Cluster analysis of the 2822 patients with heart failure included in the Multicenter French Survey OFICSEL

Background Heart Failure (HF) is a major public health problem resulting in high rates of hospitalization and mortality. So far, most HF Surveys have included a selected population of patients with HF and involved mainly one type of cardiologic practice. There is a need for more information on the whole HF population. Purpose To collect data on a large French HF population (de novo/chronic/acute; out and in-patients; consultation/hospitalization/rehabilitation; all LVEF classes and any type of cardiologic practice) and to assess low sodium diet prescription and adherence. Methods Prospective enrolment during 6 months in 2017. Data collection included biological, clinical, demographic, dietetary, echocardiographic and treatment characteristics. Results Supervised and unsupervised analysis methods from the data mining and machine learning fields can be performed to exploit data. Baseline variables involved for the analysis (NYHA classes, NTproBNP, Creatinin, BMI, …) are studied among descriptive variables (Age, Sex, Weight, …). Cluster analysis performed are based on Hierarchical Ascending Approaches in Euclidean distance and on the production of Self-Organizing Maps (SOMs) with the statistical software R. Complete data analysis with identification of new patient profiles will be shown at the congress. Conclusion Combining a large representative and non selective French HF population, cluster analysis will allow to identify specific patient profiles constituting homogeneous groups within the sampled population reflecting the severity of their heart condition. New data will be shown during the congress.