Consultation Checklists: Standardising the Human Evaluation of Medical Note Generation
This addresses the challenge of standardizing evaluations for medical note generation systems, which is crucial for reliable benchmarking in healthcare AI, though it is an incremental improvement focused on evaluation methodology rather than generation itself.
The paper tackles the problem of subjective and inconsistent human evaluation in automatic medical note generation by introducing Consultation Checklists as a common reference for assessments, resulting in improved inter-annotator agreement and better correlation of automatic metrics like ROUGE and BERTScore with human judgments.
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.