Neuropsychological Studies of Mild Traumatic Brain Injury: A Meta-Analytic Review of Research Since 1995

Abstract

A meta-analysis conducted by Binder, Rohling and Larrabee (1997) established a relationship between mild traumatic brain injury (TBI) and small reductions in cognitive functioning in individuals assessed more than 3 months post-injury. As a follow-up, this study summarized similar research that (1) was published since the previous meta-analysis, and (2) included data collected at any stage post-injury. An extensive literature search revealed 17 suitable studies from which effect sizes were aggregated. The overall effect size was g = 0.32, p < .001. Speed of processing measures had the largest effect, g = 0.47, p < .001. The merging of post-acute effect sizes with those reported in Binder et al.'s review yielded a nonsignificant result, g = 0.11. Time since injury was found to be a significant moderator variable, with effect sizes tending to zero with increasing time post injury.

Thanks go to the authors of papers included in this analysis who willingly provided additional data and information to assist in the completion of this meta-analysis. This study was supported by an Australian Postgraduate Award provided to K. Frencham.

Notes

Thanks go to the authors of papers included in this analysis who willingly provided additional data and information to assist in the completion of this meta-analysis. This study was supported by an Australian Postgraduate Award provided to K. Frencham.

Note that Binder et al. (1997) used the labels d for the effect-size statistic calculated using the standard deviation of the control group, and g for the statistic calculated using the pooled standard deviation. Our use of the labels d and g is different, but arguably more consistent with the meta-analysis literature (e.g. see Rosenthal and Dimatteo, 2002).

To simplify interpretation, non-parametric statistics are reported for all relationships with time since injury. However, time since injury was also transformed into a logarithmic function to counteract the strong positive skew of the distribution. All correlations using transformed data were found to be comparable to the non-parametric analyses reported in the body of this report.

References marked with an asterisk indicate studies included in the meta-analysis.