Modeling Dutch Medical Texts for Detecting Functional Categories and Levels of COVID-19 Patients

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 4577–4585 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 4577 Modeling Dutch Medical Texts for Detecting Functional Categories and Levels of COVID-19 Patients Jenia Kim∗, Stella Verkijk∗, Edwin Geleijn†, Marike van der Leeden†, Carel Meskers†, Caroline Meskers†, Sabina van der Veen⋄, Piek Vossen∗, Guy Widdershoven⋄ ∗Computational Linguistics and Text Mining Lab, Faculty of Humanities, Vrije Universiteit Amsterdam jenka@protonmail.com, stellaverkijk@outlook.com, p.t.j.m.vossen@vu.nl †Department of Rehabilitation Medicine, Amsterdam University Medical Centers {e.geleijn, m.vanderleeden, c.meskers, c.j.w.meskers}@amsterdamumc.nl ⋄Department of Ethics, Law and Humanities, Amsterdam University Medical Centers {s.vanderveen, g.widdershoven}@amsterdamumc.nl Abstract Electronic Health Records contain a lot of information in natural language that is not expressed in the structured clinical data. Especially in the case of new diseases such as COVID-19, this information is crucial to get a better understanding of patient recovery patterns and factors that may play a role in it. However, the language in these records is very different from standard language and generic natural language processing tools cannot easily be applied out-of-the-box. In this paper, we present a fine-tuned Dutch language model specifically developed for the language in these health records that can determine the functional level of patients according to a standard coding framework from the World Health Organization. We provide evidence that our classification performs at a sufficient level (F1-score above 80% for the main categories and error rates of less than 1 level on a 5-point Likert scale for levels) to generate patient recovery patterns that can be used to analyse factors that contribute to the rehabilitation of COVID-19 patients and to predict individual patient recovery of functioning. Keywords: medical text mining, electronic health records, Dutch language models, functional level classification, COVID-19 1. Introduction Electronic Health Records (EHRs) contain a wealth of unexplored information. Some of it is captured in the structured clinical data on patients (e.g. diagno- sis codes, lab results); other aspects are not recorded as structured data, but rather appear in free-text form. This unstructured data may be essential, especially in the case of new diseases such as COVID-19 on which we have little knowledge. Natural Language Process- ing (NLP) has a great potential to mine the free-text in EHRs for significant patterns and insights, helping us to learn more about the disease and its effects. One important aspect of health and well-being is cap- tured by functional status, i.e. the individual’s ability to engage in different activities and social roles. For example, in the case of COVID-19, it is known that al- most 80% of hospitalized patients report at least one persistent symptom (such as fatigue, sleep difficulty, anxiety etc.) six months after discharge (Huang et al., 2021). This might adversely impact their functioning, including their ability to work, to exercise, to maintain a healthy body weight, etc. Deployment of appropri- ate and personalized rehabilitation plans for these pa- tients requires insight into how their functioning devel- ops over time, so that relevant patterns and predictors can …

Content truncated to 500 words