Practical Prognostics

Virginia Commonwealth University, Richmond VA

Investigators

Linked publications & trials

Abstract

Project Summary/Abstract (Walker WC; Practical Prognostics) Traumatic brain injury (TBI) is a very heterogeneous anatomical and physiological condition with extremely variable outcomes. For survivors of higher TBI severity grades, functional recovery is usually incomplete and protracted, and can range from total dependence to full recovery. While this offers hope for specific individuals, it also creates enormous uncertainty in prognosis. Facing this uncertain future, patients and their families desperately want and need meaningful prognostic information. Unfortunately, the extensive TBI outcomes literature has largely failed to inform clinical prognosis, and predicting long- term functional outcome remains particularly challenging. For TBI severities greater than mild, the standard clinical prognosis given is ?time will tell? or ?we honestly don't know.? Clearly, better clinically relevant prognostic models are needed. This study will analyze clinical data from survivors of closed TBI enrolled in multi-center database (NIDILRR TBI Model Systems; N>14,000) and develop a set of user-friendly prognostic tools for 1, 2, and 5 year functional outcomes (Glasgow Outcome Scale and employment). Models will be built using classification tree methodology that permits multiway splits, a more robust way of estimation compared to the few past TBI studies using decision tree methods. First, TBI severity will be stratified by post-traumatic amnesia duration, then select clinical variables (injury, health, and demographic) will be entered. The result will be a set of prognostic tools that will empower providers to give meaningful prognostic information to survivors and their families. It will also help set rehabilitation expectations and serve as a basis for preliminary rehabilitative treatment planning. Using the information found in the models, several post-injury modifiable conditions will also be assessed to find patient groups at risk for having poorer outcomes from these conditions and potential candidates for targeted therapy. The specific aims of our proposed study are as follows: 1. Build the decision tree prognostic models for long-term GOS and employment outcomes 2. Assess their generalizability in large, independent datasets from TBI-MS and [exploratory aim] patients in the Transforming Research and Clinical Knowledge in TBI II database. 3. Using the classifications of outcomes from the previous method, assess how modifiable conditions (depression, anxiety, substance misuse, emotional problems) are related to the outcomes after adjusting for the demographic and baseline injury characteristics using the predictions from the prognostic tool.