SAN FRANCISCO—Prior to the publication of validated prediction rules for stroke after transient ischemic attack (TIA), patients with the highest risk for stroke following TIA were not being recommended for hospitalization, according to findings presented at the 2007 International Stroke Conference.

“I think most stroke neurologists think that at least part of the decision [to hospitalize patients] … should be based on the short-term risk of subsequent stroke following TIA,” said S. Andrew Josephson, MD, Assistant Clinical Professor of Neurology at the University of California, San Francisco. “Patients who have very little or no risk of stroke after TIA probably don’t need to be in the hospital.”

However, he said that patients with a high risk for stroke following TIA would benefit from a brief evaluation, which would enable clinicians to determine the etiology of the stroke and to administer rapid treatment, such as thrombolysis, if needed, within the first 48 hours.

During the past decade, researchers have attempted to define risk of stroke after TIA through multiple studies. From this work, the California and ABCD scores have been established. The California score identifies five factors independently associated with risk for stroke within 90 days of TIA. These include age older than 60, presence of diabetes, symptom duration of 10 minutes or longer, weakness, and speech impairment. The ABCD score evaluates risk of stroke within seven days of TIA, using four factors: Age older than 60, systolic Blood pressure greater than 140 mm Hg and diastolic blood pressure greater than 90 mm Hg, Clinical features of TIA (unilateral weakness, speech disturbance), and Duration of symptoms. Just recently, the unified ABCD2 score, which comprises the ABCD score plus 1 point for diabetes, was described in the January 27 Lancet.

TESTING THE MODEL

In their study, Dr. Josephson and his colleagues examined which factors clinicians had used prior to the publication of the risk stratification model to determine whether to hospitalize patients with TIA. “If the factors used were identical, then these models simply confirm clinical practice and are not incredibly useful to the practicing clinician.”

The researchers examined the records of 1,707 patients who had experienced TIA and presented to the emergency departments of 16 hospitals within the Kaiser-Permanente Medical Care Plan. The timeline was based on a one-year period that ended in February 1998 (prior to publication of prediction rules). Patients who were hospitalized after experiencing a subsequent stroke during their emergency department visit were excluded from the analysis. The risk of subsequent stroke was stratified according to the California and ABCD scores, and the decision to hospitalize was correlated with the risk scores using Spearman’s rho.

The results indicated that 243 patients with TIA (14%) were hospitalized. Hospitalization was weakly correlated with the California score (R2 = 0.028); 18% of patients hospitalized had no risk of stroke at seven days, and 28% had a 14% risk of stroke at seven days. There was also a weak correlation between hospitalization and ABCD score (R2 = 0.014); 16% of those hospitalized had no risk of stroke at seven days, and 22% had an 8% risk of stroke at seven days.

DEFINING FACTORS

A logistic regression model was developed using factors associated with hospitalization in a univariate analysis. Dr. Josephson reported that seven variables were independently ­associated with the decision to hospitalize following TIA: history of TIA (odds ratio [OR], 2.9), speech impairment (OR, 1.9), weakness (OR, 1.8), history of atrial fibrillation (OR, 2.1), gait disturbance (OR, 1.6), symptoms on arrival at the emergency department (OR, 1.5), and use of ticlopidine at emergency department entry (OR, 2.1).

According to Dr. Josephson, when these seven variables were compared with the ABCD2 criteria, several risk factors for stroke—age, blood pressure, and diabetes—were not accounted for as independent factors leading to the decision to hospitalize TIA patients. “This finding emphasizes the clinical utility of these prediction models and argues for their widespread dissemination among neurologists, emergency department physicians, and other health care workers evaluating patients following TIA,” he said.

Dr. Josephson noted that “future research will include [a focus] on current practices since the publication of these risk models, to assess their impact. In addition, research that focuses on which TIA patients warrant admission is important for patient care as well as allocation of health care resources.”

NR