Métodos de análisis de factores y predictores de la evolución en la supervivencia de la enfermedad cardiovascular

Resumen

Objectives: The main objective of this work is the development and implementation of a statistical model to study the evolution of patients with HF, by using left ventricular ejection fraction (LVEF), and to determine factors associated with it. Methods: This is a longitudinal, retrospective, observational and descriptive study. In it, we analyzed the clinical data of 251 patients attending the HF unit of the Hospital Virgen de la Arrixaca of Murcia during the years 2005 to 2017 with reduced LVEF (less than or equal to 40%). LVEF data were obtained from 1504 echocardiographies (average of 6.0±3.7 per patient) collected in the hospital. For all of them, mortality (occurring in 45.4% patients) and follow-up times (average of 8.9 years) were reviewed. To study these data we propose a model based on joint models for longitudinal and time-to-event data. To describe the evolution of LVEF values, we use a nonlinear function determined by an initial value, a growth rate of the values and a maximum point called saturation where they approach. To estimate and define this model we use bayesian statistics and the probabilistic language Stan. This language provides more freedom to establish different characteristics of the model. Results: The results of our analysis indicate that patients start with initial LVEF values around 25% (with 95% credible interval of [22.6, 26.8]). The increase in these values is produced in the first 6 months and it is shorter in the presence of ischemic heart disease. The average increase is estimated at 15.5\% (with 95% confidence interval of [12.7, 18.3]). The variables associated with LVEF initial values or with posterior increment are: acute myocardial infarction, hypertension, ischemic etiology, time since diagnosis or end-diastolic volume (a cardiac remodeling measure). Our results show that LVEF improvement reduces the risk of death by 18% (hazard ratio of 0.82 with a 95% credible interval of [0.71, 0.94]).