Tomas Novotny1, Raymond Bond2, Irena Andrsova3, Lumir Koc3, Martina Sisakova3, Dewar Finlay2, Daniel Guldenring2, Jindrich Spinar3, Marek Malik4. 1. Department of Internal Medicine and Cardiology, University Hospital Brno and Faculty of Medicine of Masaryk University, Brno, Czechia. Electronic address: novotny-t@seznam.cz. 2. Faculty of Computing and Engineering, Ulster University, Newtownabbey, United Kingdom. 3. Department of Internal Medicine and Cardiology, University Hospital Brno and Faculty of Medicine of Masaryk University, Brno, Czechia. 4. St. Paul's Cardiac Electrophysiology and Imperial College, London, United Kingdom.
Abstract
INTRODUCTION: Most contemporary 12-lead electrocardiogram (ECG) devices offer computerized diagnostic proposals. The reliability of these automated diagnoses is limited. It has been suggested that incorrect computer advice can influence physician decision-making. This study analyzed the role of diagnostic proposals in the decision process by a group of fellows of cardiology and other internal medicine subspecialties. MATERIALS AND METHODS: A set of 100 clinical 12-lead ECG tracings was selected covering both normal cases and common abnormalities. A team of 15 junior Cardiology Fellows and 15 Non-Cardiology Fellows interpreted the ECGs in 3 phases: without any diagnostic proposal, with a single diagnostic proposal (half of them intentionally incorrect), and with four diagnostic proposals (only one of them being correct) for each ECG. Self-rated confidence of each interpretation was collected. RESULTS: Availability of diagnostic proposals significantly increased the diagnostic accuracy (p<0.001). Nevertheless, in case of a single proposal (either correct or incorrect) the increase of accuracy was present in interpretations with correct diagnostic proposals, while the accuracy was substantially reduced with incorrect proposals. Confidence levels poorly correlated with interpretation scores (rho≈2, p<0.001). Logistic regression showed that an interpreter is most likely to be correct when the ECG offers a correct diagnostic proposal (OR=10.87) or multiple proposals (OR=4.43). CONCLUSION: Diagnostic proposals affect the diagnostic accuracy of ECG interpretations. The accuracy is significantly influenced especially when a single diagnostic proposal (either correct or incorrect) is provided. The study suggests that the presentation of multiple computerized diagnoses is likely to improve the diagnostic accuracy of interpreters.
INTRODUCTION: Most contemporary 12-lead electrocardiogram (ECG) devices offer computerized diagnostic proposals. The reliability of these automated diagnoses is limited. It has been suggested that incorrect computer advice can influence physician decision-making. This study analyzed the role of diagnostic proposals in the decision process by a group of fellows of cardiology and other internal medicine subspecialties. MATERIALS AND METHODS: A set of 100 clinical 12-lead ECG tracings was selected covering both normal cases and common abnormalities. A team of 15 junior Cardiology Fellows and 15 Non-Cardiology Fellows interpreted the ECGs in 3 phases: without any diagnostic proposal, with a single diagnostic proposal (half of them intentionally incorrect), and with four diagnostic proposals (only one of them being correct) for each ECG. Self-rated confidence of each interpretation was collected. RESULTS: Availability of diagnostic proposals significantly increased the diagnostic accuracy (p<0.001). Nevertheless, in case of a single proposal (either correct or incorrect) the increase of accuracy was present in interpretations with correct diagnostic proposals, while the accuracy was substantially reduced with incorrect proposals. Confidence levels poorly correlated with interpretation scores (rho≈2, p<0.001). Logistic regression showed that an interpreter is most likely to be correct when the ECG offers a correct diagnostic proposal (OR=10.87) or multiple proposals (OR=4.43). CONCLUSION: Diagnostic proposals affect the diagnostic accuracy of ECG interpretations. The accuracy is significantly influenced especially when a single diagnostic proposal (either correct or incorrect) is provided. The study suggests that the presentation of multiple computerized diagnoses is likely to improve the diagnostic accuracy of interpreters.
Authors: Stephen W Smith; Jeremy Rapin; Jia Li; Yann Fleureau; William Fennell; Brooks M Walsh; Arnaud Rosier; Laurent Fiorina; Christophe Gardella Journal: Int J Cardiol Heart Vasc Date: 2019-09-08
Authors: Anthony H Kashou; Wei-Yin Ko; Zachi I Attia; Michal S Cohen; Paul A Friedman; Peter A Noseworthy Journal: Cardiovasc Digit Health J Date: 2020-09-08
Authors: Anthony H Kashou; Siva K Mulpuru; Abhishek J Deshmukh; Wei-Yin Ko; Zachi I Attia; Rickey E Carter; Paul A Friedman; Peter A Noseworthy Journal: Cardiovasc Digit Health J Date: 2021-05-05