Explainable Clinical Decision Support from Text

Clinical prediction models often use structured variables and provide outcomes that are not readily interpretable by clinicians. Further, free-text medical notes may contain information not immediately available in structured variables. We propose a hierarchical CNN-transformer model with explicit attention as an interpretable, multi-task clinical language model, which achieves an AUROC of 0.75 and 0.78 on sepsis and mortality prediction, respectively. We also explore the relationships between learned features from structured and unstructured variables using projection-weighted canonical correlation analysis. Finally, we outline a protocol to evaluate model usability in a clinical decision support context. From domain-expert evaluations, our model generates informative rationales that have promising real-life applications.

[1]  Jure Leskovec,et al.  Learning Attitudes and Attributes from Multi-aspect Reviews , 2012, 2012 IEEE 12th International Conference on Data Mining.

[2]  Peter Szolovits,et al.  MIMIC-III, a freely accessible critical care database , 2016, Scientific Data.

[3]  Lukasz Kaiser,et al.  Attention is All you Need , 2017, NIPS.

[4]  John F. Hurdle,et al.  Measuring diagnoses: ICD code accuracy. , 2005, Health services research.

[5]  Phil Culliton,et al.  Predicting Severe Sepsis Using Text from the Electronic Health Record , 2017, ArXiv.

[6]  Małgorzata Marciniak,et al.  Rule-based information extraction from patients' clinical data , 2009, J. Biomed. Informatics.

[7]  M. Howell,et al.  Quick Sepsis‐related Organ Failure Assessment, Systemic Inflammatory Response Syndrome, and Early Warning Scores for Detecting Clinical Deterioration in Infected Patients outside the Intensive Care Unit , 2017, American journal of respiratory and critical care medicine.

[8]  Federico Cabitza,et al.  The Revival of the Notes Field: Leveraging the Unstructured Content in Electronic Health Records , 2019, Front. Med..

[9]  Byron C. Wallace,et al.  Attention is not Explanation , 2019, NAACL.

[10]  Wei-Hung Weng,et al.  Publicly Available Clinical BERT Embeddings , 2019, Proceedings of the 2nd Clinical Natural Language Processing Workshop.

[11]  Shamim Nemati,et al.  An Interpretable Machine Learning Model for Accurate Prediction of Sepsis in the ICU , 2017, Critical care medicine.

[12]  Ming-Wei Chang,et al.  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[13]  Maria Leonor Pacheco,et al.  of the Association for Computational Linguistics: , 2001 .

[14]  Regina Barzilay,et al.  Rationalizing Neural Predictions , 2016, EMNLP.

[15]  Florian Schmidt,et al.  Neural Document Embeddings for Intensive Care Patient Mortality Prediction , 2016, NIPS 2016.

[16]  J. Denny,et al.  Artificial intelligence, bias and clinical safety , 2019, BMJ Quality & Safety.

[17]  Yuan Luo,et al.  Clinical text classification with rule-based features and knowledge-guided convolutional neural networks , 2018, 2018 IEEE International Conference on Healthcare Informatics Workshop (ICHI-W).

[18]  Steven Horng,et al.  Creating an automated trigger for sepsis clinical decision support at emergency department triage using machine learning , 2017, PloS one.

[19]  R'emi Louf,et al.  HuggingFace's Transformers: State-of-the-art Natural Language Processing , 2019, ArXiv.

[20]  Anna Rumshisky,et al.  Unfolding physiological state: mortality modelling in intensive care units , 2014, KDD.

[21]  Yuval Pinter,et al.  Attention is not not Explanation , 2019, EMNLP.

[22]  H. Hotelling.  Relations Between Two Sets of Variates , 1936 .

[23]  Jimeng Sun,et al.  Explainable Prediction of Medical Codes from Clinical Text , 2018, NAACL.

[24]  Jaewoo Kang,et al.  BioBERT: a pre-trained biomedical language representation model for biomedical text mining , 2019, Bioinform..

[25]  Omer Levy,et al.  What Does BERT Look at? An Analysis of BERT’s Attention , 2019, BlackboxNLP@ACL.

[26]  Frank Rudzicz,et al.  Explainable Artificial Intelligence for Safe Intraoperative Decision Support. , 2019, JAMA surgery.

[27]  Dawn Xiaodong Song,et al.  Latent Attention For If-Then Program Synthesis , 2016, NIPS.

[28]  Christine A. Sinsky,et al.  Allocation of Physician Time in Ambulatory Practice: A Time and Motion Study in 4 Specialties , 2016, Annals of Internal Medicine.

[29]  Jascha Sohl-Dickstein,et al.  SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability , 2017, NIPS.

[30]  Samy Bengio,et al.  Insights on representational similarity in neural networks with canonical correlation , 2018, NeurIPS.

[31]  Kirk Roberts,et al.  Deep Patient Representation of Clinical Notes via Multi-Task Learning for Mortality Prediction. , 2019, AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science.

[32]  Uli K. Chettipally,et al.  Prediction of Sepsis in the Intensive Care Unit With Minimal Electronic Health Record Data: A Machine Learning Approach , 2016, JMIR medical informatics.

[33]  Hien Nguyen,et al.  From vital signs to clinical outcomes for patients with sepsis: a machine learning basis for a clinical decision support system , 2014, J. Am. Medical Informatics Assoc..

[34]  Ankur Teredesai,et al.  Interpretable Machine Learning in Healthcare , 2018, 2018 IEEE International Conference on Healthcare Informatics (ICHI).

[35]  Derek C. Angus,et al.  Epidemiology of sepsis: An update , 2001, Critical care medicine.

[36]  William Fleischman,et al.  Prediction of In-hospital Mortality in Emergency Department Patients With Sepsis: A Local Big Data-Driven, Machine Learning Approach. , 2016, Academic emergency medicine : official journal of the Society for Academic Emergency Medicine.