Big Data for Renal Failure Prediction

Creation of a predictive model of cardiovascular risk and mortality in patients with advanced renal failure in hemodialysis using Machine Learning: Utility of BigData

Research Partners: 2 (Universitat Autònoma de Barcelona, Parc Taulí Hospital Universitari)

Funding Agency / Institution: Institut d’Investigació I Innovació Parc Taulí (I3PT)

Period: 1 year (1/01/2018-31/12/2019)

General Project coordinator: Jose Ibeas Lopez (Parc Taulí Hospital Universitari)

UAB Local project coordinator: Javier Serrano

Research team members: Antoni Morell Perez, Javier Serrano, Dolores Rexachs, Edwar Macias, Jose Lopez Vicario.

Total funding: 5000 €


 

Project objectives

Cardiovascular disease (CV) is the main cause of morbidity and mortality in the general population. The approach to identify the future cardiovascular risk recommended by the Guidelines is based on known risk factors that are included in most predictive models. However, an important part of the population escapes their identification or receives unnecessary preventive treatment. In parallel, chronic kidney disease, especially advanced hemodialysis, which is an epidemiological problem, is closely related to cardiovascular risk and both mutually reinforce each other.

CV risk models implicitly assume that each risk factor is linearly related to events, oversimplifying what are really complex relationships that would include a huge amount of risk factors, which would also have non-linear relationships. Therefore, approaches that incorporate multiple risk factors and that identify real relationships are needed.

Machine-learning offers an alternative to standardized predictive models that have limitations. It is based on computational methods that detect complex and non-linear interactions between variables and identify latent variables that are unlikely to be directly observed. There is no experience in the study of CV risk in hemodialysis.

 The objective of this project is to evaluate if machine-learning algorithms based on neural networks can accurately establish a predictive model of cardiovascular risk and mortality in the hemodialysis patient through an ambispective study with a retrospective cohort of 537 deceased patients to establish the model and another prospective of 140 in a hemodialysis program to validate it.