ACHD Risk Score: A tool for identifying adults with moderate or complex congenital heart defects using Electronic Health Records data from the Emory Healthcare Data Warehouse, Atlanta, GA Open Access

Diallo, Alpha Oumar (2015)

Permanent URL:


Background: It is hypothesized that adults with moderate to complex congenital heart defects (CHD) are increasing, but many patients experience lapses in specialist care. There is, to date, no validated procedure to identify this population.

Objectives: To develop and validate a risk score to identify adults aged 20-60 years old with moderate to complex CHD from routine provider and health system electronic health records (EHR).

Methods: We used a case-control design (596 adults with physician-diagnosed moderate to complex CHDs receiving care at Emory's adult CHD clinic and 2,384 controls [persons without ICD-9 codes for CHD] receiving care at other Emory facilities]). We extracted data regarding age, race/ethnicity, EKG, and laboratory tests from routine outpatient visits between January 2009 and December 2012 from Emory Healthcare's EHR Data Warehouse. We used multivariable logistic regression models and a split-sample (4:1 ratio) approach to develop and validate the risk score, respectively. We generated receiver operating characteristic (ROC) curves to assess the ability of models to predict adult moderate to complex CHD.

Results: Three models (laboratory, non-laboratory, and simplified) were produced and validated internally. The non-laboratory algorithm (ACHD model) based on age, sex, and electrocardiogram markers was chosen. Validation studies of the ACHD model showed a ROC c-statistic of 0.97 [95% Confidence interval (CI): 0.95, 0.99]. The ACHD Risk Score, developed using the ACHD model, also demonstrated good accuracy with 93.69% sensitivity and positive predictive value of 69.80% at a score threshold of 11.

Conclusion: A simple non-laboratory risk score based on age, sex, and EKG marker may help accurately identify adults with moderate to complex CHD from routine EHR systems. External validation studies within large longitudinal clinical cohorts are required to assess wider performance of this tool.

Table of Contents



Data Sources and Study Design...10

Study Population...10

Data Collection...11

Study variables...11

Statistical analysis...12

Risk score development...14

Determination of Risk Score Cut-off...14



Work Cited...20


Figures and Appendix...31

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
Subfield / Discipline
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files