Evaluation and practical application of prompt-driven ChatGPTs for EMR generation

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support

Background:

• Lung nodule screening generates extensive patient data requiring efficient record management.
• Traditional electronic medical record (EMR) generation can be time-consuming, impacting clinical workflow.
• Prompt-driven artificial intelligence (AI) offers potential solutions for automating clinical documentation.

Purpose of the Study:

• To evaluate the efficacy of prompt engineering in optimizing ChatGPT for generating EMRs.
• To assess ChatGPT's performance in converting patient-provider verbal consultations into structured EMRs.
• To explore the integration of AI-generated EMRs into patient-facing tools for pre-visit information gathering.

Main Methods:

• Utilized prompt engineering techniques to guide ChatGPT in EMR generation.
• Assessed ChatGPT's accuracy and completeness in creating EMRs from transcribed verbal consultations.
• Developed and tested prototype tools (e.g., WeChat mini-programs) for patient pre-visit data input.

Main Results:

• Prompt engineering significantly improved ChatGPT's ability to generate relevant and accurate EMRs.
• The AI model demonstrated proficiency in extracting key information from patient-provider dialogues.
• Integrated tools showed potential for streamlining patient data collection prior to clinical encounters.

Conclusions:

• Prompt-optimized ChatGPT is a viable tool for enhancing EMR generation in lung nodule screening.
• This AI application can improve clinical workflow efficiency and diagnostic accuracy.
• Patient-accessible tools leveraging this technology can empower individuals and optimize healthcare delivery.