论文信息 - eXITs: An Ensemble Approach for Imputing Missing EHR Data

eXITs: An Ensemble Approach for Imputing Missing EHR Data

Missing data points are prevalent in electronic health records (EHRs) and are an impedance to utilizing machine learning for predictive and classification tasks in healthcare. For this challenge, we developed eXITs - a stacked ensemble learner that employs 6 base models to perform imputation on time series data from 13 different laboratory tests across 8, 267 patients in the MIMIC-III database provided in the ICHI 2019 Data Analytics Challenge on Missing Data Imputation (DACMI). The results show that our ensemble model (avg. nRMSE = 0.200) outperforms the reference model, 3D-MICE (avg. nRMSE = 0.222) by 9.69%.

[1] Tianqi Chen,et al. XGBoost: A Scalable Tree Boosting System , 2016, KDD.

[2] Yan Liu,et al. Recurrent Neural Networks for Multivariate Time Series with Missing Values , 2016, Scientific Reports.

[3] Ping Zhang,et al. Risk Prediction with Electronic Health Records: A Deep Learning Approach , 2016, SDM.

[4] Li Li,et al. Deep Patient: An Unsupervised Representation to Predict the Future of Patients from the Electronic Health Records , 2016, Scientific Reports.

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

[6] Wei Cao,et al. BRITS: Bidirectional Recurrent Imputation for Time Series , 2018, NeurIPS.

[7] Jeffrey Dean,et al. Scalable and accurate deep learning with electronic health records , 2018, npj Digital Medicine.

[8] Peter Szolovits,et al. 3D-MICE: integration of cross-sectional and longitudinal imputation for multi-analyte longitudinal clinical data , 2017, J. Am. Medical Informatics Assoc..