Development of Common Data Elements for Organ Transplantation

Published in the JAMA Network Open. Here is a link to the article.

Regenstrief Institute authors: Katie Ross-Driscoll, PhD, MPH, Rachel E. Patzer, PhD, MPH

Introduction

Organ transplantation research and health policy are largely based on registry data that lack granularity, and recent federal initiatives have called for modernization of the transplant data collection process.1 Common data models (CDM) are standardized metadata language that allow users to share data and their meaning across applications.2 The Consortium for the Holistic Assessment of Risk in Transplant (CHART) developed a CDM to allow data collection that begins at transplant referral for research and quality improvement.3 The purpose of this cohort study is to demonstrate the validity of the CDM.

Methods

The initial variable list was identified based on 26 semi structured interviews with multidisciplinary transplant stakeholders and group discussion at CHART steering committee meetings, then used to establish 368 data elements (eAppendix in Supplement 1). The CDM is organized into clinical domain tables with equivalence mapping variables for cross walking. Data deidentification included randomized patient identifiers and identifiers to enable linking across tables (Figure). The unit of observation is transplant referral, and phases are defined by time from referral. A shift-and-truncate method preserves temporal relations while adhering to privacy standards.4 Area-level social determinants of health (SDOH) were derived by linking patient addresses to a geocoding program written by the coordinating center.5

The initial extraction included adult patients referred for liver, kidney, or pancreas transplants between January 1, 2016, and December 31, 2022; data were collected from 1 year prior to referral through death, loss to follow-up, or administrative censor date. Approval was obtained from Duke University and each collaborating center’s institutional review board prior to data extraction. Extracted data were transmitted securely using a secure file transplant protocol to a Federal Information Security Modernization Act–compliant storage system. Informed consent was not required because no patients were contacted as part of this study and data were deidentified by participating centers prior to transfer to the data coordinating center (Duke University). Descriptive statistics were used to assess data quality for select variables.6 All analyses were performed using R version 4.3.3 (R Project for Statistical Computing). Data were analyzed from August 2023 to October 2024. This cohort study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Results

The cohort includes 91 877 referrals representing 79 787 patients referred for abdominal transplant (mean [SD] age, 54.3 [13.0] years; 32 473 female [40.7%]; 2.9% were Asian, 43.3% were Black, and 43.0% were White) at 7 transplant centers and includes 9 128 093 care encounters, 45 292 686 laboratory values, and 50 972 302 procedures. Missingness for cohort and demographic data elements ranged from 0 to 100%, with high missingness (ie, more than 40%) for other individual SDOH, such as education level and citizenship (Table). Missingness for the transplant phases ranged from 0.4% for transplant to 26.6% for committee review. For cohort and demographic variables, the proportion of values that were implausible ranged from 0% to 0.2%. Discordance of transplant phase temporal order ranged from 2.1% to 6.4% after excluding implausible values.

Discussion

Our findings suggest the feasibility of electronic health record (EHR)–based data assimilation from transplant referral through posttransplant outcomes, including a large collection of SDOH data not currently captured in existing national systems. We found high rates of plausibility and concordance but a wide range of missingness by data element; this reflects the known center-level variation in data collection practices, even among high-volume transplant centers. Missingness was generally clustered by EHR platform and center and decreased as patients progressed through the transplant care continuum. A unified, system-agnostic EHR-based CDM may be a superior alternative to the current registry-based national data collection systems and reduce the significant administrative burden associated with current data collection practices. Our study is limited by validation limited to liver, kidney, and pancreas patients and reports of quality metrics following iterative feedback and consultation with the coordinating center bioinformaticians. Further validation is required to determine whether a national EHR-based CDM for transplant data collection is plausible. We believe this approach should be carefully assessed in collaboration with bioinformaticians and private vendors as national regulatory bodies consider the best data reporting system for US transplant centers in the future.

Authors:

Lisa M. McElroy, MD, MS^1,2; Ursula Rogers, BS¹; Liz Nichols, MS¹; Nrupen A. Bhavsar, PhD, MPH^1,3; Tyler Schappe, MS³; Jessica Harding, PhD⁴; Alexandra T. Strauss, MD, PhD, MIE⁵; Elisa J. Gordon, PhD, MPH⁶; Katie Ross-Driscoll, PhD, MPH⁷; Rhiannon Deirhoi-Reed, PhD, MPH⁸; Norine W. Chan, MD¹; Jackie B. Henson, MD⁹; Yue Harn Ng, MD, MPH¹⁰; Dave J. Taber, PharmD, MS¹¹; Adam J. Milam, MD, PhD¹²; Jesse D. Schold, PhD¹³; Juan Carlos Caicedo, MD¹⁴; Andrew B. Adams, MD, PhD¹⁵; Rachel E. Patzer, PhD, MPH⁷; Samuel I. Berchuck, PhD³; Roland Matsouaka, PhD³; Jennifer Gagnon, BS, RN¹; Allan D. Kirk, MD, PhD¹

Affiliations:

• ¹Department of Surgery, Duke University School of Medicine, Durham, North Carolina
• ²Department of Population Health Sciences, Duke University School of Medicine, Durham, North Carolina
• ³Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, North Carolina
• ⁴Department of Surgery, Emory University School of Medicine, Atlanta, Georgia
• ⁵Department of Medicine, Johns Hopkins University, Baltimore, Maryland
• ⁶Department of Surgery, Vanderbilt University Medical Center, Nashville, Tennessee
• ⁷Center for Health Services Research, Regenstrief Institute, University of Indiana, Indianapolis
• ⁸Department of Surgery, University of Alabama, Birmingham
• ⁹Department of Medicine, Duke University School of Medicine, Durham, North Carolina
• ¹⁰Department of Medicine, University of Washington, Seattle
• ¹¹Department of Surgery, Medical University of South Carolina, Charleston
• ¹²Department of Anesthesiology and Perioperative Medicine, Mayo Clinic, Phoenix, Arizona
• ¹³Departments of Surgery and Epidemiology, University of Colorado Anschutz Medical Campus, Aurora
• ¹⁴Department of Surgery, Northwestern University, Chicago, Illinois
• ¹⁵Department of Surgery, University of Minnesota, Minneapolis