Using the Recommendations Inferred by a Decision Support System to Compare Breast Cancer Clinical Practice Guidelines

Numerous studies have reported that inconsistencies exist between clinical practice guidelines (CPGs) elaborated on the same topic and at the same date. These results are usually established from the analyses handled on narrative CPGs or on their semi-structured version. In the context of the European-funded DESIREE project, we have developed a guideline-based decision support system embedding various contemporary CPGs on breast cancer management. We have run the GL-DSS on a sample of 571 retrospective clinical cases and specifically assessed the level of inconsistencies between the recommendations issued by the US National Comprehensive Cancer Network (NCCN) Guidelines for Breast Cancer and those issued by Assistance Publique - Hôpitaux de Paris (AP-HP, France) CPGs. We proposed a typology with six different situations from total incompatibility to complete identity as a result of the comparison of NCCN and AP-HP CPGs for each clinical case. It was interesting to observe that we got 38% of inconsistencies with 3% of total incompatibility, and 62% of similarity with 0% of complete identity. The silence of one CPG was resolved by the other CPG providing recommendations in 21% of the cases.