Diagnostic Errors Are Mostly Due to Clinical Judgement: Opportunity for AI
[Thursday, July 18, 2019] A recent study reported that almost 90% of diagnostic errors cases are attributed to errors in clinical judgement by physicians in outpatient settings. These findings have a strong implication for developers of AI algorithms to automate diagnosis aiming to reduce human errors. Common clinical judgement errors are delayed ordering of diagnostic tests, failure to review results of the diagnostic test in the context of other symptoms, not consulting with more experienced colleagues when faced with tough diagnostic results, or misinterpretation of test reports. Most common diagnostic errors were reported for vascular events, infections and cancers. In about 33% of the cases, the errors in diagnosis led to serious permanent damage or death. The study points to an important real-life aspect of clinical diagnosis where clinicians can and do make mistakes with serious consequences for the patients. Automated software-based diagnosis can help reduce the errors by providing physicians with better tools to address the common reasons for errors. For example, software can analysis multiple data from diverse sources much faster and completely than a physician under heavy workload in an outpatient or an emergency care setting. AI algorithms can be trained using test cases and get better over time with additional data to provide assistive suggestions to the physicians in-charge. Developers of AI algorithms should use this study to understand the areas of automation to include in their protocols and run tests to address these. Also, promoting tools as assistive rather than replacements for physicians will likely have a higher acceptance by the users. Since diagnostic errors are mostly used for malpractice law suites, the benefits of reduced malpractice cases could be an added inducement for use. |
|