Predicting survival in older patients treated for cervical spine fractures: development of a clinical survival score

BACKGROUND CONTEXT Emerging literature has identified the importance of pretreatment health and functional status as influential in the prognostication of survival. A comprehensive, accessible, predictive model for survival following cervical spine fracture has yet to be developed. PURPOSE To develop an accessible and intuitive predictive model for survival in individuals aged 50 and older treated for cervical spine fractures. STUDY DESIGN Retrospective review of records from two tertiary care centers (2009–2016). PATIENT SAMPLE Patients age 50 and older who received operative or nonoperative management for cervical fractures. OUTCOME MEASURES One-year mortality was the primary outcome with 3-month and 2-year mortality considered secondarily. METHODS Multivariable logistic regression was used to identify factors independently associated with mortality. The magnitude and precision of the relationship with 1-year mortality for statistically significant variables determined weighting in the scoring system subsequently developed. Score performance was tested through multivariable regression and bootstrap simulation. In a sensitivity test, the performance of the score developed for 1-year mortality was assessed using figures for the 3-month and 2-year time-points. RESULTS We included 1,758 patients. Mortality rates were 12% at 3 months, 17% at 1 year, and 21% at 2 years. Following multivariable testing age, injury severity score and Glasgow coma scale demonstrated the strongest predictive values for a base score, followed by serum albumin and ambulatory status. The resultant composite score ranged from 0 (base score≤4, albumin≤3.5 g/dL, and dependent/nonambulator at presentation) to a maximum of 4 (base score≥5, albumin>3.5 g/dL, and independent ambulator at presentation). Following multivariable analysis, when compared to patients with a score of 4, significantly increased odds of 1-year mortality were appreciated for those with scores of 3 (odds ratio [OR] 7.35; 95% confidence interval [CI] 3.77, 14.32), 2 (OR 8.43; 95% CI 4.66, 15.25), 1 (OR 17.47; 95% CI 9.81, 31.11), and 0 (OR 26.58; 95% CI 13.87, 50.92). Score performance was unchanged in bootstrap testing and sensitivity analyses. CONCLUSIONS We have developed a useful prognostic utility capable of informing survival in individuals age 50 and older, following cervical spine fractures. The score can be applied to adjust patient expectations, anticipate outcomes, and as an adjunct to decision-making in the postinjury period.