When has service provision for transient ischaemic attack improved enough? A discrete event simulation economic modelling study

OBJECTIVES: The aim of this study was to examine the impact of transient ischaemic attack (TIA) service modification in two hospitals on costs and clinical outcomes. DESIGN: Discrete event simulation model using data from routine electronic health records from 2011. PARTICIPANTS: Patients with suspected TIA were followed from symptom onset to presentation, referral to specialist clinics, treatment and subsequent stroke. INTERVENTIONS: Included existing versus previous (less same day clinics) and hypothetical service reconfiguration (7-day service with less availability of clinics per day). OUTCOME MEASURES: The primary outcome of the model was the prevalence of major stroke after TIA. Secondary outcomes included service costs (including those of treating subsequent stroke) and time to treatment and attainment of national targets for service provision (proportion of high-risk patients (according to ABCD2 score) seen within 24 hours). RESULTS: The estimated costs of previous service provision for 490 patients (aged 74±12 years, 48.9% female and 23.6% high risk) per year at each site were £340 000 and £368 000, respectively. This resulted in 31% of high-risk patients seen within 24 hours of referral (47/150) with a median time from referral to clinic attendance/treatment of 1.15 days (IQR 0.93-2.88). The costs associated with the existing and hypothetical services decreased by £5000 at one site and increased £21 000 at the other site. Target attainment was improved to 79% (118/150). However, the median time to clinic attendance was only reduced to 0.85 days (IQR 0.17-0.99) and thus no appreciable impact on the modelled incidence of major stroke was observed (10.7 per year, 99% CI 10.5 to 10.9 (previous service) vs 10.6 per year, 99% CI 10.4 to 10.8 (existing service)). CONCLUSIONS: Reconfiguration of services for TIA is effective at increasing target attainment, but in services which are already working efficiently (treating patients within 1-2 days), it has little estimated impact on clinical outcomes and increased investment may not be worthwhile.