Mortality Prediction with Adaptive Feature Importance Recalibration for Peritoneal Dialysis Patients

Liantao Ma; Chaohe Zhang; Junyi Gao; Xianfeng Jiao; Zhihao Yu; Yinghao Zhu; Tianlong Wang; Xinyu Ma; Yasha Wang; Wen Tang; Xinju Zhao; Wenjie Ruan; Tao Wang

doi:10.1016/j.patter.2023.100892

Mortality Prediction with Adaptive Feature Importance Recalibration for Peritoneal Dialysis Patients

Liantao Ma, Chaohe Zhang, Junyi Gao^*, Xianfeng Jiao, Zhihao Yu, Yinghao Zhu, Tianlong Wang, Xinyu Ma, Yasha Wang^*, Wen Tang^*, Xinju Zhao, Wenjie Ruan, Tao Wang

^*Corresponding author for this work

School of Population Health Sciences

Research output: Contribution to journal › Article › peer-review

Abstract

The study aims to develop AICare, an interpretable mortality prediction model, using Electronic Medical Records (EMR) from follow-up visits for End-Stage Renal Disease (ESRD) patients. AICare includes a multi-channel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform a personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AI Care outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships, variations in feature importance, and provides reference values. An AI-Doctor interaction system is developed for visualizing patients’ health trajectories and risk indicators.

Original language	English
Article number	100892
Journal	Patterns
Volume	4
Issue number	12
DOIs	https://doi.org/10.1016/j.patter.2023.100892
Publication status	Published - 8 Dec 2023

Access to Document

10.1016/j.patter.2023.100892Licence: Creative Commons: Attribution (CC-BY)

Mortality prediction_MA_DOA11112023_VOR_CC-BYFinal published version, 5.51 MBLicence: Creative Commons: Attribution (CC-BY)

Embargoed Document

Mortality_Prediction_MA_DOA11112023_AAM
Accepted author manuscript, 4.68 MB

Cite this

@article{ffd56af1502c49cda0ab517bb3096cec,

title = "Mortality Prediction with Adaptive Feature Importance Recalibration for Peritoneal Dialysis Patients",

abstract = "The study aims to develop AICare, an interpretable mortality prediction model, using Electronic Medical Records (EMR) from follow-up visits for End-Stage Renal Disease (ESRD) patients. AICare includes a multi-channel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform a personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AI Care outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships, variations in feature importance, and provides reference values. An AI-Doctor interaction system is developed for visualizing patients{\textquoteright} health trajectories and risk indicators.",

author = "Liantao Ma and Chaohe Zhang and Junyi Gao and Xianfeng Jiao and Zhihao Yu and Yinghao Zhu and Tianlong Wang and Xinyu Ma and Yasha Wang and Wen Tang and Xinju Zhao and Wenjie Ruan and Tao Wang",

note = "Funding Information: This work was supported by the National Natural Science Foundation of China ( 82241052 ). W.T. was supported by the Fund from Peking University Third Hospital ( BYSYDL2023004 ) and the PKU-Baidu Foundation ( 2020BD030 ). L.M. was supported by the China Postdoctoral Science Foundation ( 2021TQ0011 and 2022M720237 ). X.Z. was supported by the Research Project of Blood Purification Center Branch of Chinese Hospital Association ( CHABP2021-11 ). W.R. was supported by the UK EPSRC project on Offshore Robotics for Certification of Assets (ORCA) ( EP/R026173/1 ). J.G. acknowledges the receipt of studentship awards from the Health Data Research UK-The Alan Turing Institute Wellcome PhD Program in Health Data Science (grant ref. 218529/Z/19/Z ). Publisher Copyright: {\textcopyright} 2023 The Authors",

year = "2023",

month = dec,

day = "8",

doi = "10.1016/j.patter.2023.100892",

language = "English",

volume = "4",

journal = "Patterns",

issn = "2666-3899",

publisher = "Cell Press",

number = "12",

}

TY - JOUR

T1 - Mortality Prediction with Adaptive Feature Importance Recalibration for Peritoneal Dialysis Patients

AU - Ma, Liantao

AU - Zhang, Chaohe

AU - Gao, Junyi

AU - Jiao, Xianfeng

AU - Yu, Zhihao

AU - Zhu, Yinghao

AU - Wang, Tianlong

AU - Ma, Xinyu

AU - Wang, Yasha

AU - Tang, Wen

AU - Zhao, Xinju

AU - Ruan, Wenjie

AU - Wang, Tao

N1 - Funding Information: This work was supported by the National Natural Science Foundation of China ( 82241052 ). W.T. was supported by the Fund from Peking University Third Hospital ( BYSYDL2023004 ) and the PKU-Baidu Foundation ( 2020BD030 ). L.M. was supported by the China Postdoctoral Science Foundation ( 2021TQ0011 and 2022M720237 ). X.Z. was supported by the Research Project of Blood Purification Center Branch of Chinese Hospital Association ( CHABP2021-11 ). W.R. was supported by the UK EPSRC project on Offshore Robotics for Certification of Assets (ORCA) ( EP/R026173/1 ). J.G. acknowledges the receipt of studentship awards from the Health Data Research UK-The Alan Turing Institute Wellcome PhD Program in Health Data Science (grant ref. 218529/Z/19/Z ). Publisher Copyright: © 2023 The Authors

PY - 2023/12/8

Y1 - 2023/12/8

N2 - The study aims to develop AICare, an interpretable mortality prediction model, using Electronic Medical Records (EMR) from follow-up visits for End-Stage Renal Disease (ESRD) patients. AICare includes a multi-channel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform a personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AI Care outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships, variations in feature importance, and provides reference values. An AI-Doctor interaction system is developed for visualizing patients’ health trajectories and risk indicators.

AB - The study aims to develop AICare, an interpretable mortality prediction model, using Electronic Medical Records (EMR) from follow-up visits for End-Stage Renal Disease (ESRD) patients. AICare includes a multi-channel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform a personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AI Care outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships, variations in feature importance, and provides reference values. An AI-Doctor interaction system is developed for visualizing patients’ health trajectories and risk indicators.

U2 - 10.1016/j.patter.2023.100892

DO - 10.1016/j.patter.2023.100892

M3 - Article

SN - 2666-3899

VL - 4

JO - Patterns

JF - Patterns

IS - 12

M1 - 100892

ER -

Mortality Prediction with Adaptive Feature Importance Recalibration for Peritoneal Dialysis Patients

Abstract

Access to Document

Embargoed Document

Fingerprint

Cite this