Background
Patients with cerebral infarction (CI) often present at the emergency room; however, it is difficult to make a correct diagnosis of CI when the patient does not show apparent symptoms. This study aimed to develop and validate a new scoring system for the risk of CI. 

Methods
Data used in this study were from 1,553 patients who were suspected of having CI and were examined using computed tomography (CT) and magnetic resonance imaging (MRI) the same day. In our hospital, all patients with suspected CI were examined using CT and MRI, and CI was diagnosed by a radiologist using MRI imaging. We randomly divided the data into 2 groups [development phase (n=777) and validation phase (n=776)]. Using logistic regression analysis with Akaike information criterion stepwise selection, we determined the best predictive model for CI. The scores of the selected predictors were determined using their odds ratio, and the CI risk score was calculated by adding them together. We examined the predictive ability of the score by generating a receiver operating characteristic (ROC) curve for both the development and validation phases. We calculated the area under the curve (AUC) of the ROC, as well as the sensitivity and specificity at the best cut-off point.

Results
The CI risk score included the following 12 factors: acute onset (<24 h), mean blood pressure, smoking, Barre's sign, visual field loss, motor disorder of the extremities, headache, seizures, history of cerebral stroke, acute low density area of CT, activated partial thromboplastin time, and prothrombin time (Table. 1). In the validation phase, the AUC was 0.75 (95% CI: 0.71–0.79) and the best cut-off point was 10.45 (sensitivity: 88%, specificity: 56%).

Conclusion
We established and validated the CI risk score with relatively high accuracy. We consider that the risk score is useful for not only predicting CI but also deciding on early treatment in clinical settings.

Development and validation of scoring systems for the risk of cerebral infarction