THE 2009 INSTITUTE OF MEDICINE report defines comparative effectiveness research as comparisons between interventions, as noted by Sherer, Dijkers, Whyte, and Nick. However, preliminary steps for conducting this type of research include "reviewing and synthesizing current research" and "identifying gaps between existing research and the needs of clinical practice," which are 2 goals of the article: Comparative Effectiveness of Traumatic Brain Injury Rehabilitation: Differential Outcomes Across TBI Model Systems Centers.1,2

The purpose of this article was to identify whether there were differences in outcomes across these centers of excellence and to highlight what is currently known about how to measure quality of care. We used methods that are well established in the medical literature.3,4 The intent is not to be punitive to centers with patient outcomes lower than expected but rather to use the better performing centers as a model. Further research by our group is examining the impact of different treatment modalities on outcomes, as well as evidence-based rehabilitation care post-TBI. Our aspirational goal is that this article establishes a multicenter baseline that can be used to assess the impact of future interventions. As we learn more about factors associated with improved outcomes, such as ones suggested by Sherer et al, this can be shared with all TBI model systems centers to collectively enhance outcomes.

Rather than calculating mean differences between observed and expected outcomes, we used each model for the outcomes of interest to calculate an expected outcome for each center, given the population it served. This value was divisible under the center's observed outcome in a ratio (O-E ratio). Sherer et al are correct that the regression models were run for all patients across centers without center effect. We did not want to adjust out center effect since it was under investigation.

We agree with Sherer et al that it is important to use functional status at admission as covariates rather than calculating difference scores in these cases. In fact, as outlined in our methods, functional status at admission to inpatient rehabilitation using Functional Independence Measure and Disability Rating Scale admission scores were included as covariates in the regression models.

Sherer et al express concern regarding speculations on process and structure indicators of quality as possible sources of differences in outcomes across centers. Given that patient demographics, TBI severity, and admission functional status were controlled for in this study, it becomes reasonable to speculate on treatment factors, just as Sherer et al speculate on the possible impact of geography and population on models of rehabilitative care as possibly contributing to these differences. We suggested that process-based variations "may" contribute to center differences, deserving further exploration, and are not "caused by" them. This is an important distinction because our colleagues are correct that we did not have information on variables such as early cognitive status and environmental supports. Consistent with one of the goals described by the 2009 Institute of Medicine report, application of comparative effectiveness research in this study helped generate new ideas for research questions, as evidenced by the comments from reviewers of this article.

REFERENCES