LLM-Assisted Systematic Literature Review System


Description

A five-module system using large language models to conduct systematic literature reviews in health technology assessment (HTA) submissions. The system includes literature search query setup, study protocol setup using PICOs criteria, LLM-assisted abstract screening, LLM-assisted data extraction, and data summarization with human-in-the-loop design.

Market positioning

Academic research tool for systematic literature reviews in healthcare

Use cases

Medical researchers, clinicians, health technology assessment professionals, and systematic review authors

Features

High performance with 90% sensitivity and 89% accuracy
Substantial agreement with human reviewers (Cohen's κ of 0.71)
Human-in-the-loop design for quality control
Real-time criteria adjustment capability
Validated across multiple medical domains
Comprehensive five-module approach
Integration with established medical databases

Pros and cons

Based on: (AI summary)

Pros

  • High performance with 90% sensitivity and 89% accuracy
  • Substantial agreement with human reviewers (Cohen's κ of 0.71)
  • Human-in-the-loop design for quality control
  • Real-time criteria adjustment capability
  • Validated across multiple medical domains

Cons

  • Still requires human oversight and validation
  • Performance may vary across different medical specialties
  • Dependent on quality of PubMed abstracts
  • May have limitations in handling complex medical terminology
  • Requires technical expertise to implement and maintain