A Deep Learning Based Unsupervised Method to Impute Missing Values in Patient Records for Improved Management of Cardiovascular Patients

Physicians increasingly depend on electronic health records (EHRs) to manage patients. However, many patient records have substantial missing values that pose a fundamental challenge to their clinical use. To address this prevailing challenge, we propose an unsupervised deep-learning method that can facilitate physicians’ use of EHRs to improve their management of cardiovascular patients. We build on the deep autoencoder framework and develop a novel method to impute missing values in patient records. To demonstrate its clinical applicability and value, we evaluate the proposed method's imputation effectiveness and predictive efficacy using data from cardiovascular patients, in comparison with several prevalent imputation techniques. By reducing imputation errors, our method empowers the use of EHRs and health analytics for improved management of cardiovascular patients. The evaluation results reinforce the importance of properly addressing missing values in patient records and illustrate how effective imputations can achieve greater predictive efficacy for assessing important patient outcomes. Our proposed method better imputes missing values in cardiovascular patient records than several prevalent techniques. Supported by this method, physicians can gain effectiveness in their patient management, because they can make timely outcome estimations (predictions) and therapeutic treatment assessments, enabled by the imputed patient data.