Incorporating unanswerable questions into EHR QA systems is crucial for testing the trustworthiness of a system, as providing non-existent responses can mislead doctors in their diagnoses. The EHRSQL dataset stands out as a promising benchmark because it is the only dataset that incorporates unanswerable questions in the EHR QA system alongside practical questions. However, in this work, we identify a data bias in these unanswerable questions; they can often be discerned simply by filtering with specific N-gram patterns. Such biases jeopardize the authenticity and reliability of QA system evaluations. To tackle this problem, we propose a simple debiasing method of adjusting the split between the validation and test sets to neutralize the undue influence of N-gram filtering. By experimenting on the MIMIC-III dataset, we demonstrate both the existing data bias in EHRSQL and the effectiveness of our data split strategy in mitigating this bias.
@article{arxiv.2405.01588,
title = {Towards Unbiased Evaluation of Detecting Unanswerable Questions in EHRSQL},
author = {Yongjin Yang and Sihyeon Kim and SangMook Kim and Gyubok Lee and Se-Young Yun and Edward Choi},
journal= {arXiv preprint arXiv:2405.01588},
year = {2024}
}