loading page

CliniXPromt: Enhancing the Comprehensibility of Electronic Health Records using GPT-3 and Chain of Thought
  • +1
  • Hirak Mazumdar,
  • Suparna Das,
  • Sathvik MSVPJ,
  • Kamil Reza Khondakar
Hirak Mazumdar
Woxsen School of Business

Corresponding Author:[email protected]

Author Profile
Suparna Das
Woxsen School of Business
Author Profile
Sathvik MSVPJ
Indian Institute of Technology Dharwad
Author Profile
Kamil Reza Khondakar
Woxsen School of Business
Author Profile


This paper presents an innovative method for enhance the comprehensibility of Electronic Health Records (EHRs), making it accessible to individuals without specialized clinical knowledge. Our approach entails predicting medical professionals’ impressions, identifying intricate medical terminol- ogy, and clarifying these complex terms. To achieve this, we fine-tuned GPT-3 for predicting doctors’ impressions and integrated the Chain Of Thought (COT) prompting technique to identify and elucidate intricate medical terms. The assessment was conducted using Rouge scores and cosine similarity scores. The outcomes reveal that our proposed approach yields a cosine similarity score surpassing 75, indicative of the model’s exceptional performance. The comparative analysis demonstrates the superiority of our approach concerning doctors’ impressions, detection of complex terminology, and provision of explanations. Furthermore, this work is pioneering in addressing and resolving intricate terminology in EHRs, marking a novel contribution to the field.
14 Mar 2024Assigned to Editor
14 Mar 2024Submission Checks Completed
31 Mar 2024Reviewer(s) Assigned
08 Apr 2024Review(s) Completed, Editorial Evaluation Pending