Trajectories and Joint Modelling Applied to Health Metrics

Mixed linear models provide an insightful understanding of trajectories through the consideration of fixed and random effects.

The longitudinal sequence may terminate due to an event or censoring. It is of interest to learn about longitudinal progression as well as risk factors and the implications of these on time-to-event. It is known that joint models, accounting for a joint distribution, for longitudinal and time-to-event data, are necessary to analyse these data.

The Latent variable models allows to obtain more accurate health indicators in the presence of measurement error, identification of classes/cluster within the data, identification of the true prevalence and sensitivity and specificity without the presence of gold-standard in diagnostic tests, and also calibration of several indicators with the presence of missing data. Bayesian and frequentist approaches can be used to estimate these types of models.

Such statistical models are being developed and applied in nutrition data, cancer, peritoneal dialysis and kidney transplantation survival and cardiovascular diseases research.

The conducted research aims for:

• The development and implementation of statistical methods in the context of relative survival analysis and joint modelling of longitudinal and time to event data in epidemiological population-based and clinical data at patient level.
• The use of latent (trait or class) variable models to obtain more accurate health indicators and identification of data patterns.
• The production of quality research in biostatistics and health metrics in the areas of Epidemiology and Public Health, Medicine and Biology, with the application, development and dissemination of mathematical and statistical techniques