Soft-prompt tuning to predict lung cancer using primary care free-text Dutch medical notes

TL;DR

This study compares NLP methods, including soft prompt-tuning and static word embeddings, for early lung cancer prediction from Dutch primary care notes, addressing class imbalance and small data scenarios.

Contribution

It demonstrates that soft prompt-tuning is an efficient alternative to fine-tuning and compares the robustness of PLMs and WEMs in imbalanced and limited data settings.

Findings

Soft prompt-tuning is an efficient alternative to fine-tuning.

PLMs have better discrimination but worse calibration in imbalanced settings.

Results on small patient samples are mixed with no clear advantage.

Abstract

We investigate different natural language processing (NLP) approaches based on contextualised word representations for the problem of early prediction of lung cancer using free-text patient medical notes of Dutch primary care physicians. Because lung cancer has a low prevalence in primary care, we also address the problem of classification under highly imbalanced classes. Specifically, we use large Transformer-based pretrained language models (PLMs) and investigate: 1) how \textit{soft prompt-tuning} -- an NLP technique used to adapt PLMs using small amounts of training data -- compares to standard model fine-tuning; 2) whether simpler static word embedding models (WEMs) can be more robust compared to PLMs in highly imbalanced settings; and 3) how models fare when trained on notes from a small number of patients. We find that 1) soft-prompt tuning is an efficient alternative to standard model fine-tuning; 2) PLMs show better discrimination but worse calibration compared to simpler static word embedding models as the classification problem becomes more imbalanced; and 3) results when training models on small number of patients are mixed and show no clear differences between PLMs and WEMs. All our code is available open source in \url{https://bitbucket.org/aumc-kik/prompt_tuning_cancer_prediction/}.