Risk-based selection for carotid revascularisation using the IMPROVE score versus standard care in symptomatic
Kelly P H Nies1,2, Bram Ramaekers3, Juul Bierens1,2
1Department of Radiology and Nuclear Medicine, Maastricht University Medical Centre+, Maastricht, LI, Netherlands.
BMJ Open
|May 28, 2026
Summary
The IMPROVE model significantly reduces ipsilateral ischaemic strokes and deaths in carotid disease patients by 34.5%. This clinical prediction model also lowers healthcare costs, proving more cost-effective than usual care.
Area of Science:
- Neurology
- Health Economics
- Clinical Decision Support
Background:
- A validated clinical prediction model, IMPROVE, exists for ipsilateral ischaemic stroke risk in symptomatic carotid disease patients.
- Current care-as-usual (CAU) involves optimal medical treatment (OMT) alone or with carotid endarterectomy (CEA).
Purpose of the Study:
- To evaluate the cost-effectiveness of using the IMPROVE model for triage compared to CAU.
- To determine optimal triage thresholds for carotid endarterectomy based on IMPROVE risk scores.
Main Methods:
- A decision-analytic model was developed using data from 678 patients with carotid disease and recent ischaemic events.
- Patients were stratified based on carotid stenosis percentage and 3-year stroke risk thresholds (IMPROVE vs. CAU).
- Monte Carlo simulations were used for probabilistic analyses over 3-year and lifetime horizons.
Main Results:
- IMPROVE-based triage reduced ipsilateral ischaemic strokes and perioperative events by 34.5% over 3 years (2.8% vs. 4.3%).
- Revascularization rates decreased by 20%, with a slight increase in Quality-Adjusted Life Years.
- Societal costs were reduced by €1441/patient (3-year) and €6101/patient (lifetime) with IMPROVE triage.
Conclusions:
- Triage using the IMPROVE model significantly prevents strokes and perioperative events in symptomatic carotid disease patients.
- IMPROVE-based triage is a cost-effective strategy, reducing both adverse events and healthcare expenditures.
- Clinical trials are recommended to validate these findings.
