Dataset Information

Ensuring electronic medical record simulation through better training, modeling, and evaluation.

ABSTRACT: OBJECTIVE:Electronic medical records (EMRs) can support medical research and discovery, but privacy risks limit the sharing of such data on a wide scale. Various approaches have been developed to mitigate risk, including record simulation via generative adversarial networks (GANs). While showing promise in certain application domains, GANs lack a principled approach for EMR data that induces subpar simulation. In this article, we improve EMR simulation through a novel pipeline that (1) enhances the learning model, (2) incorporates evaluation criteria for data utility that informs learning, and (3) refines the training process. MATERIALS AND METHODS:We propose a new electronic health record generator using a GAN with a Wasserstein divergence and layer normalization techniques. We designed 2 utility measures to characterize similarity in the structural properties of real and simulated EMRs in the original and latent space, respectively. We applied a filtering strategy to enhance GAN training for low-prevalence clinical concepts. We evaluated the new and existing GANs with utility and privacy measures (membership and disclosure attacks) using billing codes from over 1 million EMRs at Vanderbilt University Medical Center. RESULTS:The proposed model outperformed the state-of-the-art approaches with significant improvement in retaining the nature of real records, including prediction performance and structural properties, without sacrificing privacy. Additionally, the filtering strategy achieved higher utility when the EMR training dataset was small. CONCLUSIONS:These findings illustrate that EMR simulation through GANs can be substantially improved through more appropriate training, modeling, and evaluation criteria.

SUBMITTER: Zhang Z

PROVIDER: S-EPMC6913223 | biostudies-literature | 2020 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Ensuring electronic medical record simulation through better training, modeling, and evaluation.

Zhang Ziqi Z Yan Chao C Mesa Diego A DA Sun Jimeng J Malin Bradley A BA

Journal of the American Medical Informatics Association : JAMIA 20200101 1

<h4>Objective</h4>Electronic medical records (EMRs) can support medical research and discovery, but privacy risks limit the sharing of such data on a wide scale. Various approaches have been developed to mitigate risk, including record simulation via generative adversarial networks (GANs). While showing promise in certain application domains, GANs lack a principled approach for EMR data that induces subpar simulation. In this article, we improve EMR simulation through a novel pipeline that (1) e ...[more]

PMID: 31592533

Similar Datasets

Project description:Background/aimCost-efficient methods are essential for successful participant recruitment in clinical trials. Patient portal messages are an emerging means of recruiting potentially eligible patients into trials. We assessed the response rate and complaint rate from direct-to-patient, targeted recruitment through patient portals of an electronic medical record for a clinical trial, and compared response rates by differences in message content.MethodsThe Study to Understand Fall Reduction and Vitamin D in You (STURDY) trial is a National Institutes of Health-sponsored, community-based study of vitamin D supplementation for fall prevention in older adults conducted at Johns Hopkins. Potential participants were identified using the Epic electronic medical record at the Johns Hopkins Health System based on age (≥70 years), ZIP code (30-mile radius of study site), and prior activation of a patient portal account. We prepared a shorter message and a longer message. Both had basic information about study participation, but the longer message also contained information about the significance of the study and a personal invitation from the STURDY principal investigator. The Hopkins Institutional Review Board did not require prior consent from the patient or their providers. We calculated the response rate and tracked the number of complaints and requests for removal from future messages. We also determined response rate according to message content.ResultsOf the 5.5 million individuals receiving care at the Johns Hopkins Health System, a sample of 6896 met our inclusion criteria and were sent one patient portal recruitment message between 6 April 2017 and 3 August 2017. Assessment of enrollment by this method ended on 1 December 2017. There were 116 patients who expressed interest in the study (response rate: 1.7%). Twelve (0.2%) recipients were randomized. There were two complaints (0.03%) and one request to unsubscribe from future recruitment messages (0.01%). Response rate was higher with the longer message than the shorter message (2.1% vs 1.2%; p = 0.005).ConclusionPatient portal messages inviting seniors to participate in a randomized controlled trial resulted in a response rate similar to commercial email marketing and resulted in very few complaints or opt-out requests. Furthermore, a longer message with more content enhanced response rate. Recruitment through patient portals might be an effective strategy to enroll trial participants.

Dataset Information

Ensuring electronic medical record simulation through better training, modeling, and evaluation.

Publications

Ensuring electronic medical record simulation through better training, modeling, and evaluation.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure