LO87: Use of a clinical prediction rule would lead to more effective CTA utilization for urgent brain imaging of suspected TIA/mild stroke in the emergency department

Abstract

Introduction: Canadian stroke best practice guidelines recommend patients suspected of Acute Cerebrovascular Syndrome (ACVS) receive urgent brain imaging, preferably CTA. Yet, high requisition rates for non-ACVS patients overburdens limited radiological resources. We hypothesize that our clinical prediction rule (CPR) previously developed for diagnosis of ACVS in the emergency department (ED), and which incorporates Canadian guidelines, could improve CTA utilization. Methods: Our data consists of records for 1978 ED-referred patients to our TIA clinic in Victoria, BC from 2015-2016. Clinic referral forms captured all data needed for the CPR. For patients who received CTA, orders were placed in the ED or at the TIA clinic upon arrival. We use McNemar’s test to compare the sensitivity (sens) and specificity (spec) of our CPR vs. the baseline CTA orders for identifying ACVS. Results: Our sample (49.5% male, 60.6% ACVS) has a mean age of 70.9±13.6 yrs. Clinicians ordered 1190 CTAs (baseline) for these patients (60%). Where CTA was ordered, 65% of patients (n=768) were diagnosed as ACVS. To evaluate our CPR, predicted probabilities of ACVS were computed using the ED referral data. Those patients with probabilities greater than the decision threshold and presenting with at least one focal neurological deficit clinically symptomatic of ACVS were flagged as would have received a CTA. Our CPR would have ordered 1208 CTAs (vs. 1190 baseline). Where CTA would have been ordered, 74% of patients (n=893) had an ACVS diagnosis. This is a significantly improved performance over baseline (sens 74.5% vs. 64.1%, p<0.001; spec 59.6% vs. 45.9%, p<0.001). Specifically, the CPR would have ordered an additional 18 CTAs over the 2-yr period, while simultaneously increasing the number of imaged-ACVS patients by 125 with imaging 107 fewer non-ACVS patients. Conclusion: Using ED physician referral data, our CPR demonstrates significantly higher sensitivity and specificity for CTA imaging of ACVS patients than baseline CTA utilization. Moreover, our CPR would assist ED physicians to apply and practice the Canadian stroke best practice guidelines. ED physician use of our CPR would increase the number of ACVS patients receiving CTA imaging before ED discharge (rather than later at TIA clinics), and ultimately reduce the burden of false-positives on radiological departments.