Abstract P100: Identification of Patients with Familial Hypercholesterolemia (FH) Using the Dutch Lipid Network (DLN) Criteria in Electronic Health Records (EHR)
Introduction: Heterozygous FH is common but under-diagnosed. No uniformly accepted diagnostic criteria for FH exist, and the ability of current algorithms to identify patients using EHR data remains largely untested. Here we present results and challenges in using the DLN criteria to identify FH patients in an EHR.
Methods: In the Vanderbilt University Medical Center de-identified EHR, patients >18 years with 1+ low-density lipoprotein cholesterol (LDL-C) measure from 1996 to 2014 were eligible for study. Using diagnostic and procedure codes, narrative text from clinical care, and laboratory data, patients were assigned points for DLN components (see Table), which were summed to classify definite (DLN≥8) or probable (DLN=6-7) FH. LDL-C values were extracted directly from structured laboratory records. Evidence of physical signs or testing for LDL-C receptor mutations were assessed via keyword searches with simple negation detection rules, the latter including vendor names known to conduct such mutation analyses.
Results: Among 218,652 eligible patients, 10,812 (5%) had 1+ LDL-C>190 mg/dl. We identified 622 probable and 430 definite FH patients (overall prevalence 0.48%). Family history was not ascertainable in a systematic way for most patients and thus not used in scoring. Physical exam signs minimally impacted DLN scores as they were noted infrequently. No definite or probable FH patient records included mention of LDL-C receptor mutation status. Only 5 definite and 3 probable FH patients had narrative keywords noting the presence or suspicion of FH.
Conclusions: EHRs may be useful to identify FH patients but existing algorithms lack sensitivity when applied to EHRs as they rely on key elements rarely captured. Further, the sizable population with LDL-C>190 suggests that FH in this group may not be adequately captured by the DLN criteria. Work is needed to determine how best to utilize EHRs to reliably ascertain FH, via development of new algorithms or by clarifying the positive predictive value of existing algorithms.
Author Disclosures: S.S. Cohen: C. Other Research Support; Modest; Amgen Inc. J. Shirey-Rice: None. J. Hardin: A. Employment; Significant; Amgen Inc. K. Monda: A. Employment; Significant; Amgen Inc. J.P. Fryzek: C. Other Research Support; Modest; Amgen Inc. S. Fazio: None. J.C. Denny: None. W. Wei: None. L. Lipworth: None.
- © 2016 by American Heart Association, Inc.