Long-Term Survival After Traumatic Brain Injury Part I: External Validity of Prognostic Models

Abstract

ObjectivesTo develop prognostic models for long-term survival in adults with traumatic brain injury (TBI) and to assess their external validity in 2 independent populations. DesignSurvival analysis. SettingPost-discharge from rehabilitation units and long-term follow-up at regional centers. ParticipantsTwo cohorts of long-term survivors of TBI (N=12,481): the Traumatic Brain Injury Model Systems (TBIMS) cohort comprised 7365 persons who were admitted to a TBIMS facility and were assessed at ≥1 years postinjury, and the California Department of Developmental Services (CDDS) cohort comprised 5116 persons who sustained a TBI and received long-term services from the CDDS. InterventionsNot applicable. Main Outcome MeasuresSurvival/mortality. ResultsOlder age, male sex, and severity of disability in walking and feeding were significant predictors of increased long-term mortality rates (all P<.05, both databases). The CDDS model predicted 623 deaths for persons in the TBIMS cohort, with an observed-to-expected ratio of .94 (95% confidence interval [CI], 0.87–1.02). The TBIMS model predicted a total of 525 deaths for persons in the CDDS cohort, with an observed-to-expected ratio of 1.08 (95% CI, 0.99–1.17). Regression calibration statistics were satisfactory, and both models ranked survival times well from shortest to longest (TBIMS: C index, .78; 95% CI, .76–.80; CDDS: C index, .80; 95% CI, .78–.82). ConclusionsLong-term survival prognosis in TBI is related to age, sex, and severity of disability. When compared on the basis of these factors, the survival estimates derived from the TBIMS and CDDS cohorts are found to be similar. The close agreement between model predictions and actual mortality rates confirm the external validity of the prognostic models presented herein.