A Clinical Prediction Rule to Major Spine Trauma

Abstract

Introduction

Spinal trauma can be devastating and the ultimate solution is the prevention of delayed or missed diagnosis and treatment. Our purpose was to develop a reliable prediction rule to assess a patient's identify risk factors for spinal injuries that could guide trauma patient's screening for timely spinal fracture diagnosis and treatment.

Patients and Methods

Retrospective study from January 2005 to March 2014 of patients with trauma-associated spinal fractures who were admitted into the spinal surgery service of a referral institute. Two clinical prediction rules based on a set of demographic and clinical variables (injury mechanism, accident site, occupation, toxic substances intake, and presence of other fractures) were developed with a multivariate logistic regression model. AO type A spinal fracture and thoracolumbar spinal segment were the outcomes involved. Model performances were quantified with respect to discrimination area under receiver-operating-characteristics curve (AUROC).

Results

In a total of 209 patients, 78.5% were males, median age 36.7 ± 13.9 years and the most affected age group was that of patients aged 20 to 49 years (43.9%). The fracture distribution for axial compression with intact posterior ligamentous complex type A corresponded to 56.5% of the cases; the thoracolumbar junction was the most affected one (44.5%) at one level and 5.3% at two levels, mainly L1 (19.1%). A combination of fall from 1 to 5 m (OR = 3.9, p = 0.000), domestic occupation (OR = 3.1, p = 0.026), recreational place (OR = 5.02, p = 0.000), and toxic substances intake (OR =1.95, p = 0.035), altogether as a prediction rule, showed excellent discrimination in diagnosing AO type A spinal fracture (AUROC = 0.75) with 59% of sensitivity and 77% specificity. In the second model, a combination of falls from 1 to 5 m (OR = 5.8, p = 0.000) and presence of other fractures (pelvic, head, lower, and upper limb fractures) (OR = 5.1, p = 0.006); showed excellent discrimination in diagnosing a spinal fracture at the thoracolumbar junction (AUROC = 0.72) with sensitivity 70% and specificity 74%.

Conclusion

Our prediction rules, including a set of demographic and clinical variables at hand in the emergency room can guide trauma patient's screening for timely spinal fracture diagnosis and treatment. Our models, although stable and with acceptable discrimination should be validated and put to test in other populations.