Consultation Checklists: Standardising the Human Evaluation of Medical Note Generation

TL;DR

This paper introduces a standardized evaluation protocol for medical note generation using Consultation Checklists, which improves inter-annotator agreement and correlates better with human judgments than traditional metrics.

Contribution

The paper proposes a novel evaluation protocol grounded in Consultation Checklists to enhance objectivity and consistency in assessing medical note generation systems.

Findings

High inter-annotator agreement achieved with the protocol

Consultation Checklists improve correlation of automatic metrics with human judgments

Protocol facilitates more reliable evaluation of medical note generation systems

Abstract

Evaluating automatically generated text is generally hard due to the inherently subjective nature of many aspects of the output quality. This difficulty is compounded in automatic consultation note generation by differing opinions between medical experts both about which patient statements should be included in generated notes and about their respective importance in arriving at a diagnosis. Previous real-world evaluations of note-generation systems saw substantial disagreement between expert evaluators. In this paper we propose a protocol that aims to increase objectivity by grounding evaluations in Consultation Checklists, which are created in a preliminary step and then used as a common point of reference during quality assessment. We observed good levels of inter-annotator agreement in a first evaluation study using the protocol; further, using Consultation Checklists produced in the study as reference for automatic metrics such as ROUGE or BERTScore improves their correlation with human judgements compared to using the original human note.