OBJECTIVE: To assess the performance of large language models (LLMs) for zero-shot disambiguation of acronyms in clinical narratives. MATERIALS AND METHODS: Clinical narratives in English, German, and Portuguese were applied for testing the performance of four LLMs: GPT-3.5, GPT-4, Llama-2-7b-chat, and Llama-2-70b-chat. For English, the anonymized Clinical Abbreviation Sense Inventory (CASI, University of Minnesota) was used. For German and Portuguese, at least 500 text spans were processed. The output of LLM models, prompted with contextual information, was analyzed to compare their acronym disambiguation capability, grouped by document-level metadata, the source language, and the LLM. RESULTS: On CASI, GPT-3.5 achieved 0.91 in accuracy. GPT-4 outperformed GPT-3.5 across all datasets, reaching 0.98 in accuracy for CASI, 0.86 and 0.65 for two German datasets, and 0.88 for Portuguese. Llama models only reached 0.73 for CASI and failed severely for German and Portuguese. Across LLMs, performance decreased from English to German and Portuguese processing languages. There was no evidence that additional document-level metadata had a significant effect. CONCLUSION: For English clinical narratives, acronym resolution by GPT-4 can be recommended to improve readability of clinical text by patients and professionals. For German and Portuguese, better models are needed. Llama models, which are particularly interesting for processing sensitive content on premise, cannot yet be recommended for acronym resolution.

Medical texts often contain expressions that are not listed in biomedical dictionaries and terminology systems. In this investigation, the detection of such entities is examined with online health forum entries in the Croatian language. Emphasis is put on short-form content, lexical variations, brand names, and proper names. By leveraging Transformer architectures on token and entity level, noteworthy results (>90% in F1-measure) could be achieved, which is on par with state-of-the-art named entity recognition in high-resource languages. Additionally, this investigation showcases a way to recognize non-lexicalized tokens from texts, which can be of use for further research questions in combination with word sense disambiguation, word-embeddings, and terminology expansion workflows.

A log-likelihood based co-occurrence analysis of ∼1.9 million de-identified ICD-10 codes and related short textual problem list entries generated possible term candidates at a significance level of p<0.01. These top 10 term candidates, consisting of 1 to 5-grams, were used as seed terms for an embedding based nearest neighbor approach to fetch additional synonyms, hypernyms and hyponyms in the respective n-gram embedding spaces by leveraging two different language models. This was done to analyze the lexicality of the resulting term candidates and to compare the term classifications of both models. We found no difference in system performance during the processing of lexical and non-lexical content, i.e. abbreviations, acronyms, etc. Additionally, an application-oriented analysis of the SapBERT (Self-Alignment Pretraining for Biomedical Entity Representations) language model indicates suitable performance for the extraction of all term classifications such as synonyms, hypernyms, and hyponyms.

OBJECTIVE: Automated clinical phenotyping is challenging because word-based features quickly turn it into a high-dimensional problem, in which the small, privacy-restricted, training datasets might lead to overfitting. Pretrained embeddings might solve this issue by reusing input representation schemes trained on a larger dataset. We sought to evaluate shallow and deep learning text classifiers and the impact of pretrained embeddings in a small clinical dataset. MATERIALS AND METHODS: We participated in the 2018 National NLP Clinical Challenges (n2c2) Shared Task on cohort selection and received an annotated dataset with medical narratives of 202 patients for multilabel binary text classification. We set our baseline to a majority classifier, to which we compared a rule-based classifier and orthogonal machine learning strategies: support vector machines, logistic regression, and long short-term memory neural networks. We evaluated logistic regression and long short-term memory using both self-trained and pretrained BioWordVec word embeddings as input representation schemes. RESULTS: Rule-based classifier showed the highest overall micro F1 score (0.9100), with which we finished first in the challenge. Shallow machine learning strategies showed lower overall micro F1 scores, but still higher than deep learning strategies and the baseline. We could not show a difference in classification efficiency between self-trained and pretrained embeddings. DISCUSSION: Clinical context, negation, and value-based criteria hindered shallow machine learning approaches, while deep learning strategies could not capture the term diversity due to the small training dataset. CONCLUSION: Shallow methods for clinical phenotyping can still outperform deep learning methods in small imbalanced data, even when supported by pretrained embeddings.