Evaluation of Transfer Learning for Adverse Drug Event (ADE) and Medication Entity Extraction

Narayanan, Sankaran; Mannam, Kaivalya; Sunitha, R; Rangan, P. Venkat

doi:10.18653/v1/2020.clinicalnlp-1.6

Cited by 12 publications

(3 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The introduction of SpanBERT [9] architecture to ADE extraction task outperformed competing models [10] on Social Media Mining for Health (SMM4H) and CSIRO Adverse Drug Event Corpus (CADEC). [11,12] Narayanan et al [13] evaluated various biomedical contextual embeddings and models using the 2018 National NLP Clinical Challenges (n2c2) shared task Track 2 data on ADEs and Medication Extraction, [14] demonstrating the importance of BERT structure. In recent work on adverse event (AE) detection, Chopard et al [15] confirmed the feasibility of automating coding of AEs described in the narrative section of serious AE report forms.…”

Section: Background and Significancementioning

confidence: 99%

Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records

Wu,

Ruan,

McNeer

et al. 2024

Preprint

View full text Add to dashboard Cite

ObjectiveTo develop a transformer-based natural language processing (NLP) system for detecting adverse drug events (ADEs) from clinical notes in electronic health records (EHRs).Materials and MethodsWe fine-tuned BERT Short-Formers and Clinical-Longformer using the processed dataset of the 2018 National NLP Clinical Challenges (n2c2) shared task Track 2. We investigated two data processing methods, window-based and split-based approaches, to find an optimal processing method. We evaluated the generalization capabilities on a dataset extracted from Vanderbilt University Medical Center (VUMC) EHRs.ResultsOn the n2c2 dataset, the best average macro F-scores of 0.832 and 0.868 were achieved using a 15-word window with PubMedBERT and a 10-chunk split with Clinical-Longformer. On the VUMC dataset, the best average macro F-scores of 0.720 and 0.786 were achieved using a 4-chunk split with PubMedBERT and Clinical-Longformer.DiscussionOur study provided a comparative analysis of data processing methods. The fine-tuned transformer models showed good performance for ADE-related tasks. Especially, Clinical-Longformer model with split-based approach had a great potential for practical implementation of ADE detection. While the token limit was crucial, the chunk size also significantly influenced model performance, even when the text length was within the token limit.ConclusionWe provided guidance on model development, including data processing methods for ADE detection from clinical notes using transformer-based models. Our results on two datasets indicated that data processing methods and models should be carefully selected based on the type of clinical notes and the allocation trade-offs of human and computational power in annotation and model fine-tuning.

show abstract

Section: Background and Significancementioning

confidence: 99%

Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records

Wu,

Ruan,

McNeer

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In a similar vein, Peng et al [16] have proposed a benchmark setting (called BLUE) to evaluate pre-trained model in a clinical setting, showing that BERT model pre-trained on PubMed abstracts and MIMIC-III was superior (we refer to it as BlueBert in the rest of the paper). Recently, biomedical contextual embeddings have also been applied to improve the performance of adverse drug identification and the medication extraction task [9], [17].…”

Section: Related Workmentioning

confidence: 99%

Neural Medication Extraction: A Comparison of Recent Models in Supervised and Semi-supervised Learning Settings

Kocabiyikoglu¹,

Portet²,

Qader³

et al. 2021

Preprint

View full text Add to dashboard Cite

Drug prescriptions are essential information that must be encoded in electronic medical records. However, much of this information is hidden within free-text reports. This is why the medication extraction task has emerged. To date, most of the research effort has focused on small amount of data and has only recently considered deep learning methods. In this paper, we present an independent and comprehensive evaluation of state-ofthe-art neural architectures on the I2B2 medical prescription extraction task both in the supervised and semi-supervised settings.The study shows the very competitive performance of simple DNN models on the task as well as the high interest of pre-trained models. Adapting the latter models on the I2B2 dataset enables to push medication extraction performances above the state-ofthe-art. Finally, the study also confirms that semi-supervised techniques are promising to leverage large amounts of unlabeled data in particular in low resource setting when labeled data is too costly to acquire.

show abstract

“…Transfer learning can be effective for medication extraction from clinical texts [ 8]. In this paper, we address how named entity recognition and relation extraction can be performed for medication-related incident reports, even when few gold-standard data are available [4].…”

Section: Introductionmentioning

confidence: 99%

Rule-Based Natural Language Processing Pipeline to Detect Medication-Related Named Entities: Insights for Transfer Learning

Wong,

Waters,

Kuo

et al. 2024

Studies in Health Technology and Informatics

View full text Add to dashboard Cite

We document the procedure and performance of a rule-based NLP system that, using transfer learning, automatically extracts essential named entities related to drug errors from Japanese free-text incident reports. Subsequently, we used the rule-based annotated data to fine-tune a pre-trained BERT model and examined the performance of medication-related incident report prediction. The rule-based pipeline achieved a macro-F1-score of 0.81 in an internal dataset and the BERT model fine-tuned with rule-annotated data achieved a macro-F1-score of 0.97 and 0.75 for named entity recognition and relation extraction tasks, respectively. The model can be deployed to other, similar problems in medication-related clinical texts.

show abstract

Evaluation of Transfer Learning for Adverse Drug Event (ADE) and Medication Entity Extraction

Cited by 12 publications

References 30 publications

Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records

Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records

Neural Medication Extraction: A Comparison of Recent Models in Supervised and Semi-supervised Learning Settings

Rule-Based Natural Language Processing Pipeline to Detect Medication-Related Named Entities: Insights for Transfer Learning

Contact Info

Product

Resources

About