Cancer patients' messages about radiology/pathology reports: Insights for AI.

Abstract

1570 Background: Cancer patients often use portals to view results prior to discussing with physicians, leading to messages with questions or concerns. 1 These messages vary widely in content and urgency, creating challenges for healthcare providers to respond effectively. 2 Categorizing and triaging these messages through AI-enhanced tools could streamline communication and improve patient care and satisfaction. Methods: This study assessed common themes in 1 week (April 1-8, 2023) of patients’ portal messages about “rapidly read” pathology and radiology reports (viewed by patients within 6 hours of posting to the portal, as a proxy for viewing before discussing with physicians) at Memorial Sloan Kettering Cancer Center in New York City. Results: Five notable themes emerged across a total of 48 messages about rapidly read radiology and pathology results: Interpretation (24/48, 50%): Half of the messages contained questions like, “What does this mean?” Patients sought explanations of pathology and radiology findings, reflecting a need for clear, accessible interpretations. Implications (14/48, 29%): With questions like, "What are the next steps?” patients often asked how findings might alter treatment plans, highlighting a need for guidance on the care implications of their reports. Concern (5/48, 10%): Some patients expressed worry or pessimism about pathology and radiology reports: "I am very worried" and “ Maybe it’s time to give up. ” Such statements indicated a need for supportive communication. Relief (3/48, 6%): In other messages, patients shared positive emotions regarding favorable results – “It is a huge weight off my mind.” These responses offer clinicians opportunities to reinforce patient satisfaction. Errors/Omissions (3/48, 6%): Occasionally, patients perceived errors or omissions in their reports -- "The radiologist totally misread the size of the lesion" -- which impacted their trust in the information. Addressing these concerns promptly can help strengthen the patient-provider relationship. Conclusions: This novel study highlights opportunities for AI-enhanced tools to triage messages and facilitate timely, effective responses. This study found common themes in patients’ diverse questions about rapidly read pathology and radiology reports. By implementing AI to categorize and triage message patterns, providers could support patients more efficiently. Methods in development could be used to classify the message content. 3 For instance, AI-driven natural language processing tools could recognize queries related to "What does this mean?" and offer clear, accessible explanations of medical terms. Similarly, AI could be trained to flag high-priority messages based on distress signals, ensuring that these messages are addressed swiftly. Implementing such AI-based solutions could help meet patients' immediate needs while they await conversations with their providers.

Ähnliche Arbeiten