Purpose of review: Prognostic models for predicting outcome after severe traumatic brain injury (TBI) may be useful in several areas. However, established risk prediction models for general critical illness show significant limitations in neurotrauma. Development of specific risk prediction models for TBI has been difficult due to the variability of injury, which predicates a large sample for construction of robust models. Previous development of prognostic models for TBI has suffered from small sample sizes, poor study design and follow up, difficulty in application to clinical practice, limited inclusion of patients from low income countries, and lack of external validation.
Recent findings: This situation has changed substantially in the last year with the development and validation of two new risk prediction models based on large datasets.The Corticosteroid Randomization after Significant Head Injury (CRASH) trial and the International Mission on Prognosis and Analysis of Clinical Trials in TBI (IMPACT) databases have been used to create prediction with or without computed tomography data, and been cross validated. In addition, the CRASH database was used to develop models for low/middle income countries.
Summary: The outcome prediction models that have evolved from these databases are undergoing further refinement and validation, and it is likely that these advances will prove valuable in training clinicians, counselling patients' families, auditing unit performance, designing better clinical trials, and rational allocation of resources.