Abstract 11337: Sensitivity Analysis of an Electronic Health Record-Based Algorithm to Facilitate Detection of Familial Hypercholesterolemia: Results in Genetically Confirmed Familial Hypercholesterolemia

Abstract

Introduction: Familial hypercholesterolemia (FH) is a genetic disorder resulting in elevated LDL-C levels and increased risk of cardiovascular disease. Algorithms have been developed to facilitate FH detection. We previously developed a continuous, electronic health record (EHR)-based algorithm using the Dutch Lipid Clinical Network (DLCN) criteria, with automated correction of LDL-C levels for use of lipid-lowering-therapy (LLT). In this study, we performed a sensitivity analysis of the algorithm in patients with previously genetically confirmed FH. Methods: We selected all patients with genetically confirmed FH, seen during 2018 and 2019. DLCN scores were calculated before and after correction for LLT at the most recent visit. The primary outcome was the number of FH patients with DLCN criteria ≥6 points for LDL-C (which would have been an indication for genetic testing) before and after correction for LLT. Secondary endpoint was the additional number of patients with ≥6 points after also adding data on patient- and family history. Results: We included 222 patients with genetically confirmed FH; mean (SD) age: 50 (17) years; female: n=134 (60,4%). In total, 210 patients (94,6%) used some type of LLT. Mean (SD) LDL-C was 3,0 (1,4) mmol/L before correction for LLT and 7,5 (3,7) mmol/L after correction for LLT. Before correction for LLT, the algorithm identified 3 patients (1,4%) with ≥6 DLCN points, increasing to 59 patients (26,6%) after correction for LLT and to 76 patients (34,2%) after adding data on patient-, and family history. Conclusion: We conclude that, even after correction for LLT and addition of data on patient-, and family history, the sensitivity of a DLCN-based algorithm appeared relatively low in a population with genetically confirmed FH. Correction for LLT is important to increase sensitivity of the algorithm. Interestingly, FH detection algorithms inevitably rely on phenotypical characteristics of patients, whereas the present data show that genotypical FH does not necessarily express as phenotypical FH recognized by the DLCN-based algorithm. These data highlight the need for improvement of the DLCN-based algorithm and the importance of cascade testing for FH.