Background: Through collaborative neurosurgical comanagement services, hospital medicine has had an increasing role in the hospital-based care of the neurosurgical patient. This care model has evolved with the change in demographics of traumatic brain injury (TBI), in which patients are increasingly likely to be medically complex older individuals sustaining their injury after ground-level fall. Compared to younger patients, elderly TBI patients are more likely to be hospitalized, recover more slowly, and exhibit worse outcomes. The International Mission for Prognosis and Clinical Trial Design in TBI (IMPACT) prognostic models have been used to predict outcome following TBI. These models were derived in a cohort primarily composed of younger, healthier randomized controlled trial participants and may not be valid for older patients seen in clinical practice.
Methods: Using data from the National Study on Costs and Outcomes of Trauma (NSCOT) we identified adult patients presenting to US hospitals between July 2001 and November 2002 with non-penetrating moderate or severe TBI (GCS ≤12). The cohort was split into older (65-84 years) and younger (18-64 years) age strata and the predicted risks of death and unfavorable outcome were calculated using the IMPACT core and lab models. Model calibration and discrimination in the older stratum was compared to that in the younger stratum.
Results: We identified 202 older patients (weighted n = 268) and 613 younger patients (weighted n = 1,682) with moderate or severe non-penetrating TBI. Older patients were more likely to have multiple co-morbidities and antiplatelet or anticoagulant use prior to injury. Older patients were more likely to be injured in a fall and more than three times more likely to die within 6 months of injury. IMPACT model discrimination did not differ significantly between older and younger age strata and was generally adequate. IMPACT model calibration was poor for both older and younger strata.
Conclusions: The examined IMPACT prognostic models demonstrated adequate discrimination but poor calibration in both older and younger strata of a population-based sample of patients with moderate-to-severe TBI. These models should be used with caution when risk stratifying geriatric TBI populations.