Abstract: TH-PO480
Identifying Sudden Cardiac Death using Electronic Health Records
Session Information
- CKD: Epidemiology, Outcomes - Cardiovascular - I
November 02, 2017 | Location: Hall H, Morial Convention Center
Abstract Time: 10:00 AM - 10:00 AM
Category: Chronic Kidney Disease (Non-Dialysis)
- 303 CKD: Epidemiology, Outcomes - Cardiovascular
Authors
- Makar, Melissa, Duke University, Durham, North Carolina, United States
- Pun, Patrick H., Duke University, Durham, North Carolina, United States
Background
Sudden cardiac death (SCD) is a leading cause of mortality in the US, and disproportionally affects CKD patients. Ascertaining SCD events in electronic health records (EHRs) allows for a better understanding of modifiable risk factors and prevention; however, the complexity of the SCD phenotype makes accurate identification difficult using ICD-9 data alone. By comparing the identification of SCD using a non-machine algorithm versus physician adjudication, we aim to standardize the process of identifying cases of sudden cardiac death in large cohorts as the first step in ultimately reducing SCD incidence among CKD patients.
Methods
An automated search for ICD-9 code 427.5 (cardiac arrest) was applied to the Duke Databank for Cardiovascular Disease, a cohort of over 36,000 patients at a single institution that have undergone cardiac catheterization over the past twenty years. Non-physician, trained staff then applied an algorithm to review the EHRs of detected cases. This algorithm used questions about location, timing, and patient details to guide reviewers to pinpoint SCD cases. This was compared to the gold standard of adjudication by physicians using death certificate data, discharge summaries, and eye witness accounts.
Results
The ICD-9 search identified 1,334 potential events. Of these, 617 unique cases remained after excluding duplicate and missing records. Applying the algorithm then narrowed this to 209 cases of true sudden cardiac arrest events. A total of 88 of the 209 cardiac arrests (42%) resulted in death within 24 hours and qualified as SCD. In comparison, only 63 of the 88 cases (72%) were identified as SCD by the physician adjudication method. Additional information gathered identified the type of arrhythmia (43% VF/VT; 45% PEA or asystole; 12% other) and the type of death (75% witnessed/resuscitation attempted, 25% unobserved).
Conclusion
On its own, ICD-9 data mining to identify SCD is fraught with false positives. Adding an algorithm, based on a precise definition of sudden cardiac death, improves detection of SCD cases. This systematic approach compares favorably to the gold standard of physician adjudication. By following a standard approach to isolate cases of SCD in large cohorts, we hope to delineate a clear phenotype that can then be studied further to eventually lead to reduced SCD incidence among CKD patients.
Funding
- Other NIH Support