Adult prediction rules for cervical spine injury (CSI) exist; however, pediatric rules do not. Our objectives were to determine test accuracies of retrospectively identified CSI risk factors in a prospective pediatric cohort and compare them to a de novo risk model. We conducted a 4-center, prospective observational study of children 0 to 17 years old who experienced blunt trauma and underwent emergency medical services scene response, trauma evaluation, and/or cervical imaging. Emergency department providers recorded CSI risk factors. CSIs were classified by reviewing imaging, consultations, and/or telephone follow-up. We calculated bivariable relative risks, multivariable odds ratios, and test characteristics for the retrospective risk model and a de novo model. Of 4091 enrolled children, 74 (1.8%) had CSIs. Fourteen factors had bivariable associations with CSIs: diving, axial load, clotheslining, loss of consciousness, neck pain, inability to move neck, altered mental status, signs of basilar skull fracture, torso injury, thoracic injury, intubation, respiratory distress, decreased oxygen saturation, and neurologic deficits. The retrospective model (high-risk motor vehicle crash, diving, predisposing condition, neck pain, decreased neck mobility (report or exam), altered mental status, neurologic deficits, or torso injury) was 90.5% (95% confidence interval: 83.9%-97.2%) sensitive and 45.6% (44.0%-47.1%) specific for CSIs. The de novo model (diving, axial load, neck pain, inability to move neck, altered mental status, intubation, or respiratory distress) was 92.0% (85.7%-98.1%) sensitive and 50.3% (48.7%-51.8%) specific. Our findings support previously identified pediatric CSI risk factors and prospective pediatric CSI prediction rule development.