diff --git "a/AdFJT4oBgHgl3EQfrS3C/content/tmp_files/load_file.txt" "b/AdFJT4oBgHgl3EQfrS3C/content/tmp_files/load_file.txt" new file mode 100644--- /dev/null +++ "b/AdFJT4oBgHgl3EQfrS3C/content/tmp_files/load_file.txt" @@ -0,0 +1,1178 @@ +filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf,len=1177 +page_content='A Multi-View Joint Learning Framework for Embedding Clinical Codes and Text Using Graph Neural Networks Lecheng Kong, Christopher King, Bradley Fritz, Yixin Chen Washington University in St.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Louis One Brookings Drive St.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Louis, Missouri 63130, USA {jerry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='kong, christopherking, bafritz, ychen25}@wustl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='edu Abstract Learning to represent free text is a core task in many clini- cal machine learning (ML) applications, as clinical text con- tains observations and plans not otherwise available for in- ference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' State-of-the-art methods use large language models developed with immense computational resources and train- ing data;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' however, applying these models is challenging be- cause of the highly varying syntax and vocabulary in clinical free text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Structured information such as International Clas- sification of Disease (ICD) codes often succinctly abstracts the most important facts of a clinical encounter and yields good performance, but is often not as available as clinical text in real-world scenarios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We propose a multi-view learning framework that jointly learns from codes and text to com- bine the availability and forward-looking nature of text and better performance of ICD codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The learned text embed- dings can be used as inputs to predictive algorithms indepen- dent of the ICD codes during inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our approach uses a Graph Neural Network (GNN) to process ICD codes, and Bi- LSTM to process text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We apply Deep Canonical Correlation Analysis (DCCA) to enforce the two views to learn a similar representation of each patient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In experiments using planned surgical procedure text, our model outperforms BERT models fine-tuned to clinical data, and in experiments using diverse text in MIMIC-III, our model is competitive to a fine-tuned BERT at a tiny fraction of its computational effort.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We also find that the multi-view approach is beneficial for stabilizing inferences on codes that were unseen during train- ing, which is a real problem within highly detailed coding systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We propose a labeling training scheme in which we block part of the training code during DCCA to improve the generalizability of the GNN to unseen codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In experi- ments with unseen codes, the proposed scheme consistently achieves superior performance on code inference tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 1 Introduction An electronic health record (EHR) stores a patient’s com- prehensive information within a healthcare system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' It pro- vides rich contexts for evaluating the patient’s status and fu- ture clinical plans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The information in an EHR can be clas- sified as structured or unstructured.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Over the past decade, ML techniques have been widely applied to uncover pat- terns behind structured information such as lab results (Yu, Copyright © 2023, Association for the Advancement of Artificial Intelligence (www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='aaai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='org).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' All rights reserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Beam, and Kohane 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Shickel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Goldstein et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Recently, the surge of deep learning and large-scale pre-trained networks has allowed unstructured data, mainly clinical notes, to be effectively used for learning (Huang, Al- tosaar, and Ranganath 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Si et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, most methods focus on either structured or un- structured data only.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A particularly informative type of structured data is the International Classification of Diseases (ICD) codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ICD is an expert-identified hierarchical medical concept ontology used to systematically organize medical concepts into cate- gories and encode valuable domain knowledge about a pa- tient’s diseases and procedures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Because ICD codes are highly specific and unambigu- ous, ML models that use ICD codes to predict procedure outcomes often yield more accurate results than those do not (Deschepper et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, the availability of ICD codes is not always guaranteed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, billing ICD codes are generated after the clinical encounter, meaning that we cannot use the ICD codes to predict post-operative outcomes before the surgery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A more subtle but crucial drawback of using ICD codes is that there might be unseen codes during inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' When a future pro- cedure is associated with a code outside the trained subset, most existing models using procedure codes cannot accu- rately represent the case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Shifts in coding practices can also cause data during inference to not overlap the trained set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' On the other hand, unstructured text data are readily and consistently available.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Clinical notes are generated as free text and potentially carry a doctor’s complete insight about a patient’s condition, including possible but not known di- agnoses and planned procedures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Unfortunately, the clinical text is a challenging natural language source, containing am- biguous abbreviations, input errors, and words and phrases rarely seen in pre-training sources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' It is consequently diffi- cult to train a robust model that predicts surgery outcomes from the large volume of free texts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Most current models rely on large-scale pre-trained models (Huang, Altosaar, and Ranganath 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Such methods require a considerable corpus of relevant texts to fine-tune, which might not be available at a particular facility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hence, mod- els that only consider clinical texts suffer from poor perfor- mance and incur huge computation costs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' To overcome the problems of models using only text or arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='11608v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='CL] 27 Jan 2023 codes, we propose to learn from the ICD codes and clini- cal text in a multi-view joint learning framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We ob- serve that despite having different formats, the text and code data are complementary and broadly describe the same un- derlying facts about the patient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This enables each learner (view) to use the other view’s representation as a regulariza- tion function where less information is present.