@article{6c0f4950989d428daa125dec6212a5df,
title = "A Knowledge Distillation Ensemble Framework for Predicting Short- and Long-Term Hospitalization Outcomes from Electronic Health Records Data",
abstract = "The ability to perform accurate prognosis is crucial for proactive clinical decision making, informed resource management and personalised care. Existing outcome prediction models suffer from a low recall of infrequent positive outcomes. We present a highly-scalable and robust machine learning framework to automatically predict adversity represented by mortality and ICU admission and readmission from time-series of vital signs and laboratory results obtained within the first 24 hours of hospital admission. The stacked ensemble platform comprises two components: a) an unsupervised LSTM Autoencoder that learns an optimal representation of the time-series, using it to differentiate the less frequent patterns which conclude with an adverse event from the majority patterns that do not, and b) a gradient boosting model, which relies on the constructed representation to refine prediction by incorporating static features. The model is used to assess a patient's risk of adversity and provides visual justifications of its prediction. Results of three case studies show that the model outperforms existing platforms in ICU and general ward settings, achieving average Precision-Recall Areas Under the Curve (PR-AUCs) of 0.891 (95% CI: 0.878-0.939) for mortality and 0.908 (95% CI: 0.870-0.935) in predicting ICU admission and readmission.",
keywords = "Biological system modeling, Clinical Outcome Prediction, Data models, Ensemble Learning, Gradient Boost, Hospitals, Imbalanced time-series, Long Short Term Memory networks (LSTM), Machine Learning, Mortality Prediction, Outlier Detection, Oxygen, Physiology, Predictive models, Stacked Ensemble, Tools",
author = "Ibrahim, {Zina M.} and Daniel Bean and Thomas Searle and Linglong Qian and Honghan Wu and Anthony Shek and Zeljko Kraljevic and James Galloway and Sam Norton and Teo, {James T.} and Dobson, {Richard J. B.}",
note = "Funding Information: The work of Zina M. Ibrahim and Richard JB Dobson was supported by in part by the NIHR Biomedical Research Centre at SLaM, in part by Kings College London, London, U.K., and in part by the NIHR University College London Hospitals Biomedical Research Centre. The work of Richard JB Dobson was also supported in part by Health Data Research (HDR), U.K., and in part by The BigData@Heart Consortium under Grant 116074. The work of Daniel Bean was supported by a UKRI Innovation Fellowship as part of Health Data Research U.K. under Grant MR/S00310X/1. The work of Honghan Wu was supported by MRC and HDR U.K. under Grant MR/S004149/1, and in part by the Wellcome Institutional Translation Partnership Award (PIII054). The work of Anthony Shek was supported by a King?s Medical Research Trust studentship. The work of JTHT was supported by the London AI Medical Imaging Centre for Value-Based Healthcare (AI4VBH), and in part by the NIHR Applied Research Collaboration South London at King?s College Hospital NHS Foundation Trust. Funding Information: Manuscript received November 18, 2020; revised March 26, 2021 and May 22, 2021; accepted June 6, 2021. Date of publication June 15, 2021; date of current version January 5, 2022. The work of Zina M. Ibrahim and Richard JB Dobson was supported by in part by the NIHR Biomedical Research Centre at SLaM, in part by Kings College London, London, U.K., and in part by the NIHR University College London Hospitals Biomedical Research Centre. The work of Richard JB Dobson was also supported in part by Health Data Research (HDR), U.K., and in part by The BigData@Heart Consortium under Grant 116074. The work of Daniel Bean was supported by a UKRI Innovation Fellowship as part of Health Data Research U.K. under Grant MR/S00310X/1. The work of Honghan Wu was supported by MRC and HDR U.K. under Grant MR/S004149/1, and in part by the Wellcome Institutional Translation Partnership Award (PIII054). The work of Anthony Shek was supported by a King{\textquoteright}s Medical Research Trust studentship. The work of JTHT was Publisher Copyright: {\textcopyright} 2013 IEEE.",
year = "2022",
month = jan,
doi = "10.1109/JBHI.2021.3089287",
language = "English",
volume = "26",
pages = "423--435",
journal = "IEEE Journal of Biomedical and Health Informatics",
issn = "2168-2194",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "1",
}