Summary
A prominent multinational medical device manufacturer faced significant time and cost challenges in manually generating literature summaries for Clinical Evaluation Reports (CERs), which are required for regulatory compliance. One of the primary activities in CER report creation is literature summarization, which consumes a significant portion of the total CER generation time. Our client was manually generating summaries of the articles, which was very time-consuming and incurred significant costs and efforts on the part of the medical writers performing the task.
The Challenge
Tackling time, compliance and precision challenges in literature summarization
Clinical Evaluation Report (CER) involves an average of about 70 articles per device for clinical evaluation. Each article takes about 1.5 hours for manual summarization. Overall, literature summarization consumes about 25% of the total time and effort of the CER report generation process. The main challenges for the customer were to:
- Reduce the literature summarization time significantly
- Ensure that all medical writers generate summaries in the required format as mandated by the regulatory requirements and get a measure of how complete the summaries are
- Ensure that all the acceptance criteria related to the summarized article were met during the manual summarization process
The Objective
Precision in summarization with seamless GenAI automation
To address the client's challenges, the following objectives were set in place as a starting point:
- Automating the pre-processing of literature articles so that the complete articles are available for summarization without any significant loss of information
- Reviewing the manual-generated summaries and identifying the key attributes needed for creating a comprehensive automated summary
- Identifying 16 acceptance criteria that need to be available in the summary so that it can be considered complete
- Interacting with key medical writers on the team to assess the details that must be generated for each of the attributes in the summary so that no important information is missed
- Identifying and isolating different formats of documents, size limits and other aspects of the documents under summarization so that all types of documents are covered under the automated summarization process
The Solution
Empowering precision and efficiency with HCLTech AI Force
Our AI Force support platform was brought in to provide a GenAI-based automated literature summarization solution. It had the following solution tenets:
- Automatic extraction of text from various document formats before summarization, ensuring that no significant content is lost
- A novel approach to handling the summarization of very large-sized documents by overcoming the size limit enforced by GenAI models to consume large-size documents
- Providing flexible (configurable) control to the medical writers to extract information as per their requirements, which ensures that the summaries contain the required level of detail for the acceptance criteria
- Providing dual formats of summary — a comprehensive summary of the entire document and section-specific summaries to help medical writers assess the coverage of all the acceptance criteria
The Impact
Effort reduction in summarization efforts and improvement in efficiency
By implementing the comprehensive solution at scale, we empowered the client to achieve remarkable outcomes:
- Summarization time was reduced by 44%, with a range of 30%-55% effort reduction across different articles
- 60%-80% of the acceptance criteria was met for summarization
- Deployed across multiple projects, with 100+ medical writers using the solution effectively
