Object. The aims of this study were to analyze the relevance of risk factors in mild head injury (MHI) by studying the possibility of establishing prediction models based on these factors and to evaluate the reliability of the clinical guidelines proposed for the management of MHI. Methods. A series of 1101 patients with MHI were prospectively enrolled in this study. In all cases clinical data were collected and a computerized tomography (CT) scan was obtained. The relationship between clinical findings and the presence of intracranial lesions was studied to establish prediction models based on logistic regression and recursive partitioning analysis. Recently proposed guidelines and recommendations for the treatment of MHI were selected, calculating their diagnostic efficiency when applying each of them to our series. The incidence of acute intracranial lesions was 7.5% (83 patients). A Glasgow Coma Scale score of 14, loss of consciousness, vomiting, headache, signs of basilar skull fracture, neurological deficit, coagulopathies, hydrocephalus treated with shunt insertion, associated extracranial lesions, and patient age greater than 65 years were identified as independent risk factors. Prediction models built on clinical variables were able to indicate patients with clinically important lesions, but failed to achieve 100% sensitivity in the detection of all patients with CT scans positive for intracranial lesions within reasonable specificity limits. Conclusions. Clinical variables are insufficient to predict all cases of intracranial lesions following MHI, although they can be used to detect patients with relevant injuries. Avoiding systematic CT scan indication implies a rate of misdiagnosis that should be known and assumed when planning treatment in these patients by using guidelines based on clinical parameters.