32 Diagnostic missed opportunities for diagnosis – addressing the unknown unknown with a safety net team

<h3>Background</h3> Medical errors are the third leading cause of death in the US, with delayed diagnoses causing significant harm. Missed opportunities for diagnoses (MODs) are common in diagnostic imaging, particularly when incidental findings require follow-up. <h3>Objectives</h3> To implement a centralized solution identifying and mitigating potential MODs in diagnostic imaging, focusing on ”delayed surveillance” recommendations from radiologist interpretations. <h3>Methods</h3> Parkland Health’s approach: a) Create a model to flag diagnostic imaging ”delayed surveillance” recommendations. b) Operationalize this flag in the Electronic Health Record. c) Establish a centralized team to monitor identified MODs. d) Develop an Artificial Intelligence Large Language Model (AI LLM) solution to capture MODs, ensuring comprehensive identification and reducing radiologists’ workload. <h3>Results</h3> Historically, Parkland identified MODs in 1.7% of CT or MRI studies; 17% of these MODs were not addressed. By capturing MODs in the Electronic Health Record (HER) and creating a centralized monitoring team, Parkland ensured potential diagnoses are not overlooked. Using the radiologist interpretation to detect recommended follow-up visits, the AI LLM solution demonstrated 94.8% accuracy, with a positive predictive value of 0.989 and negative predictive value of 0.944. <h3>Conclusions and Implications</h3> Parkland’s initiative demonstrates an effective approach to reducing MODs in diagnostic imaging. The implemented system, including the highly accurate AI solution, addresses a significant patient safety concern and can improve outcomes while reducing radiologist burden. This model allows for replication by other healthcare organizations, suggesting broader implications for improving diagnostic accuracy and patient safety across the healthcare system.