Document Type : Letter to the Editor

Authors

• Mohammad Reza Rahmanian ¹
• Abolfazl Shahedi ²

¹ School of medicine, Shiraz University of Medical Sciences, Shiraz, Iran.

² School of medicine, Shiraz University of medical Sciences, Shiraz,Iran.

Abstract

We evaluated the recent prospective study by Hamidnezhad et al. comparing Trauma and Injury Severity Score (TRISS) and Madras Head Injury Prognostic Scale (MHIPS) in traumatic brain injury. The authors showed that both tools are good predictors of mortality and ICU admission, with higher specificity for TRISS and higher sensitivity for MHIPS. Their two-endpoint methodology, sound approach, and robust statistical methodology greatly enhance confidence in the result. However, several areas for future improvement still exist. Calibration, necessary to evaluate agreement between predicted and observed outcomes, was not assessed, although it is key to prognostic modelling. Further, Decision Curve Analysis might illuminate net clinical benefit over threshold probabilities. Wider issues in trauma prognostic models, like small sample sizes, sparse external validation, unsatisfactory treatment of missing data, and dichotomizing of continuous predictors, also need to be improved. Lastly, stability of models deserves more consideration to permit transportability to other settings. We praise the authors' contribution and issue these improvements as recommendations for future prognostic studies in head injury.

Keywords

• Traumatic Brain Injury
• Prognostic Models
• Model Calibration