MODELING PERFORMANCE OF IMAGING : A NEW PROPOSITION FOR PERFORMANCE APPRAISAL

Purpose:   Evaluation of carotid stenosis degree by imaging technologies is mandatory prior to eventual surgery, in order to assess patient with a high stroke risk. The Carmedas study compared Doppler Ultrasound (DUS), Magnetic Resonance Angiography (MRA) and Computed Tomography Angiography (CTA) with Digital Subtraction Angiography (DSA), i.e. the gold standard, but invasive, and conclude on best accuracy with association of non-invasive DUS, CTA and MRA. However, literature gives separate figures of techniques performance by their interval of sensibility and specificity. The purpose of our study is to conduct a medico-economic investigation of different imaging strategies.

Method:   As decision of surgery is a consequence of imaging, we built a decision tree and established 11 imaging strategies (4 with one exam, 3 with two and decision made on the second, 4 with three and decision made on two concordant). The performance was the rate of stroke risk reduction (per 1000). We evaluate the potential impact of addition of uncertainty using a Bayesian approach. We built a design of experiment  testing variation of sensibility and specificity of different techniques in intervals given by literature. The model was applied to symptomatic patients, which disease prevalence known. The simulation gives worst, average, best performance (for 1000) and an average cost-effectiveness ratio in € per avoided risk of stroke.

Result:   Despite his status, DSA alone [23,05/29,23/37,1/9851] never appeared as the best solution. DUS alone [20,49/25,71/32,1/3300] has the best CE, but a poor performance. MRA alone has the best performance [26,06/33,26/38,81/8497], but the CE is high. Similarly to the Carmedas  study, with a low CE ratio , DUS-DSA-MRA [19,79/29,96/40,58/5475] appears with the highest best performance but not for the average and lowest performance.

Conclusion: Our study demonstrates that the best CE ratios goes with imaging strategies that are not the best in terms of medical performance. The model illustrates also the different perspectives of actors interfering with medical decision. Doctors prefers the solution that give an individual to be in position of have a chance to get the best performance. But as uncertainty occurs, the best solution for the whole population of symptomatic patient would be the best average ratio. The economic rationality gives a third choice, but have to be link to the consequences of stroke, in terms of disabilities .