A simple hospital discharge score predicts Glasgow Outcome Scale at 12 months in patients with traumatic brain injury

Introduction

This study examines whether a hospital discharge score predicts future Glasgow Outcome Scale (GOS) following traumatic brain injury (TBI). Loss to follow-up is a major problem in TBI studies and an early prediction score could be used for interpolation of missing outcome data or as a surrogate outcome measure.

Methods

Adults admitted to the Queens Medical Centre between 1993 and 2002 with TBI and GCS <12 were included prospectively in the Nottingham Head Injury Register. At hospital discharge the modified Barthel Index was recorded by nursing staff and a Discharge Disability Score (DDS) calculated: 1. Independent, 2. Dependent <24 hour care, 3. Fully dependent, 4. Dead [1]. Glasgow Outcome Scale (GOS) was calculated at 12 months by the patient's general practitioner [2]. Unfavourable outcome was defined as DDS 3, 4 or GOS 1, 2, 3. Correlation between DDS and GOS was measured and the ability of DDS to predict GOS at 12 months in survivors was assessed using logistic regression. Sensitivity and specificity of dichotomised DDS for prediction of unfavourable outcome at 12 months was also calculated.

Results

Data were available on 1,227 patients. A highly correlated linear relationship was evident between DDS and GOS (spearman correlation coefficient 0.77, P < 0.0001). Unfavourable outcome measured by DDS showed sensitivity of 73% (95% CI 69 to 76%) and specificity of 56% (53 to 59%) for prediction of unfavourable outcome at 12 months defined by GOS. DDS showed a strong association with GOS at 12 months for survivors at discharge, with fully dependent patients having 15 times higher odds of an unfavourable outcome. Adjusting for known TBI prognostic factors attenuated the strength of the association, however DDS remained a statistically significant strong predictor of outcome.

Conclusions

DDS is strongly associated with 12-month GOS and could potentially be used to replace missing data or provide a surrogate outcome measure in TBI trials. A valid prediction score may also aid the clinician's ability to discuss patient prognosis at hospital discharge.