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Under our framework, even when one view is missing, the other view can perform inference independently and maintain the effec- tive data representation learned from the different perspec- tives, which allows us to train reliable text models without a vast corpus and computation cost required by other text-only models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Specifically, we make the following contributions in this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (1) We propose a multi-view learning framework us- ing Deep Canonical Correlation Analysis (DCCA) for ICD codes and clinical notes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (2) We propose a novel tree-like structure to encode ICD codes by relational graph and ap- ply Relational Graph Convolution Network (RGCN) to em- bed ICD codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (3) We use a two-stage Bi-LSTM to en- code lengthy clinical texts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (4) To solve the unseen code pre- diction problem, we propose a labeling training scheme in which we simulate unseen node prediction during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Combined with the DCCA optimization process, the training scheme teaches the RGCN to discriminate between unseen and seen codes during inference and achieves better perfor- mance than plain RGCN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2 Related Works Deep learning on clinical notes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Many works focus on applying deep learning to learn representations of clini- cal texts for downstream tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Early work (Boag et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018) compared the performance of classic NLP meth- ods including bag-of-words (Zhang, Jin, and Zhou 2010), Word2Vec (Mikolov et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013), and Long-Short-Term- Memory (LSTM) (Hochreiter and Schmidhuber 1997) on clinical prediction tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' These methods solely learn from the training text, but as the clinical texts are very noisy, they either tend to overfit the data or fail to uncover valuable pat- terns behind the text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Inspired by large-scale pre-trained lan- guage models such as BERT (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018), a series of works developed transformer models pre-trained on medical notes, including ClinicalBERT (Huang, Altosaar, and Ran- ganath 2019), BioBERT (Lee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020), and PubBERT (Alsentzer et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' These models fine-tune general lan- guage models on a large corpus of clinical texts and achieve superior performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Despite the general nature of these models, the fine-tuning portion may not translate well to new settings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, PubBERT is trained on the clinical texts of a single tertiary hospital, and the colloquial terms used and procedures typically performed may not map to different hospitals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' BioBERT is trained on Pubmed abstracts and articles, which also is likely poorly representative of the topics and terms used to, for example, describe a planned surgery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Some other models propose to use joint learning models to learn from the clinical text, and structured data (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', mea- sured blood pressure and procedure codes) (Wei et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Since the structured data are less noisy, these models can produce better and more stable results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, most assume the co-existence of text and struc- tured data at the inference time, while procedure codes for a patient are frequently incomplete until much later.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Machine learning and procedure codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Procedure codes are a handy resource for EHR data mining.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Most works focus on automatic coding, using machine learning models to predict a patient’s diagnostic codes from clini- cal notes (Pascual, Luck, and Wattenhofer 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li and Yu 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Some other works directly use the billing code to pre- dict clinical outcomes (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deschepper et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019), whereas our work focuses on using the high correla- tion of codes and text data to augment the performance of each.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Most of these works exploit the code hierarchies by human-defined logic based on domain knowledge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In con- trast, our proposed framework uses GNN and can encode arbitrary relations between codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Graph neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A series of works (Xu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Gilmer et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017) summarize GNN structures in which each node iteratively aggregates neighbor nodes’ em- bedding and summarizes information in a neighborhood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The resulting node embeddings can be used to predict down- stream tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' RGCN (Schlichtkrull et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018) generalizes GNN to heterogeneous graphs where nodes and edges can have different types.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our model utilizes such heterogeneous properties on our proposed hierarchy graph encoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Some works (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Choi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020) applied GNN to model interaction between EHRs, whereas our model uses GNN on the code hierarchy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Privileged information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our approach is related to the Learning Under Privileged Information (LUPI) (Vapnik and Vashist 2009) paradigm, where the privileged information is only accessible during training (in this case, billing code data).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Many works have applied LUPI to other fields like computer vision (Lambert, Sener, and Savarese 2018) and metric learning (Fouad et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 3 Methods Admissions with ICD codes and clinical text can be repre- sented as D = {(C1, A1, y1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', (Cn, An, yn)}, where Ci is a set of ICD codes for admission i, Ai is a set of clin- ical texts, and yi is the desired task label (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' mortality, re-admission, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=').' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The ultimate goal is to minimize task- appropriate losses L defined as: min fC,gC � i L(fC(gC(Ci)), yi) (1) and min fA,gA � i L(fA(gA(Ai)), yi), (2) where gC and gA embed codes and texts to vector repre- sentations respectively, and fC and fA map representations to the task labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Note that (gC, fC) and (gA, fA) should operate independently during inference, meaning that even when one type of data is missing, we can still make accurate predictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In this section, we first propose a novel ICD ontology graph encoding method and describe how we use Graph Figure 1: Overall multi-view joint learning framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Blue boxes/arrows represent the text prediction pipeline, and green represents the code prediction pipeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Dashed boxes and arrows denote processes only happening during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' By removing the dashed parts, text and code pipelines can predict tasks independently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Neural Network (GNN) to parameterize gC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We then de- scribe the two-stage Bi-LSTM (gA) to embed lengthy clini- cal texts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We then describe how to use DCCA on the repre- sentation from gC and gA to generate representations that are less noisy and more informative, so the downstream models fC and fA are able to make accurate predictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Figure 1 shows the overall architecture of our multi-view joint learn- ing framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ICD Ontology as Graphs The ICD ontology has a hierarchical scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We can rep- resent it as a tree graph as shown in Figure 2, where each node is a medical concept and a node’s children are finer di- visions of the concept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' All top-level nodes are connected to a root node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In this tree graph, only the leaf nodes correspond to observable codes in the coding system, all other nodes are the hierarchy of the ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This representation is widely adopted by many machine learning systems (Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, Ma, and Gao 2021) as a refinement of the earlier approach of grouping together all codes at the top level of the hierarchy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A tree graph is ideal for algorithms based on mes- sage passing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' It allows pooling of information within disjoint groups, and encodes a compact set of neighbors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, it (1) ignores the granularity of different levels of classifica- tion, and (2) cannot encode similarities of nodes that are dis- tant from each other.