A statistical model for predicting the occurrence of intracranial lesions in patients with mild head trauma

Objective(s): To develop a statistical model to differentiate between patients who need CT scan and those who did not. Methods: 320 consecutive minor head trauma victims admitted in two teaching hospitals were entered into the study. After completion of medical care, clinical information were collected by physicians. Brain CT Scan was taken from all patients and after clinical use, was filed to be reported by a single radiologist for this research purpose. Logistic Regression method was utilized to analyze results and develop the prediction model. Results: Using logistic regression, a model was developed which showed 100% sensitivity and 74% specificity to detect a positive finding on brain CT scan. Using this model negative finding prevalence in minor head trauma brain scans decreased by 66% while no cases were missed. Conclusion: The proposed statistical model reveals the need for brain CT scan in 70% of minor head trauma victims while preventing any positive findings being missed. This model would be considered as a Clinical Practice Guideline (CPG) in rural areas to refer minor head trauma victims for Brain CT Scan. However, more investigation of this model with more sample size is recommended.