LAUSR

STUDY DESIGN Retrospective analysis on prospectively collected data. OBJECTIVE Construct a clinical prediction model for 90-day mortality in elderly patients with traumatic spine injuries. SUMMARY OF BACKGROUND DATA Spine trauma in the elderly population is increasing. Comparing elderly spine trauma patients to younger patients with similar injuries proves challenging due to the extensive comorbidities and frailty found in the elderly. There is a paucity of evidence to predict survival of elderly patients following traumatic spinal injuries. METHODS All patients 65+ with spine trauma presenting to a level I trauma center from 2010-2019 were reviewed from a prospectively maintained trauma registry. Retrospective chart review was performed to record injury, frailty scores, comorbidities, presence of spinal cord injury, imaging evidence of sarcopenia and osteopenia, mortality, and complications. We preselected 13 variables for our multivariable logistic regression model: hypotension on admission, gender, marital status, age, max Abbreviated Injury Scale, Modified Frailty Index, surgical treatment, hematocrit, white blood count, spinal cord injury, closed head injury, injury level and presence of high energy mechanism. The performance of the prediction model was evaluated using a concordance index and calibration plot. The model was internally validated via bootstrap approach. RESULTS Over the 9-year period, 1746 patients met inclusion criteria. 359 (20.6%) of patients died within 90 days after spine trauma presentation. The most important predictors for 90-day mortality were age, hypotension, closed head injury, max AIS and hematocrit. There was an optimism-corrected C-index of 0.77. A calculator was created to predict a personalized mortality risk. CONCLUSIONS The incidence of spine trauma in elderly patients continues to increase. Previous publications described preexisting conditions that imply increased mortality, but ours is the first to develop a predictive calculator. Prospective research is planned to externally validate this model to better determine its predictive value and utility in the clinical setting.