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This latter point comes about because a tree system may split on factors that are not the most rele- Figure 2: Top: Conventional encoding of ICD ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bot- tom Left: ICD ontology encoded with relations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Relation types for different levels are denoted by different colors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bottom Right: Jump connection creates additional edges to leaf nodes’ predecessors, denoted by dashed color lines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' vant for a given task, such as the same procedure in an arm versus a leg, or because cross-system concepts are empiri- cally very correlated in medical syndromes, such as kidney failure and certain endocrine disorders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' To overcome the aforementioned problems, we propose to augment the tree graph with edge types and jump connec- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Unlike conventional tree graphs, where all edges have the same edge type, we use different edge types for connec- tions between different levels in the tree graph as shown in the bottom left of Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, ICD-10 codes have seven characters and hence eight levels in the graph (includ- ing the root level).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The edges between the root node and its children have edge Type 1, and the edges between the sev- enth level and the last level (actual code level) have edge Type 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Different edge types not only encode whether two procedures are related but also encode the level of similarity between codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' With multiple edge types introduced to the graph, we are able to further extend the graph structure by jump connec- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For each leaf node, we add one additional edge be- tween the node and each of its predecessors up to the root node, as shown in the bottom right of Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The edge type depends on the level that the predecessor resides.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, in the ICD-10 tree graph, a leaf node will have seven additional connections to its predecessors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Its edge to the root node will have Type 8 (the first seven types are used to represent connections between levels), and its edge to the third level node will have Type 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jump connections signifi- cantly increase the connectivity of the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Meanwhile, we still maintain the good hierarchical information of the origi- Adm 1: Posterior Cervical Decompression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Adm 1: 0QSH06Z Adm 2: Thoracic Laminectomy for.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Adm 2: 00CU0ZZ,009U3ZX,02HV33Z Adm 3: .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Adm 3: .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Word2Vec ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Code to Node Index ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Two-Stage LSTM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='RGCN ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Codes Embedding ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Text Embedding ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Generated from ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Sum/Max Pooling ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Text Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='DCCA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Code Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Projected Text ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Projected Codes ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Embedding ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Embedding ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='MLP ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='MLP ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Downstream Task Predictior ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Downstream Task PredictionRoot node r connects ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='to all level-1 ontology ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Bottom level nodes ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='represent actual codes ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0RGAOT0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='ORGA0T1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5A09357 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5A09358 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Relation-Augmented Graph ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='Jump Connection Graphnal tree graph because the jump connections are represented ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='by a different set of edge types.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Using jump connection helps uncover relationships between codes that are not presented in the ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, the relationship between ane- mia and renal failure can be learned using jump connec- tion even though these diverge at the root node in ICD-9 and ICD-10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Moreover, GNNs suffer from over-smoothing, where all node representations converge to the same value when the GNN has too many layers (Li, Han, and Wu 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' If we do not employ jump connections, the maximal distance between one leaf node to another is twice the number of levels in the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' To capture the connection between the nodes, we will need a GNN with that many layers, which is computationally expensive and prone to over-smoothing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jump connections make the distance between two leaf nodes two, and this ensures that the GNN is able to embed any correlation between two nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We will discuss this in more detail in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 Embedding ICD Codes using GNN We use GNN to embed medical concepts in the ICD ontol- ogy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Let G = {V, E, R} be a graph, where V is its set of the vertex (medical concepts in the ICD graph), E ⊆ {V ×V } is its set of edges (connects medical concept to its sub-classes), and R is the set of edge type in the graph (edges in different levels and jump connection).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' As each ICD code corresponds to one node in the graph, we use code and node interchange- ably.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We adopt RGCN (Schlichtkrull et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018), which itera- tively updates a node��s embedding from its neighbor nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Specifically, the kth layer of RGCN on node u ∈ V is: h(k+1) u = σ � �� r∈R � v∈N r u 1 cu,r W (k) r h(k) v + W (k)h(k) u � � (3) where N r i is the set of neighbors of i that connects to i by re- lation r, h(k) i is the embedding of node i after k GNN layers, h0 i is a randomly initialized trainable embedding, W (k) r is a linear transformation on embeddings of nodes in N r i , W (k) updates the embedding of u, and σ is a nonlinear activation function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We have c = |N r i | as a normalization factor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' After T iterations, h(T ) u can be used to learn down- stream tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Since a patient can have a set of codes, Ci = {vi1, vi2, vi3, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='} ⊆ V , we use sum and max pooling to sum- marize Ci in an embedding function gC: gC(Ci) = � v∈Ci h(T ) v ⊕ max({h(T ) v |v ∈ Ci}), (4) where max is the element-wise maximization, and ⊕ rep- resents vector concatenation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Summation more accurately summarizes the codes’ information, while maximization provides regularization and stability in DCCA, which we will discuss in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Training RGCN helps embed the ICD codes into vectors based on the defined ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Nodes that are close together in the graph will be assigned similar embeddings because of their similar neighborhood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Moreover, distant nodes that appear together frequently in the health record can also be assigned correlated embeddings because the jump connec- tion keeps the maximal distance between two nodes at two.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Consider a set of codes C = {u, v}, because of the sum- mation in the code pooling, using a 2-layer RGCN, we will have non-zero gradients of hT u and hT v with respect to h0 v and h0 u, respectively, which connects the embeddings of u and v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In contrast, applying RGCN on a graph without jump connections will result in zero gradients when the distance between u and v is greater than two.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 Embedding Clinical Notes using Bi-LSTM Patients can have different numbers of clinical texts in each encounter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Where applicable, we sort the texts in an en- counter in ascending order by time, and have a set of texts Ai = (ai1, ai2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', ain).