Citation

Silverberg N, Gardner AJ, Brubacher J, Panenka W, Li JJ, Iverson GL. J. Neurotrauma 2014; 32(8): 517-526.

Affiliation

UBC, Medicine , 4255 Laurel St , Vancouver, British Columbia, Canada, V5Z 4R2 ; noah.silverberg@vch.ca.

Copyright

(Copyright © 2014, Mary Ann Liebert Publishers)

DOI

10.1089/neu.2014.3600

PMID

25222514

Abstract

Prognostic models can guide clinical management and increase statistical power in clinical trials. The availability and adequacy of prognostic models for mild traumatic brain injury (MTBI) is uncertain. The present study aimed to (i) identify and evaluate multivariable prognostic models for MTBI, and (ii) determine which pre-, peri-, and early post-injury variables have independent prognostic value in the context of multivariable models. An electronic search of MEDLINE, PsycINFO, PubMed, EMBASE, and CINAHL databases for English-language MTBI cohort studies from 1970-2013 was supplemented by Web of Science citation and hand searching. This search strategy identified 7,789 articles after removing duplicates. Of 182 full-text articles reviewed, 26 met eligibility criteria including (i) prospective inception cohort design, (ii) prognostic information collected within one month post-injury, and (iii) 2+ variables combined to predict clinical outcome (e.g., postconcussion syndrome) at least one month later. Independent reviewers extracted sample characteristics, study design features, clinical outcome variables, predictor selection methods, and prognostic model discrimination, calibration, and cross-validation. These data elements were synthesized qualitatively. The present review found no multivariable prognostic model that adequately predicts individual patient outcomes from MTBI. Suboptimal methodology limits their reproducibility and clinical usefulness. The most robust prognostic factors in the context of multivariable models were pre-injury mental health and early post-injury neuropsychological functioning. Women and adults with early post-injury anxiety also have worse prognoses. Relative to these factors, the severity of MTBI had little long-term prognostic value. Future prognostic studies should consider a broad range of biopsychosocial predictors in large inception cohorts.

Language: en