers after trauma: A modeling study based on data from the French National Programme for Medicalization of Information

Arthur James, David Hajage, Noémie Simon-Tillaux, Sophie Tezenas du Montcel, Pierre Rufat, Bruno Riou and Mathieu Raux

PLOS ONE, 2026, vol. 21, issue 8, 1-14

Abstract: Background: Reliable mortality prediction is central to trauma research, quality assessment, and resource allocation. While the French national hospital discharge database (PMSI) provides comprehensive nationwide data, it does not capture clinical parameters required for established severity scores such as TRISS, which is used for predicting in-hospital survival after trauma and necessary for implementing national-level evaluations of health system performance. This study tested whether TRISS could nonetheless be accurately derived from PMSI data. Methods: We conducted a retrospective study of adult patients admitted for severe trauma to an academic trauma center between 2012 and 2023. Clinical data from medical records were linked with corresponding PMSI records. Surrogate predictors of TRISS were derived from PMSI codes. An extreme gradient boosting (XGBoost) model was used, model performance was assessed using bootstrap validation and interpretability was explored through SHapley Additive exPlanations analyses. Results: A total of 2,523 trauma patients were successfully matched between PMSI and medical records. Median age was 35 years, 79% were male, and in-hospital mortality was 15%. Correlation between Simplified Acute Physiology Score II (SAPS II) in PMSI and medical records was strong (ρ = 0.88), while concordance for shock and coma was weak (κ ≈ 0.25). The XGBoost model using individual PMSI codes achieved the best performance (optimism-adjusted R2 = 0.72). SAPS II was the main contributor, with additional contributions from shock, coma, and anatomical injury codes. Conclusion: TRISS can be reliably approximated from PMSI data using an XGBoost model. This approach enables patient-specific assessment of trauma severity and mortality risk from PMSI, opening perspectives for nationwide surveillance and quality evaluation of trauma care in France.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0354700 (text/html)
https://journals.plos.org/plosone/article/file?id= ... 54700&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pone00:0354700

DOI: 10.1371/journal.pone.0354700

Access Statistics for this article

More articles in PLOS ONE from Public Library of Science
Bibliographic data for series maintained by plosone ().