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In our examples, we concatenate the texts together to a single document Hi, and we have Hi = CAT(Ai) = � j={1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='n} aij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We leave to future work the possibility of further modeling the collection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The concatenated text might be very lengthy with over ten thousands word tokens, and RNN suffers from dimin- ishing gradients with LSTM-type modifications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' While at- tention mechanisms are effective for arbitrary long-range dependence, they require large sample size and expensive computational resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hence, following previously suc- cessful approach (Huang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019), we adopt a two-stage model which stacks a low-frequency RNN on a local RNN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Given Hi, we first split it into blocks of equal size b, Hi = {Hi1, Hi2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', HiK}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The last block HiK is padded to length b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The two-stage model first generates block-wise text em- beddings by lHik = LSTM({w(Hik1), w(Hik2), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', w(Hikb)}), (5) where w(·) is a Word2Vec (Mikolov et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013) trainable embedding function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The representation of Ai is given by gA(Ai) = LSTM({lHi1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', lHiK}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (6) The two-stage learning scheme minimizes the effect of di- minishing gradients while maintaining the temporal order of the text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 DCCA between Graph and Text Data As previously mentioned, ICD codes may not be available at the time when models would be most useful, but are struc- tured and easier to analyze, while the clinical text is read- ily available but very noisy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Despite different data formats, they usually describe the same information: the main diag- noses and treatments for an encounter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Borrowing ideas from multi-view learning, we can use them to supplement each other.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Many existing multi-view learning methods require the presence of both views during inference and are not able to adapt to the applications we envision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Specifically, we use DCCA (Andrew et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2015) on gA(Ai) and gC(Ci) to learn a joint representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' DCCA solves the following optimization problem, max gC,gA,U,V 1 N tr(U T M T C MAV ) s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' U T ( 1 N M T C MC + rCI)U = I, V T ( 1 N M T AMA + rAI)V = I, uT i M T C MAvj = 0, ∀i ̸= j, 1 ≤ i, j ≤ L MC = stack{gC(Ci)|∀i}, MA = stack{gA(Ai)|∀i}, (7) where MC and MA are the matrices stacked by vector rep- resentations of codes and texts, (rC, rA) > 0 are regulariza- tion parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' U = [u1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', uL] and V = [v1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=', vL] maps GNN and Bi-LSTM output to maximally correlated embed- ding, and L is a hyper-parameter controlling the number of correlated dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use gC(Ci)U, gA(Ai)V as the fi- nal embedding of codes and texts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' By maximizing their cor- relation, we force the weak learner (usually the LSTM) to learn a similar representation as the strong learner (usually the GNN) and to filter out inputs unrelated to the structured data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hence, when a record’s codes can yield correct results, its text embedding is highly correlated with that of the codes, and the text should also be likely to produce correct predic- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' During development, we found that a batch of ICD data often contains many repeated codes with the same embed- ding and that a SUM pooling tended to obtain a less than full rank embedding matrix MC and MA, which causes in- stability in solving the optimization problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A nonlinear max pooling function helps prevent this.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The above optimization problem suggests full-batch train- ing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, the computation graph will be too large for the text and code data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Following (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2015), we use large mini-batches to train the model, and from the experi- mental results, they sufficiently represent the overall distri- bution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' After training, we stack MC, MA again from all data output and obtain U and V as fixed projection matrix from equation (7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' After obtaining the projection matrices and embedding models, we attach two MLPs (fA and fC) to the embedding models as the classifier, and train/fine-tune fA (fC) and gA (gC) together in an end-to-end fashion with respect to the learning task using the loss functions in (1) and (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 4 Predicting Unseen Codes In the previous section, we discuss the formulation of ICD ontology and how we can use DCCA to generate embed- dings that share representations across views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In this section, we will demonstrate another use case for DCCA-regularized embeddings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In real-world settings, the set of codes that re- searchers observe in training is usually a small subset of the entire ICD ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In part, this is due to the extreme speci- ficity of some ontologies, with ICD-10-PCS having 87,000 distinct procedures and ICD-10-CM 68,000 diagnostic pos- sibilities before considering that some codes represent a fur- ther modification of another entity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In even large training samples, some codes will likely be seen zero or a small num- ber of times in training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Traditional models using indepen- dent code embedding are expected to function poorly on rare codes and have arbitrary output on previously unseen nodes, even if similar entities are contained in the training data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our proposed model and the graph-embedded hierarchy can naturally address the above challenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Its two features enable predictions of novel codes at inference: Relational embedding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' By embedding the novel code in the ontology graph, we are able to exploit the repre- sentation of its neighbors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For example, a rare diagnostic procedure’s embedding is highly influenced by other pro- cedures that are nearby in the ontology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jump connection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' While other methods also exploit the proximity defined by the hierarchy, as we suggested above, codes can be highly correlated but remain distant in the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jump connections increase the graph con- nectivity;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' hence, our model can seek the whole hierarchy for potential connection to the missing code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Because the connections across different levels are assigned different relation types, our GNN can also differentiate the likeli- hood of connections across different levels and distances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Meanwhile, during inference, the potential problem is that the model does not automatically differentiate between the novel and the previously seen codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Because the model never uses novel codes to generate any gC(Ci), the embed- dings of the seen and novel nodes experience different gra- dient update processes and hence are from different distri- butions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Nevertheless, during inference, the model will treat them as if they are from the same distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' However, such transferability and credibility of novel node embed- dings are not guaranteed, and applying them homogeneously may result in untrustworthy predictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hence, we propose a labeling training scheme to teach the model how to handle novel nodes during inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Let G = {V, E, R} be the ICD graph and U be the set of unique nodes in the training set, U ⊆ V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We select a random subset Us from U to form the seen nodes during training, and Uu = V \\ Us be treated as unseen nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We augment the initial node embeddings with 1-0 labels, formally, h0+ u = h0 u ⊕ 1 ∀u ∈ Us h0+ v = h0 v ⊕ 0 ∀v ∈ V \\ Us (8) Note that we still use h0 u as the trainable node embedding, while the input to the RGCN is augmented to h0+ u .