Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma

<h3>Background</h3><p dir="ltr">Predictive modeling with longitudinal electronic health record (EHR) data offers great promise for accelerating personalized medicine and better informs clinical decision-making. Recently, deep learning models have achieved state-of-the-art...

Full description

Saved in:
Bibliographic Details
Main Author: Rawan AlSaad (14159019) (author)
Other Authors: Qutaibah Malluhi (3158757) (author), Ibrahim Janahi (3744933) (author), Sabri Boughorbel (846228) (author)
Published: 2019
Subjects:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1864513566608982016
author Rawan AlSaad (14159019)
author2 Qutaibah Malluhi (3158757)
Ibrahim Janahi (3744933)
Sabri Boughorbel (846228)
author2_role author
author
author
author_facet Rawan AlSaad (14159019)
Qutaibah Malluhi (3158757)
Ibrahim Janahi (3744933)
Sabri Boughorbel (846228)
author_role author
dc.creator.none.fl_str_mv Rawan AlSaad (14159019)
Qutaibah Malluhi (3158757)
Ibrahim Janahi (3744933)
Sabri Boughorbel (846228)
dc.date.none.fl_str_mv 2019-11-08T06:00:00Z
dc.identifier.none.fl_str_mv 10.1186/s12911-019-0951-4
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Interpreting_patient-Specific_risk_prediction_using_contextual_decomposition_of_BiLSTMs_application_to_children_with_asthma/21598380
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Health sciences
Health services and systems
Information and computing sciences
Machine learning
Interpretability
Deep learning
Predictive models
Electronic health record
dc.title.none.fl_str_mv Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <h3>Background</h3><p dir="ltr">Predictive modeling with longitudinal electronic health record (EHR) data offers great promise for accelerating personalized medicine and better informs clinical decision-making. Recently, deep learning models have achieved state-of-the-art performance for many healthcare prediction tasks. However, deep models lack interpretability, which is integral to successful decision-making and can lead to better patient care. In this paper, we build upon the contextual decomposition (CD) method, an algorithm for producing importance scores from long short-term memory networks (LSTMs). We extend the method to bidirectional LSTMs (BiLSTMs) and use it in the context of predicting future clinical outcomes using patients’ EHR historical visits.</p><h3>Methods</h3><p dir="ltr">We use a real EHR dataset comprising 11071 patients, to evaluate and compare CD interpretations from LSTM and BiLSTM models. First, we train LSTM and BiLSTM models for the task of predicting which pre-school children with respiratory system-related complications will have asthma at school-age. After that, we conduct quantitative and qualitative analysis to evaluate the CD interpretations produced by the contextual decomposition of the trained models. In addition, we develop an interactive visualization to demonstrate the utility of CD scores in explaining predicted outcomes.</p><h3>Results</h3><p dir="ltr">Our experimental evaluation demonstrate that whenever a clear visit-level pattern exists, the models learn that pattern and the contextual decomposition can appropriately attribute the prediction to the correct pattern. In addition, the results confirm that the CD scores agree to a large extent with the importance scores generated using logistic regression coefficients. Our main insight was that rather than interpreting the attribution of individual visits to the predicted outcome, we could instead attribute a model’s prediction to a group of visits.</p><h3>Conclusion</h3><p dir="ltr">We presented a quantitative and qualitative evidence that CD interpretations can explain patient-specific predictions using CD attributions of individual visits or a group of visits.</p><h2>Other Information</h2><p dir="ltr">Published in: BMC Medical Informatics and Decision Making<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="http://dx.doi.org/10.1186/s12911-019-0951-4" target="_blank">http://dx.doi.org/10.1186/s12911-019-0951-4</a></p>
eu_rights_str_mv openAccess
id Manara2_fabb9b5ebfa3bc7b416f702956af0749
identifier_str_mv 10.1186/s12911-019-0951-4
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/21598380
publishDate 2019
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthmaRawan AlSaad (14159019)Qutaibah Malluhi (3158757)Ibrahim Janahi (3744933)Sabri Boughorbel (846228)Health sciencesHealth services and systemsInformation and computing sciencesMachine learningInterpretabilityDeep learningPredictive modelsElectronic health record<h3>Background</h3><p dir="ltr">Predictive modeling with longitudinal electronic health record (EHR) data offers great promise for accelerating personalized medicine and better informs clinical decision-making. Recently, deep learning models have achieved state-of-the-art performance for many healthcare prediction tasks. However, deep models lack interpretability, which is integral to successful decision-making and can lead to better patient care. In this paper, we build upon the contextual decomposition (CD) method, an algorithm for producing importance scores from long short-term memory networks (LSTMs). We extend the method to bidirectional LSTMs (BiLSTMs) and use it in the context of predicting future clinical outcomes using patients’ EHR historical visits.</p><h3>Methods</h3><p dir="ltr">We use a real EHR dataset comprising 11071 patients, to evaluate and compare CD interpretations from LSTM and BiLSTM models. First, we train LSTM and BiLSTM models for the task of predicting which pre-school children with respiratory system-related complications will have asthma at school-age. After that, we conduct quantitative and qualitative analysis to evaluate the CD interpretations produced by the contextual decomposition of the trained models. In addition, we develop an interactive visualization to demonstrate the utility of CD scores in explaining predicted outcomes.</p><h3>Results</h3><p dir="ltr">Our experimental evaluation demonstrate that whenever a clear visit-level pattern exists, the models learn that pattern and the contextual decomposition can appropriately attribute the prediction to the correct pattern. In addition, the results confirm that the CD scores agree to a large extent with the importance scores generated using logistic regression coefficients. Our main insight was that rather than interpreting the attribution of individual visits to the predicted outcome, we could instead attribute a model’s prediction to a group of visits.</p><h3>Conclusion</h3><p dir="ltr">We presented a quantitative and qualitative evidence that CD interpretations can explain patient-specific predictions using CD attributions of individual visits or a group of visits.</p><h2>Other Information</h2><p dir="ltr">Published in: BMC Medical Informatics and Decision Making<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="http://dx.doi.org/10.1186/s12911-019-0951-4" target="_blank">http://dx.doi.org/10.1186/s12911-019-0951-4</a></p>2019-11-08T06:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1186/s12911-019-0951-4https://figshare.com/articles/journal_contribution/Interpreting_patient-Specific_risk_prediction_using_contextual_decomposition_of_BiLSTMs_application_to_children_with_asthma/21598380CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/215983802019-11-08T06:00:00Z
spellingShingle Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
Rawan AlSaad (14159019)
Health sciences
Health services and systems
Information and computing sciences
Machine learning
Interpretability
Deep learning
Predictive models
Electronic health record
status_str publishedVersion
title Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
title_full Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
title_fullStr Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
title_full_unstemmed Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
title_short Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
title_sort Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma
topic Health sciences
Health services and systems
Information and computing sciences
Machine learning
Interpretability
Deep learning
Predictive models
Electronic health record