Optimizing TBI Inpatient Rehabilitation: Longitudinal Analysis of Intervention

项目总结

PROJECT SUMMARY Inpatient rehabilitation for traumatic brain injury has been studied largely as an undifferentiated black box. The prospective, observational TBI Practice-Based Evidence (PBE) study exhaustively documented hundreds of thousands of therapy sessions, a rich set of longitudinal outcomes, and an unrivaled characterization of injury severity and disease comorbidity in order to shed light on this “black box” of care. The analyses to date demonstrated that six “high impact” activities were associated with better outcomes in nearly all severity subgroups and across many different outcomes. The proposed project intends to identify and characterize specific populations of treatment responders and non-responders in the TBI-PBE database and to identify therapy interventions that are associated with better weekly FIM and 9-month outcomes, controlling for patient characteristics and treatment response class. We will use growth mixture modeling (GMM) to identify responder and non-responder classes. The GMM approach to identification of responders and non-responders has advantages over more traditional end-point analysis of treatment response: it reduces bias and improves signal detection over end-point analyses because it accounts for correlations between measures from the same individual and unequal variances of the outcome over time. We will use analysis of covariance (ANCOVA) linear mixed effect regression models to compare post-baseline FIM measures with cumulative time on each of the set of interventions. As a secondary analysis, we will look at FIM associations at other time points to characterize early treatment interventions. The completion of this proposed project will advance knowledge about inpatient rehabilitation care for persons who have experienced moderate and severe TBI and will be highly valuable for rehabilitation treatment planning during this incredibly dynamic period of recovery. The project makes use of what may likely be the richest data set on TBI inpatient rehabilitation ever assembled, which due to its origin in clinical practice, fosters high levels of acceptability from clinicians and demonstrates feasibility to health care system stakeholders. The more accurate and detailed information about how different individuals respond to interventions would allow more patient-specific treatment planning during acute rehabilitation and after discharge. Better understanding of the nature and timing of interventions would inform the staging of therapeutic modalities and would result in more effective use of limited resources. In addition, the proposed research builds directly on the work of the study team to date and has the potential to overcome significant barriers to realizing more informed, effective and equitable care.