Bayesian averaging over decision tree models: an application for estimating uncertainty in trauma severity scoring

2.50
Hdl Handle:
http://hdl.handle.net/10547/622483
Title:
Bayesian averaging over decision tree models: an application for estimating uncertainty in trauma severity scoring
Authors:
Schetinin, Vitaly ( 0000-0003-1826-0153 ) ; Jakaite, Livija; Krzanowski, Wojtek
Abstract:
Introduction For making reliable decisions, practitioners need to estimate uncertainties that exist in data and decision models. In this paper we analyse uncertainties of predicting survival probability for patients in trauma care. The existing prediction methodology employs logistic regression modelling of Trauma and Injury Severity Score(external) (TRISS), which is based on theoretical assumptions. These assumptions limit the capability of TRISS methodology to provide accurate and reliable predictions. Methods We adopt the methodology of Bayesian model averaging and show how this methodology can be applied to decision trees in order to provide practitioners with new insights into the uncertainty. The proposed method has been validated on a large set of 447,176 cases registered in the US National Trauma Data Bank in terms of discrimination ability evaluated with receiver operating characteristic (ROC) and precision–recall (PRC) curves. Results Areas under curves were improved for ROC from 0.951 to 0.956 (p = 3.89 × 10−18) and for PRC from 0.564 to 0.605 (p = 3.89 × 10−18). The new model has significantly better calibration in terms of the Hosmer–Lemeshow Hˆ" role="presentation"> statistic, showing an improvement from 223.14 (the standard method) to 11.59 (p = 2.31 × 10−18). Conclusion The proposed Bayesian method is capable of improving the accuracy and reliability of survival prediction. The new method has been made available for evaluation purposes as a web application.
Affiliation:
University of Bedfordshire; University of Exeter
Citation:
Schetinin V, Jakaite L, Krzanowski W (2018) 'Bayesian averaging over decision tree models: an application for estimating uncertainty in trauma severity scoring', International Journal of Medical Informatics, 112, pp.6-14.
Publisher:
Elsevier
Journal:
International Journal of Medical Informatics
Issue Date:
11-Jan-2018
URI:
http://hdl.handle.net/10547/622483
DOI:
10.1016/j.ijmedinf.2018.01.009
Additional Links:
https://www.sciencedirect.com/science/article/pii/S1386505618300091
Type:
Article
Language:
en
ISSN:
1386-5056
Appears in Collections:
Computing

Full metadata record

DC FieldValue Language
dc.contributor.authorSchetinin, Vitalyen
dc.contributor.authorJakaite, Livijaen
dc.contributor.authorKrzanowski, Wojteken
dc.date.accessioned2018-02-12T10:26:34Z-
dc.date.available2018-02-12T10:26:34Z-
dc.date.issued2018-01-11-
dc.identifier.citationSchetinin V, Jakaite L, Krzanowski W (2018) 'Bayesian averaging over decision tree models: an application for estimating uncertainty in trauma severity scoring', International Journal of Medical Informatics, 112, pp.6-14.en
dc.identifier.issn1386-5056-
dc.identifier.doi10.1016/j.ijmedinf.2018.01.009-
dc.identifier.urihttp://hdl.handle.net/10547/622483-
dc.description.abstractIntroduction For making reliable decisions, practitioners need to estimate uncertainties that exist in data and decision models. In this paper we analyse uncertainties of predicting survival probability for patients in trauma care. The existing prediction methodology employs logistic regression modelling of Trauma and Injury Severity Score(external) (TRISS), which is based on theoretical assumptions. These assumptions limit the capability of TRISS methodology to provide accurate and reliable predictions. Methods We adopt the methodology of Bayesian model averaging and show how this methodology can be applied to decision trees in order to provide practitioners with new insights into the uncertainty. The proposed method has been validated on a large set of 447,176 cases registered in the US National Trauma Data Bank in terms of discrimination ability evaluated with receiver operating characteristic (ROC) and precision–recall (PRC) curves. Results Areas under curves were improved for ROC from 0.951 to 0.956 (p = 3.89 × 10−18) and for PRC from 0.564 to 0.605 (p = 3.89 × 10−18). The new model has significantly better calibration in terms of the Hosmer–Lemeshow Hˆ" role="presentation"> statistic, showing an improvement from 223.14 (the standard method) to 11.59 (p = 2.31 × 10−18). Conclusion The proposed Bayesian method is capable of improving the accuracy and reliability of survival prediction. The new method has been made available for evaluation purposes as a web application.en
dc.language.isoenen
dc.publisherElsevieren
dc.relation.urlhttps://www.sciencedirect.com/science/article/pii/S1386505618300091en
dc.rightsGreen - can archive pre-print and post-print or publisher's version/PDF-
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/*
dc.subjectBayesian model averagingen
dc.subjectdecision treesen
dc.subjectpredictive posterior distributionen
dc.subjecttrauma and injury severity scoringen
dc.titleBayesian averaging over decision tree models: an application for estimating uncertainty in trauma severity scoringen
dc.typeArticleen
dc.contributor.departmentUniversity of Bedfordshireen
dc.contributor.departmentUniversity of Exeteren
dc.identifier.journalInternational Journal of Medical Informaticsen
dc.date.updated2018-02-12T10:21:23Z-
This item is licensed under a Creative Commons License
Creative Commons
All Items in UOBREP are protected by copyright, with all rights reserved, unless otherwise indicated.