PRISM: Mitigating EHR Data Sparsity via Learning from Missing Feature Calibrated Prototype Patient Representations

TL;DR

PRISM is a novel framework that improves predictive modeling on sparse EHR data by learning from prototype patient representations and assessing feature reliability, outperforming traditional imputation methods.

Contribution

PRISM introduces a prototype-based data imputation approach with a feature confidence learner and a new patient similarity metric tailored for sparse EHR data.

Findings

Outperforms existing methods in mortality and readmission prediction

Effectively handles EHR data sparsity

Demonstrates robustness across multiple datasets

Abstract

Electronic Health Records (EHRs) contain a wealth of patient data; however, the sparsity of EHRs data often presents significant challenges for predictive modeling. Conventional imputation methods inadequately distinguish between real and imputed data, leading to potential inaccuracies of patient representations. To address these issues, we introduce PRISM, a framework that indirectly imputes data by leveraging prototype representations of similar patients, thus ensuring compact representations that preserve patient information. PRISM also includes a feature confidence learner module, which evaluates the reliability of each feature considering missing statuses. Additionally, PRISM introduces a new patient similarity metric that accounts for feature confidence, avoiding over-reliance on imprecise imputed values. Our extensive experiments on the MIMIC-III, MIMIC-IV, PhysioNet Challenge 2012, eICU datasets demonstrate PRISM's superior performance in predicting in-hospital mortality and 30-day readmission tasks, showcasing its effectiveness in handling EHR data sparsity. For the sake of reproducibility and further research, we have publicly released the code at https://github.com/yhzhu99/PRISM.