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We fur- ther extract data that only contain the seen nodes to form the seen data: Ds = {(Ci, Ai, yi)|c ∈ Us∀c ∈ Ci}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We, again, use DCCA on Ds to maximize the correlation between the text representation and the code representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' After obtaining the projection matrix, we train on the en- tire dataset D to minimize the prediction loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Note that D contains nodes that do not appear in the DCCA process and are labeled differently from the seen nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The different la- bels allow the RGCN to tell whether a node is unseen during the DCCA process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' If unseen nodes hurt the prediction, it will be reflected in the prediction loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Intuitively, if unseen nodes are less credible, data with more 0-labeled nodes will Method Local Data MIMIC-III DEL DIA TH D30 MORT R30 Corr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 BERT 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ClinicalBERT 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 Bi-LSTM 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 DCCA+Bi-LSTM 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 RGCN 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 DCCA+RGCN 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 RGCN+Bi-LSTM 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 DCCA+RGCN+Bi-LSTM 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 Table 1: DCCA Joint Learning and baseline AUROC (%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Top 4 lines use clinical notes only during inference, middle 2 ICD codes only, and bottom 2 both.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Corr = Sum of correlation of latent representations over 20 dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Method Local Data MIMIC-III DEL DIA TH D30 MORT R30 RGCN 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 RGCN+Labling 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 DCCA+RGCN 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 DCCA+RGCN+Labeling 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 Table 2: Ablation Study of the Labeling Training Scheme under Unseen Code Setting in AUROC (%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' have poor prediction results;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' GNN can capture this charac- teristic and reflect it in the prediction by assigning less pos- itive/negative scores to queries with more 0-labeled nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The labeling training scheme essentially blocks a part of the training code during DCCA and thus obtains embeddings for Us and Uu from different distributions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' And we train on the entire training dataset so that the model learns to handle seen and unseen codes heterogeneously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This setup mimics the actual inference scenario.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Note that despite being differ- ent, the distributions of seen and unseen node embeddings can be similar and overlapped.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Thus, the additional 1-0 la- beling is necessary to differentiate them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 5 Experimental Results Datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use two datasets to evaluate the performance of our framework: Proprietary Dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This dataset con- tains medical records of 38,551 admissions at the local Hos- pital from 2018 to 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Each entry is also associated with a free text procedural description and a set of ICD-10 pro- cedure codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We aim to use our framework to predict a set of post-operative outcomes, including delirium (DEL), dial- ysis (DIA), troponin high (TH), and death in 30 days (D30).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' MIMIC-III dataset (Johnson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This dataset con- tains medical records of 58,976 unique ICU hospital ad- mission from 38,597 patients at the Beth Israel Deaconess Medical Center between 2001 and 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Each admission record is associated with a set of ICD-9 diagnoses codes and multiple clinical notes from different sources, includ- ing case management, consult, ECG, discharge summary, general nursing, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We aim to predict two outcomes from the codes and texts: (1) In-hospital mortality (MORT).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use admissions with hospital expire flag=1 in the MIMIC- III dataset as the positive data and sample the same number of negative data to form the final dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' All clinical notes generated on the last day of admission are filtered out to avoid directly mentioning the outcome.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use all clinical notes ordered by time and take the first 2,500-word tokens as the input text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' (2) 30-day readmission (R30).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We follow (Huang, Altosaar, and Ranganath 2019), label admissions where a patient is readmitted within 30 days as positive, and sample an equal number of negative admissions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Newborn and death admissions are filtered out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We only use clinical notes of type Discharge Summary and take the first 2,500- word tokens as the input text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Sample sizes can be found in Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Effectiveness of DCCA training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We split the dataset with a train/validation/test ratio of 8:1:1 and use 5-fold cross-validation to evaluate our model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' GNN and Bi-LSTM are optimized in the DCCA process using the training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The checkpoint model with the best validation correlation is picked to compute the projection matrix only from the training dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Then we attach an MLP head to the tar- get prediction model (either the GNN or the Bi-LSTM) and fine-tune the model in an end-to-end fashion to minimize the prediction loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For this task, we compare our framework to popular pre- trained models ClinicalBERT and BERT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We also compare it to the base GNN and Bi-LSTM to show the effective- ness of our proposed framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We additionally provide experimental results where both text and code embedding are used to make predictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We compare our model with a vanilla multi-view model without DCCA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For all baselines, we report their Area Under Receiver Operating Characteris- tic (AUROC) as evaluation metrics, and Average Precision (AP) can be found in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For all datasets, we set L, the number of correlated dimensions to 20, and report the total amount of correlation obtained (Corr).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Table 1 shows the main results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For clinical notes predic- tion, we can see that the codes augmented model can con- sistently outperform the base Bi-LSTM, with an average rel- ative performance increase of 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4% on the proprietary data and 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6% on the MIMIC-III data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our proposed method out- performs BERT on most tasks and achieves very competi- tive performance compared to that of ClinicalBERT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Note that our model only trains on the labeled EHR data without unsupervised training on extra data like BERT and Clini- calBERT do.