Summarizing Patients’ Problems from Hospital Progress Notes Using Pre-trained Sequence-to-Sequence Models

Proceedings of the 29th International Conference on Computational Linguistics, pages 2979–2991 October 12–17, 2022. 2979 Summarizing Patients’ Problems from Hospital Progress Notes Using Pre-trained Sequence-to-Sequence Models Yanjun Gao* ICU Data Science Lab School of Medicine and Public Health University of Wisconsin-Madison ygao@medicine.wisc.edu Dmitriy Dligach Loyola University Chicago ddligach@luc.edu Timothy Miller, Dongfang Xu Boston Children’s Hospital and Harvard Medical School Timothy.Miller, Dongfang.Xu @childrens.harvard.edu Matthew M. Churpek, Majid Afshar ICU Data Science Lab School of Medicine and Public Health University of Wisconsin-Madison mchurpek, mafshar@medicine.wisc.edu Abstract Automatically summarizing patients’ main problems from daily progress notes using nat- ural language processing methods helps to battle against information and cognitive over- load in hospital settings and potentially assists providers with computerized diagnostic deci- sion support. Problem list summarization re- quires a model to understand, abstract, and gen- erate clinical documentation. In this work, we propose a new NLP task that aims to generate a list of problems in a patient’s daily care plan using input from the provider’s progress notes during hospitalization. We investigate the per- formance of T5 and BART, two state-of-the-art seq2seq transformer architectures, in solving this problem. We provide a corpus built on top of progress notes from publicly available electronic health record progress notes in the Medical Information Mart for Intensive Care (MIMIC)-III. T5 and BART are trained on gen- eral domain text, and we experiment with a data augmentation method and a domain adaptation pre-training method to increase exposure to medical vocabulary and knowledge. Evaluation methods include ROUGE, BERTScore, cosine similarity on sentence embedding, and F-score on medical concepts. Results show that T5 with domain adaptive pre-training achieves sig- nificant performance gains compared to a rule- based system and general domain pre-trained language models, indicating a promising direc- tion for tackling the problem summarization task. 1 Introduction The progress note is a common note type in the electronic health record (EHR) that also contains the necessary details for medical billing; there- fore, every hospital day will contain at least one Figure 1: When a sick patient arrives to the hospital, diagnos- tic evaluations are performed to assess the patient’s condition and deduce the problems causing the illness. progress note for a patient. Healthcare providers write them to document a patient’s daily progress and care plan (Brown et al., 2014). The progress note contains both subjective and objective infor- mation gathered by the care team, and it is up- dated daily and serves as the most viewed clini- cal document by providers. The complexity of the progress note increases as the patient’s illness wors- ens with progress notes collected in the intensive care unit (ICU) representing the sickest patients in the hospital. In the ICU, information and cognitive overload occur frequently, with more opportunities for missed diagnoses and medical errors (Furlow, 2020; Hultman et al., 2019). Automatically gen- erating a set of diagnoses/problems in a progress note may assist providers in overcoming cogni- tive biases and heuristics and apply evidence-based medicine via information synthesis to accurately understand a patient’s condition. These …

Content truncated to 500 words