Letter to the Editor

To the editors,

I wish to express concern about the methods reported by Perkes et al. [1] in the February 2011 issue of Brain Injury. Several important problems with the internal validity of the research call into question the authors’ conclusions.

The authors state that their overall objective was to compare traumatic brain injury (TBI) rates and sequelae in prisoners with a matched community control group. The authors matched the two groups to assure subjects came from similarly low SES neighbourhoods. However, a major concern with the article is that the groups were not comparable in terms of multiple, other important characteristics. The percentages and p-values in Table II show these discrepancies. Statistically significant and clinically important differences were evident for age, marital status, Aboriginal or Torres Strait Islander descent, education, participation in contact sports, alcohol consumption and illicit drug use. Most, if not all, of these demographic characteristics have been associated with the occurrence of TBI. While some of the variables in question were used in subsequent regression modelling, there was no explanation for why marital status, Aboriginal or Torres Strait Islander descent and participation in contact sports were not. Regardless, statistical control for all of these demographic differences would have a very limited likelihood of success, especially given the sample sizes used.

Several differences should be acknowledged in how data about TBI were collected from the custody and community samples. First, the data collected by different research entities. The authors do not mention if training or other methods were employed to standardize the training of interviewers. A computer-assisted telephone interview (CATI) is mentioned in relation to the community sample, but not the custody sample. Second, two different interview modalities were employed––one was conducted by telephone, the other was a face-to-face interview. The potential effect of this difference is unknown. However, a third, and more troubling discrepancy in the data collection method was difference in the structure of the elicitation question. Both groups were asked for an overall number of TBIs from which they became ‘dazed or confused without loss of consciousness, or unconscious or blacked out’ but for the prisoner group further inquiry was limited to details about up to five individual TBI episodes (the most recent, first, most severe, second most severe and third most severe). In contrast, the community sample was asked about their first five TBIs, although it is unclear if these were the first five mentioned or the first five in their lifetime. As the authors point out in the discussion section, ‘large differences in the rates of past TBI among non-prisoner samples are likely to be due, at least in part, to differences in the precise wording of the inquiry about TBI’ (p. 139). It is not clear what effects all of these differences in elicitation method had on the resulting data, but, when viewed together, serious concerns arise about non-equivalent methods.

The differences noted here call into question the ability to compare results from the two groups and undermine the authors’ conclusion that there was ‘no significant association between TBI frequency or severity and custody/community group membership’(p. 140). They further state, ‘One cannot invoke causality based on the results of cross-sectional analyses, but these analyses provide little support for the notion that TBI leads to much offending behaviour and invite speculation that personality characteristics (impulsivity and dissocial) may be considerably more salient’ (p. 139). This statement is misleading, as the authors violate their own caution. More important, given that non-comparable samples and differing methods of data elicitation were utilized, any conclusion about similarities or differences between the groups is significantly compromised. The methods used in this study highlight the need for validated tools for eliciting history of TBI.