Contextual Evaluation of Large Language Models for Classifying Tropical and Infectious Diseases

TL;DR

This study evaluates large language models' effectiveness in classifying tropical and infectious diseases, emphasizing the importance of contextual information and introducing a new research tool for health-related LLM analysis.

Contribution

It expands the TRINDs dataset with demographic and semantic data, compares various LLMs, and develops a prototype tool to analyze context effects on medical LLM outputs.

Findings

Context improves LLM classification accuracy.

Demographic and location data enhance model responses.

The TRINDs-LM prototype aids in understanding context impact.

Abstract

While large language models (LLMs) have shown promise for medical question answering, there is limited work focused on tropical and infectious disease-specific exploration. We build on an opensource tropical and infectious diseases (TRINDs) dataset, expanding it to include demographic and semantic clinical and consumer augmentations yielding 11000+ prompts. We evaluate LLM performance on these, comparing generalist and medical LLMs, as well as LLM outcomes to human experts. We demonstrate through systematic experimentation, the benefit of contextual information such as demographics, location, gender, risk factors for optimal LLM response. Finally we develop a prototype of TRINDs-LM, a research tool that provides a playground to navigate how context impacts LLM outputs for health.