On the role of arbitration of discordant double readings of screening mammography: experience from two Italian programmes

Abstract

Purpose: The authors sought to assess the role of arbitration by a third reader of discordant double readings to reduce the rate of recalls to diagnostic assessment.

Materials and methods: A consecutive series of 7,660 double readings of screening examinations were considered. Discordant recalls were arbitrated by an expert reader (negative/positive). Diagnostic assessment was performed irrespective of arbitration results, and its outcome was used as reference standard for the study purpose. Assuming that negative arbitration would deny recall, its impact was assessed in terms of reduced recall rate and reduced cancer detection rate. Cost analysis of introducing arbitration was performed according to these results.

Results: Recalls at double reading were 528 (6.8%), of which 230 (43.5%) were concordant and 298 (56.5%) were discordant. The latter underwent arbitration, which was negative in 216 (72.4%) and positive in 82 (27.6%) cases, respectively. Overall, 49 cancers were detected (6.39 ‰ screened, 9.2% recalled): 43 cancers were detected among concordant (5.6 ‰ screened, 18.6% concordant) and six among discordant recalls (0.7 ‰ screened, 2.0% discordant). Six cancers were observed among arbitrated cases: five (6%) in positive and one (4.6 ‰) in negative arbitrations. Negative arbitration would have spared 216 assessment procedures (2.8% absolute, 40.9% relative reduction of recall rate) while missing one cancer case (0.13 ‰ absolute, 2.0% relative reduction of cancer detection rate). Arbitration cost was 74 euro, whereas 216 spared assessment procedures would have cost 14,558.4-23,346 euro.

Conclusions: Arbitration is a cost-effective procedure that could be employed as a first measure to counterbalance excess recall rate observed in a double-reading scenario.