Abstract:
Objectives: To identify the dimensionality of outcomes following serious TBI. To determine whether outcomes are highly multidimensional, or not. To determine which items fit which dimension, what items are redundant, and what items do not “fit”. To determine whether there are anomalous cases of TBI whose outcome cannot be measured using usual methods.

Design: One year follow up data from the National TBI Model Systems database were analyzed. We selected all items indicative of global outcomes, social activities, disability, and handicap or community participation (excluding items indicating highly specific neuropsychological or medical impairment). Data were analyzed using Rasch analysis (WinSteps), an advanced form of psychometric analysis that constructs equal-interval measures from samples varying greatly in ability, and identifies person-misfit as well as item misfit.

Results: Although we expected to find many outcome dimensions, what we found was an impressive degree of unidimensionality. It is clear that there is a powerful singular central dimension to TBI outcomes. A suitable name for this dimension is “general functioning”. It incorporates the classic medical outcome measure – the Glasgow Outcome Scale (GOS) – as well as other general outcome measures (DRS, rescaled SRS, most items from the FIM and CIQ, and selected items from NFI, as well as indicators of living arrangement, and hours worked.) The actual dimensionality is very different from the conventional designation of “brand name” scales. Many items were redundant.

Significance: One of the major problems in TBI outcomes and effectiveness research has been uncertainty about what outcome measure to use, as hundreds have been developed and used. The true dimensionality of outcomes is now known. With additional research, it should in the near future be possible to develop outcome measures with greatly reduced numbers of items. Outcomes measurement will be briefer and yet more reliable and meaningful. Outcomes can be truly measured, using an equal-interval measure, which has not been possible in the past. Artifacts will be lessened, and studies of the effectiveness of treatment will be substantially facilitated.