Development of a Predictive Algorithm for the Risk of Rehospitalization of Patients With Heart Failure

Heart failure is a chronic disease whose prevalence, due to the aging of the population, is increasing. In France, the prevalence of this pathology is 2.3% (it reaches 10% in the over 75 years) and affects nearly a million patients.

The rehospitalization of patients with heart failure affects 25% of patients within 1-3 months of hospital discharge, and 66% at 1 year while 75% of hospitalizations are preventable. These readmissions result in decreased quality of life and increased mortality; from an economic point of view, hospitalization accounts for 70% of expenses related to the management of heart failure. Avoiding rehospitalization is therefore a major public health issue. The current predictive scores remain perfectible, even though risk factors for readmission have already been the subject of numerous studies. The identification of patients at risk of rehospitalization is still an issue, especially for patients with preserved left ventricular ejection fraction. Targeting patients requiring appropriate care remains an issue.

The rise of innovative statistical techniques around Big Data in health opens new perspectives for the scientific exploitation of data available in electronic medical records, for example in the field of prediction. This study aims to explore the risk of rehospitalization in heart failure patients by analyzing routine data collected in medical records and by mobilizing artificial intelligence algorithms. A review of the literature confirms the innovative nature of such an approach: the majority of the studies identified implemented a prospective collection of data; only 20% of the studies mobilized the medical file; no French study used the new machine learning algorithms.