Role of NSI in Differentiating Between Mild Traumatic Brain Injury And Behavioral Health Conditions

Archival data will be collected from all OEF/OIF Veterans who were seen in the VA ECHCS TBI Clinic by Nancy Cutter, M.D. and her team between January 1, 2009 and December 31, 2010. Electronic Medical Record (EMR) progress note titles "OEF/OIF TBI 2nd Level Evaluation Consult Report" and "TBI Consult Report" will be used to identify potential subjects.

Power is based on the primary hypothesis that those with PTSD only (Group 3) and those with target, service-related mTBI only (Group 2) will have significantly more PC symptom reporting than those with no history of target, service-related mTBI or PTSD (Group 4). Belanger et al. (2010) reported a standard deviation of 15.3 on the NSI total score in a sample of 134 mTBI subjects. Assuming variability similar to the Belanger study, a significance level of 0.025 to correct for the two comparisons and 80% power, 180 subjects per group will detect a clinically significant difference of 5 points. For simplicity, this was calculated using a two-sided, two-sample t-test. Given that we are modeling all four groups together, pooled standard error estimates will be used, which will result in slightly higher power.

All analyses will assume a two-sided test of hypothesis with an overall significance level of 0.05, unless otherwise noted, and will be performed in either Statistical Analysis Software (SAS) v9.2 or above (SAS Institute, Inc., Cary, NC).

Demographic characteristics will be reported as means and standard deviations; medians and ranges; and proportions, as appropriate. Likewise, characteristics will be compared between the four aforementioned subject groups using Analysis of Variance (ANOVA), chi-square tests and/or nonparametric tests as appropriate.

Hypothesis 1: An ANOVA with reference cell coding will be utilized to model NSI total score as a function of the four groups (Group 1: Individuals with co-occurring target, service-related, mTBI and PTSD; Group 2: those with target, service-related, mTBI only; Group 3: those with PTSD only; and Group 4: those with no history of target, service-related, mTBI or PTSD). Potential confounders will then be assessed individually by adding them to the model with the groups. If any of the group parameter estimates changes by more than 10% with the inclusion of the potential confounder, the variable will be utilized in the final model. Once the final model is determined, a contrast will be set up within the model to test (1) PTSD only vs. Neither and (2) target, service-related, mTBI only vs. Neither, a Bonferroni correction will be employed such that the significance level will be adjusted to 0.025 for these two primary tests. Estimated mean differences will be reported with 95% CIs. Potential confounding variables to be considered are: age; gender; total number of deployments; time since last deployment; time since earliest documented mTBI; time since target, service-related, mTBI; and total number of mTBIs.

Hypothesis 2: Within the same final model above, a contrast will be used to test (1) those with co-occurring target, service-related, mTBI and PTSD vs. PTSD only and (2) those with co-occurring target, service-related, mTBI and PTSD vs. target, service-related, mTBI only.