A new classification tool for clinical differentiation between haemorrhagic and ischaemic stroke

To develop a simple and reliable diagnostic tool for differentiation of cerebral infarction (CIF) from intracerebral haemorrhage (ICH) in order to aid clinicians to decide about starting antiplatelet therapy in settings where rapid access to computed tomography (CT) is lacking.

Thirty variables regarding each patient admitted with acute stroke were recorded and considered in a logistic regression analysis using ICH as end-point (internal study). CT was used as the golden standard. The score derived was validated with data from the next consecutive stroke patients and was compared with the three preexisting scores (external validation study).

Amongst 235 patients (119 males, mean age 70.6 +/- 11.2 years) of the internal study, 43 (18.3%) had ICH. Four independent correlates of ICH were identified and used for the derivation of the following integer-based scoring system: number of points=6 * (neurological deterioration within 3 h from admission) + 4 * (vomiting) + 4 * (WBC > 12 000) + 3 * (decreased level of consciousness). In the external study [168 patients, 85 males, mean age 70.2 +/- 10.8 years, 31 (18.5%) with ICH], when the cut-offs < or =3 points for CIF and > or =11 points for ICH were used, sensitivity, specificity, and positive and negative predictive values of the score for detection of stroke type were 97, 99, 97 and 99%, respectively; exceeding noticeably the three previously proposed systems.

The proposed model provides an easy to use tool for sufficiently accurate differentiation between haemorrhagic and nonhaemorrhagic stroke on the basis of information available to all physicians early after admission.