RESUMO
PURPOSE: To address the feasibility, reliability and internal validity of natural language processing (NLP) for automated functional assessment of hospitalised COVID-19 patients in key International Classification of Functioning, Disability and Health (ICF) categories and levels from unstructured text in electronic health records (EHR) from a large teaching hospital. MATERIALS AND METHODS: Eight human annotators assigned four ICF categories to relevant sentences: Emotional functions, Exercise tolerance, Walking and Moving, Work and Employment and their ICF levels (Functional Ambulation Categories for Walking and Moving, metabolic equivalents for Exercise tolerance). A linguistic neural network-based model was trained on 80% of the annotated sentences; inter-annotator agreement (IAA, Cohen's kappa), a weighted score of precision and recall (F1) and RMSE for level detection were assessed for the remaining 20%. RESULTS: In total 4112 sentences of non-COVID-19 and 1061 of COVID-19 patients were annotated. Average IAA was 0.81; F1 scores were 0.7 for Walking and Moving and Emotional functions; RMSE for Walking and Moving (5- level scale) was 1.17 for COVID-19 patients. CONCLUSION: Using a limited amount of annotated EHR sentences, a proof-of-concept was obtained for automated functional assessment of COVID-19 patients in ICF categories and levels. This allows for instantaneous assessment of the functional consequences of new diseases like COVID-19 for large numbers of patients.Key messagesHospitalised Covid-19 survivors may persistently suffer from low physical and mental functioning and a reduction in overall quality of life requiring appropriate and personalised rehabilitation strategies.For this, assessment of functioning within multiple domains and categories of the International Classification of Function is required, which is cumbersome using structured data.We show a proof-of-concept using Natural Language Processing techniques to automatically derive the aforementioned information from free-text notes within the Electronic Health Record of a large academic teaching hospital.
Assuntos
COVID-19 , Registros Eletrônicos de Saúde , Avaliação da Deficiência , Humanos , Processamento de Linguagem Natural , Qualidade de Vida , Reprodutibilidade dos Testes , SARS-CoV-2RESUMO
Will reading different stories about the same event in the world result in a similar image of the world? Will reading the same story by different people result in a similar proxy for experiencing the story? The answer to both questions is no because language is abstract by definition and relies on our episodic experience to turn a story into a more concrete mental movie. Since our episodic knowledge differs, also the mental movie will be different. Language leaves out details, and this becomes specifically clear when building machines that read texts to represent events and to establish event relations across mentions, such as co-reference, causality, subevents, scripts, timelines, and storylines. There is a lot of information and knowledge on the event that is not in the text but is needed to reconstruct these relations and understand the story. Machines lack this knowledge and experience and likewise make explicit what it takes to understand stories from text. In this paper, we report on experiments to automatically model event descriptions and instances across different news articles. We will show that event information is scattered over the text but also varies a lot in the degree it abstracts from details, which makes establishing event identity and relations extremely difficult. The variation in granularity of event descriptions seems to vary with pragmatic communicative strategies and defines the problem at different levels of complexity.