2024 World Conference on Lung Cancer (WCLC) - Posters-P4.17F.03 Utilizing AI for Automated Data Entry and Analysis to Pre-Screen Lung Cancer Clinical Trial Candidates

Pdf Summary

The study evaluates an AI-driven automated data entry (ADE) system for selecting lung cancer patients for clinical trials, compared to manual data entry (MDE). Conducted at Gustave Roussy, patients with thoracic cancer were considered, using data sourced from unstructured medical letters and various reports for the AI input, complemented by structured MDE in RedCap.<br /><br />The AI system aimed to enhance accuracy, efficiency, and scalability in clinical trial pre-screening processes by automatically structuring patient data such as demographics, disease characteristics, and treatment history. With data consisting of 1344 patients and 160 variables, totaling 282,445 data points, discordances were observed at varying rates for both methods.<br /><br />MDE often had access to more detailed imaging or pathology reports unavailable to ADE. ADE errors were primarily due to gaps in medical notes, a limitation that hindered complete accuracy. Despite this, ADE demonstrated potential by identifying clinical trial candidates with over 80% accuracy when critical variables were correctly entered simultaneously. It exhibited high performance in demographics, histology, molecular alterations, and comorbidities. However, it faced challenges in handling missing data due to absent information, suggesting scope for further improvement.<br /><br />The study found that ADE can significantly streamline patient selection processes in clinical oncology, maintaining over 80% correctness. It underscores the potential of AI in reducing the time physicians spend manually identifying eligible trial patients, thus accelerating inclusion times and increasing trial volumes. Moreover, ADE's performance is promising for automating complex data entry tasks in clinical settings, though limitations remain due to incomplete data access. With focused improvements, AED could enhance patient selection's speed and accuracy, making AI a valuable tool in modern medical research.

Asset Subtitle

Mihaela Aldea

Meta Tag

Speaker Mihaela Aldea

Topic Global Health, Health Services & Health Economics

Keywords

AI-driven data entry

lung cancer trials

automated data entry

clinical trial pre-screening

Gustave Roussy

medical data structuring

clinical oncology

patient selection

data accuracy

AI in healthcare