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ClinicalBERT has been previously trained and fine-tuned on the entire MIMIC dataset, including the dis- charge summaries, and therefore these results may overesti- mate its performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For ICD code prediction, we see that DCCA brings a 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5% performance increase on the proprietary data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Since the codes model significantly outperforms the language model on all tasks, the RGCN is a much stronger learner and has less information to learn from the text model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Comparing the results of the proprietary and the MIMIC datasets, we can see that DCCA brings a more significant performance boost to the proprietary dataset, presumably because of the larger amount of correlation obtained in the proprietary dataset (85% versus 58%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Moreover, an important difference in these datasets is the ontology used: MIMIC-III uses ICD-9 and the proprietary dataset uses ICD-10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The ICD-9 ontol- ogy tree has a height of four, which is much smaller than that of ICD-10 and is more coarsely classified.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This may also ex- plain the smaller performance gains in MIMIC-III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The combined model with DCCA only brings a slight per- formance boost compared to the one without because the amount of information for the models to learn is equiva- lent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Nevertheless, the DCCA model encourages the two views’ embeddings to agree and allows independent predic- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In contrast, a vanilla multi-view model does not help the weaker learner learn from the stronger learner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Unseen Codes Experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We identify the set of unique codes in the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We split the codes into k-fold and ran k experiments on each split.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For each experiment, we pick one fold as the unseen code set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Data that contain at least one unseen code are used as the evaluation set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The evaluation set is split into two halves as the valid and test sets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The rest of the data forms the training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We pick an- other fold from the code split as the DCCA unseen code set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Training set data that do not contain any DCCA unseen code form the DCCA training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Then, the entire training set is used for task fine-tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Because the distribution of codes is not uniform, the number of data for each split is not equal across different folds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use k=10 for the proprietary dataset and k=20 for the MIMIC-III dataset to generate a reasonable data division.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We provide average split sizes in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For this task, we compare our method with the base GNN, # Admission # Pos.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Samples # Unique codes DEL 11,064 5,367 5,637 DIA 38,551 1,387 9,320 TH 38,551 1,235 9,320 D30 38,551 1,444 9,320 MORT 5,926 2,963 4,448 R30 10,998 5,499 3,645 Table 3: Statistics of different datasets and tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' base GNN augmented with the same labeling training strat- egy, and DCCA-optimized GNN to demonstrate the out- standing performance of our framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Similarly, we re- port AUROC and include AP in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Table 2 summarizes the results of the unseen codes exper- iments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Note that all test data contain at least one code that never appears in the training process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In such a more diffi- cult inference scenario, comparing the plain RGCN with the DCCA-augmented RGCN, we see a 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2% average relative performance increase on the proprietary dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' With the labeling learning method, we can further improve the per- formance gain to 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' On the MIMIC-III dataset, the per- formance boost of our model over the plain RGCN is 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6%, demonstrating our method’s ability to differentiate seen and unseen codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We also notice that DCCA alone only slightly improves the performance on the MIMIC-III dataset (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We suspect that while the labeling training scheme helps dis- tinguish seen and unseen codes, the number of data used in the DCCA process is also reduced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' As MORT and R30 datasets are smaller and a small DCCA training set may not faithfully represent the actual data distribution, the regular- ization effect of DCCA diminishes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 6 Conclusions Predicting patient outcomes from EHR data is an essen- tial task in clinical ML.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Conventional methods that solely learn from clinical texts suffer from poor performance, and those that learn from codes have limited application in real- world clinical settings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In this paper, we propose a multi- view framework that jointly learns from the clinical notes and ICD codes of EHR data using Bi-LSTM and GNN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use DCCA to create shared information but maintain each view’s independence during inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' This allows accurate prediction using clinical notes when the ICD codes are miss- ing, which is commonly the case in pre-operative analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We also propose a label augmentation method for our frame- work, which allows the GNN model to make effective infer- ences on codes that are not seen during training, enhancing generalizability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Experiments are conducted on two different datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Our methods show consistent effectiveness across tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In the future, we plan to incorporate more data types in the EHR and combine them with other multi-view learning methods to make more accurate predictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' References Alsentzer, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Murphy, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Boag, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Weng, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jin, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Naumann, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and McDermott, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Pub- licly available clinical BERT embeddings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' arXiv preprint arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='03323.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Andrew, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Arora, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bilmes, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Livescu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deep canonical correlation analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In International con- ference on machine learning, 1247–1255.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Boag, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Doss, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Naumann, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Szolovits, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' What’s in a note?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' unpacking predictive value in clinical note representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' AMIA Summits on Translational Science Proceedings, 2018: 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Choi, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Xu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Dusenberry, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Flores, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Xue, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Dai, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Learning the graphical structure of elec- tronic health records with graph convolutional transformer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In Proceedings of the AAAI conference on artificial intelli- gence, volume 34, 606–613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deschepper, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Eeckloo, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Vogelaers, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Waege- man, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A hospital wide predictive model for un- planned readmission using hierarchical ICD data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Computer methods and programs in biomedicine, 173: 177–183.