Background and aims: Familial hypercholesterolemia (FH) is one of the most frequent diseases with monogenic inheritance. Previous data indicated that the heterozygous form occurred in 1:250 people. Based on these reports, around 36,000-40,000 people are estimated to have FH in Hungary, however, there are no exact data about the frequency of the disease in our country. Therefore, we initiated a cooperation with a clinical site partner company that provides modern data mining methods, on the basis of medical and statistical records, and we applied them to two major hospitals in the Northern Great Plain region of Hungary to find patients with a possible diagnosis of FH.
Methods: Medical records of 1,342,124 patients were included in our study. From the mined data, we calculated Dutch Lipid Clinic Network (DLCN) scores for each patient and grouped them according to the criteria to assess the likelihood of the diagnosis of FH. We also calculated the mean lipid levels before the diagnosis and treatment.
Results: We identified 225 patients with a DLCN score of 6-8 (mean total cholesterol: 9.38 ± 3.0 mmol/L, mean LDL-C: 7.61 ± 2.4 mmol/L), and 11,706 patients with a DLCN score of 3-5 (mean total cholesterol: 7.34 ± 1.2 mmol/L, mean LDL-C: 5.26 ± 0.8 mmol/L).
Conclusions: The analysis of more regional and country-wide data and more frequent measurements of total cholesterol and LDL-C levels would increase the number of FH cases discovered. Data mining seems to be ideal for filtering and screening of FH in Hungary.
Keywords: Data mining; Deep learning; Dutch Lipid Clinic Network criteria; Familial hypercholesterolemia; Low-density lipoprotein; Screening.
Copyright © 2018 Elsevier B.V. All rights reserved.