Assessing Concordance With Watson for Oncology, a Cognitive Computing Decision Support System for Colon Cancer Treatment in Korea.

Abstract

IBM Watson for Oncology (WFO) is a clinical decision-support computing system that provides oncologists with evidence-based treatment recommendations for a variety of cancer diagnoses. The evidence-based supported treatment recommendations are presented in three categories: Recommended, representing the Memorial Sloan Kettering Cancer Center (MSKCC) preferred approach; For Consideration, evidence-based alternative treatments; and Not Recommended, alternative therapies that may be unacceptable. We examined the absolute concordance of treatment options with that of the recommendations of a multidisciplinary team of oncologists from Gachon University, Gil Medical Centre, Incheon, South Korea. We enrolled 656 patients with stage II, III, and IV colon cancer between 2009 and 2016. Cases were processed using WFO and, using retrospective clinical data, outputs were compared with the actual treatment the patient received. Absolute concordance was defined as an alignment of recommendation in the Recommended MSKCC preferred-approach category. Treatment recommendations that were represented in the For Consideration category were not the focus of this study. The absolute concordance between the WFO-derived MSKCC preferred approach and Gil Medical Centre treatment recommendations was 48.9%. The percentage of cases found to be acceptable was 65.8% (432 of 656) and the stage-specific concordance rate was 32.5% for patients with stage II disease who had risk factors and 58.8% for patients with stage III disease. Patients 70 years of age and older had a concordance rate of only 20.2%, whereas younger patients had a concordance rate of 63.8% ( P = .0001). The main reasons attributed to the low concordance rate were age, reimbursement plan, omitting chemotherapy after liver resection, and not recommending biologic agents (ie, cetuximab and bevacizumab).