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Devlin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Chang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Toutanova, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' CoRR, abs/1810.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='04805.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Fouad, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Tino, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Raychaudhury, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Schneider, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Incorporating Privileged Information Through Met- ric Learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' IEEE Transactions on Neural Networks and Learning Systems, 24(7): 1086–1098.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Gilmer, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Schoenholz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Riley, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Vinyals, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Dahl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Neural message passing for quantum chemistry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In International conference on machine learn- ing, 1263–1272.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Goldstein, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Navar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Pencina, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Ioan- nidis, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Journal of the American Medical Infor- matics Association, 24(1): 198–208.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hochreiter, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Schmidhuber, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 1997.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Long short-term memory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Neural computation, 9(8): 1735–1780.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Huang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Altosaar, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Ranganath, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Clinical- bert: Modeling clinical notes and predicting hospital read- mission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' arXiv preprint arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='05342.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Huang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Moseley, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deng, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' George, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Lindvall, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Clinical XLNet: Mod- eling sequential clinical notes and predicting prolonged me- chanical ventilation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' arXiv preprint arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='11975.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Johnson, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Pollard, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Shen, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lehman, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Feng, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Ghassemi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Moody, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Szolovits, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' An- thony Celi, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Mark, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' MIMIC-III, a freely accessible critical care database.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Scientific data, 3(1): 1–9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lambert, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Sener, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Savarese, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deep learning under privileged information using heteroscedastic dropout.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 8886–8895.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Lee, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Yoon, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Kim, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Kim, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Kim, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' So, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Kang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' BioBERT: a pre-trained biomedical language representation model for biomedical text mining.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bioinfor- matics, 36(4): 1234–1240.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Yu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ICD coding from clinical text using multi-filter residual convolutional neural network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In Pro- ceedings of the AAAI Conference on Artificial Intelligence, volume 34, 8180–8187.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Han, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Wu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deeper insights into graph convolutional networks for semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In Thirty-Second AAAI conference on artificial intelligence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Ma, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Gao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Integrating Multimodal Electronic Health Records for Diagnosis Prediction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In AMIA Annual Symposium Proceedings, volume 2021, 726.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' American Medical Informatics Association.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Liu, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Stansbury, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Singh, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Ryan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Sukul, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Mahmoudi, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Waljee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zhu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Nallamothu, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Predicting 30-day hospital readmissions using arti- ficial neural networks with medical code embedding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' PloS one, 15(4): e0221606.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Liu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Peng, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' He, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Yu, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Het- erogeneous Similarity Graph Neural Network on Electronic Health Records.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In 2020 IEEE International Conference on Big Data (Big Data), 1196–1205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Mikolov, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Sutskever, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Chen, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Corrado, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Dean, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Distributed representations of words and phrases and their compositionality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Advances in neural in- formation processing systems, 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Pascual, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Luck, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Wattenhofer, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Towards BERT-based automatic ICD coding: Limitations and oppor- tunities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' arXiv preprint arXiv:2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='06709.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Schlichtkrull, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Kipf, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bloem, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Berg, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Titov, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Welling, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Modeling relational data with graph convolutional networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In European semantic web conference, 593–607.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Shickel, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Tighe, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Bihorac, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Rashidi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Deep EHR: a survey of recent advances in deep learn- ing techniques for electronic health record (EHR) analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' IEEE journal of biomedical and health informatics, 22(5): 1589–1604.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Si, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Xu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Roberts, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Enhanc- ing clinical concept extraction with contextual embeddings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Journal of the American Medical Informatics Association, 26(11): 1297–1304.