Accuracy of a clinical decision support system in identifying patients at risk of cancer in primary care.

Abstract

1562 Background: 85% of patients diagnosed with cancer are symptomatic at presentation. Yet a delay in diagnosis (e.g. missing a presentation) can result in late stage detection and poorer outcomes. Clinical decision support systems (CDSS) have been used in other fields to calculate risk and support standardization of care. However, little has been reported about successful implementation of CDSS with determining cancer risk. C the Signs is a CDSS which uses artificial intelligence, mapped with the latest evidence and guidance, to assist healthcare professionals in identifying patients at risk of cancer and determining the most appropriate pathway for diagnosis. The aim of this study is to assess the accuracy of C the Signs in a real-world setting. Methods: A retrospective observational study was conducted in the National Health Service in England utilizing all patients risk assessed for cancer through the Electronic Healthcare Record (EHR) integrated C the Signs platform between 1st January 2021 and 31st August 2022. A total of 118,677 patients were risk assessed. All patients were followed up for 6 months post risk assessment to determine outcome. A sub-analysis was performed on the patients who were flagged at risk of cancer through C the Signs and subsequently diagnosed with cancer to establish the CDSS’s cancer origin accuracy (the ability for the system to accurately predict which cancer the patient is at risk of). Results: 7,295 patients were diagnosed with cancer within 6 months of a C the Signs risk assessment. 7,056 patients were successfully identified at risk of cancer by the CDSS with a sensitivity of 96.7%. There were 103,168 false positives with a specificity of 7.37%. Of the 8,453 patients who were identified as not at risk of cancer using the CDSS, 239 had a confirmed cancer diagnosis within 6 months of risk assessment (with a negative predictive value 97.2%). The cancer origin accuracy of C the Signs was 85.6% across all cancers. Conclusions: A cancer CDSS like C the Signs can be used to accurately identify patients at risk of cancer. More research is needed to understand the specificity and false positive rate, and the impact of identifying these patients with non-cancer disease.