Transforming diagnostic efficacy in Breast Screening using an optimized double reading strategy

Double reading of mammography images, where two radiologists independently read the same patient images is current practice in the Breast Screening service in Australia. Compared with single reads, the double read strategy improves outputs from breast screening programs, e.g. if a woman’s cancer is missed by one radiologist, the second radiologist may pick it up by thus leading to improved cancer detection rates. Unfortunately however double reading only works if paired radiologists complement each other, i.e. they do not make the same type of error, therefore sophisticated matching is required. Such sophisticated matching has never been implemented due to historic reasons: since mammographic films can only be physically stored at one location (often where they are produced), matching of radiologists has been performed out of geographic convenience rather than maximizing diagnostic potential. However in 2014, almost all mammographic images are produced, transmitted and stored electronically (digitally) meaning that a woman’s images can be read by anyone, anywhere (regardless of where they are produced). This means that radiologists can be matched on reading characteristics and individual errors rather than geography.

By bringing together scientists from Australia and the US we will:

• establish maximum efficacy obtainable when radiologists are optimally paired;
• define individual reading patterns linked to specific error types so optimal pairings are established prospectively.

Streamlining the double reading strategy, will result in substantial benefits for the 1.6 million Australian women who participate in breast screening over a two-year cycle, as well as the millions who undergo mammography world-wide each year.