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Vapnik, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Vashist, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A new learning paradigm: Learning using privileged information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Neural networks, 22(5-6): 544–557.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Arora, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Livescu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Bilmes, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' On deep multi-view representation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In International conference on machine learning, 1083–1092.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Wei, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Teixeira, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Mo, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Cronin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Warner, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Denny, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Combining billing codes, clin- ical notes, and medications from electronic health records provides superior phenotyping performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Journal of the American Medical Informatics Association, 23(e1): e20– e27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Xu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hu, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Leskovec, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Jegelka, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' How powerful are graph neural networks?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' arXiv preprint arXiv:1810.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='00826.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Yu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Beam, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Kohane, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Artificial intelligence in healthcare.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Nature biomedical engineering, 2(10): 719–731.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zhang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Yin, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zeng, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Yuan, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Zhang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Combining structured and unstructured data for predictive models: a deep learning approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' BMC medical informatics and decision making, 20(1): 1–11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zhang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' King, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Avidan, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2020b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hierarchical attention propagation for healthcare represen- tation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 249–256.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Jin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' and Zhou, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Understanding bag-of-words model: a statistical framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' International journal of machine learning and cybernetics, 1(1): 43–52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' A Average Precision Score Results AP results demonstrate a similar pattern to AUROC results, where DCCA augmented model can consistently outper- form the base model while achieving very competitive re- sults compared to ClinicalBERT for the text data as shown in Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The proposed labeling training scheme can also consistently improve our model’s performance on the un- seen codes experiments, as shown in Table 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' B Hyperparameters We use grid search for hyperparameter tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' For missing view experiments on text, we fix the number of RGCN layers to be 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We use 32 for all hidden dimensions as we found that varying hidden size has minimal impact on the performance of the data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Text and Code represent the hyperparameters used for text and code inference tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Table 7 summarizes the set of hyperparameters used for tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' C Unseen Code Sample Size We use 10-fold code split for the local data and 20-fold code split for the MIMIC-III data so that the split sizes are reason- able for training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' We report the average number of samples for all tasks in Table 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' DCCA Train Full Train Test DEL 3,458.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 4,624.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 3,219.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 DIA 19,305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 23,717.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 7,416.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 TH 19,305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 23,717.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 7,416.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 D30 19,305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 23,717.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 7,416.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 MORT 4,603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 6,148.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 2,424.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 R30 2,528.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 3,264.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 1,330.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 Table 4: Average Split Size in Unseen Codes Experiment .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' D Data And Implementation We adopted the local dataset because it is the only dataset we have access to that uses both clinical free texts and ICD-10 codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' The implementation details of the MIMIC-III dataset experiments can be found in the supplementary material (code).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Local Data MIMIC-III DEL DIA TH D30 MORT R30 Corr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 BERT 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ClinicalBERT 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 LSTM 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 DCCA+LSTM 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 RGCN 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 DCCA+RGCN 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 RGCN+Bi-LSTM 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 DCCA+RGCN+Bi-LSTM 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 Table 5: Effect of DCCA Joint Learning Compared to Different Baselines in AP (%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Method Local Data MIMIC-III DEL DIA TH D30 MORT R30 RGCN 72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 RGCN+Labling 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 54.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 ± 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 DCCA+RGCN 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 ± 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6 ± 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 DCCA+RGCN+Labeling 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='3 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='7 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='0 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='5 ± 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='9 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='1 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4 Table 6: Ablation Study of the Labeling Training Scheme under Unseen Code Setting in AP (%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content=' Hyperparameter Local-Text Local-Code MIMIC-III-Text MIMIC-III-Code GNN #layers 3 {2,3,4} 3 {2,3,4} LSTM block size(b) 30 30 MLP #layers 2 2 1 1 dropout {0,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4} {0,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4} {0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8} {0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='2,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='4,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='6,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='8} DCCA learning rate 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='001 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='001 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='001 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'} +page_content='001 batch size 1024 1024 400 400 Task learning rate {1e-3,1e-4,1e-5} {1e-3,1e-4,1e-5} {1e-3,1e-4,1e-5} {1e-3,1e-4,1e-5} batch size 256 256 32 32 Table 7: Hyperparameters used for tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/AdFJT4oBgHgl3EQfrS3C/content/2301.11608v1.pdf'}