Predicting Outcomes in Patients with Diffuse Axonal Injury: External Validation of the Widely Used Prognostic Instruments
Rita Vieira 1Regina Sousa 2Wellingson Paiva 3Leonardo Pipek 4Daniel Oliveira 3Daniel Godoy 5Camila Souza 6Jacob Stubbs 7William Panenka 7
1 CAPES Foundation, Ministry of Brazil, 70040-020 Brasilia, Brazil; School of Nursing, University of Sao Paulo, 05508-220 Sao Paulo, Brazil; Faculty of Medicine, University of British Columbia, Vancouver, BC V6Z 3B7, Canada
2 School of Nursing, University of Sao Paulo, 05508-220 Sao Paulo, Brazil
3 Department of Neurosurgery, Clinical Hospital, University of Sao Paulo School of Medicine, 01246-903 Sao Paulo, Brazil
4 Department of Neurology, Clinical Hospital, University of Sao Paulo School of Medicine, 01246-903 Sao Paulo, Brazil
5 Unidad de Cuidados Neurointensivos Sanatorio Pasteur, Unidad de Terapia Intensiva, Hospital Interzonal de Agudos "San Juan Bautista", K4703AFI Catamarca, Argentina
6 Department of Statistical and Actuarial Sciences, University of Western Ontario, London, ON N6A 3K7, Canada
7 Faculty of Medicine, University of British Columbia, Vancouver, BC V6Z 3B7, Canada
Ann. Ital. Chir., 2024, 95(3), 100195; https://doi.org/10.62713/aic.3510
Published: 20 Jun 2024
Copyright © 2024 The Author(s).
Abstract
AIM: Accurate prognosis of diffuse axonal injury (DAI) is important in directing clinical care, allocating resources appropriately, and communicating with families and surrogate decision-makers. METHODS: A study was conducted on patients with clinical DAI due to closed-head traumatic brain injury treated at a trauma center in Brazil from July 2013 to September 2015. The objective efficacy of the Glasgow Coma Scale (GCS), Trauma and Injury Severity Scoring system (TRISS), New Trauma and Injury Severity Scoring system (NTRISS), Abbreviated Injury Scale (AIS)/head, Corticosteroid Randomization After Significant Head Injury (CRASH), and International Mission on Prognosis and Analysis of Clinical Trials (IMPACT) models in the prediction of mortality at 14 days and 6-months and unfavorable outcomes at 6 months was tested. RESULTS: Our cohort comprised 95 prospectively recruited adults (85 males, 10 females, mean age 30.3 ± 10.9 years) admitted with DAI. Model efficacy was assessed through discrimination (area under the curve [AUC]), and Cox calibration. The AIS/head, TRISS, NTRISS, CRASH, and IMPACT models were able to discriminate both mortality and unfavorable outcomes (AUC 0.78–0.87). IMPACT models resulted in a statistically perfect calibration for both 6-month outcome variables; mortality and 6-month unfavorable outcome. Calibration also revealed that TRISS, NTRISS, and CRASH systematically overpredicted both outcomes, except for 6-month unfavorable outcome with TRISS. CONCLUSIONS: The results of this study suggest that TRISS, NTRISS, CRASH, and IMPACT models satisfactorily discriminate between mortality and unfavorable outcomes. However, only the TRISS and IMPACT models showed accurate calibration when predicting 6-month unfavorable outcome.
