Evaluating the Robustness of Adverse Drug Event Classification Models Using Templates

TL;DR

This paper introduces a template-based evaluation framework for ADE classification models, revealing that models with similar overall performance can differ significantly in specific capabilities like negation and sentiment detection.

Contribution

It presents a novel, detailed evaluation approach using hand-crafted templates to assess models on key ADE detection capabilities.

Findings

Models vary in capabilities despite similar overall performance

Templates reveal differences in temporal, negation, sentiment, and beneficial effect detection

Thorough evaluation is crucial for high-stakes medical NLP applications

Abstract

An adverse drug effect (ADE) is any harmful event resulting from medical drug treatment. Despite their importance, ADEs are often under-reported in official channels. Some research has therefore turned to detecting discussions of ADEs in social media. Impressive results have been achieved in various attempts to detect ADEs. In a high-stakes domain such as medicine, however, an in-depth evaluation of a model's abilities is crucial. We address the issue of thorough performance evaluation in English-language ADE detection with hand-crafted templates for four capabilities: Temporal order, negation, sentiment, and beneficial effect. We find that models with similar performance on held-out test sets have varying results on these capabilities.