Innovations in Bayesian joint models with applications to big data and cardiovascular health

Michael Daniels Chair
University of Florida
 
Michael Daniels Organizer
University of Florida
 
Wednesday, Aug 6: 10:30 AM - 12:20 PM
0238 
Invited Paper Session 
Music City Center 
Room: CC-202B 

Applied

Yes

Main Sponsor

Biometrics Section

Co Sponsors

Section on Bayesian Statistical Science
Section on Statistics in Epidemiology

Presentations

Bayesian Nonparametric High-Dimensional Joint Modeling

We consider the setting of joint modeling of a time-to-event outcome and a large number of longitudinally measured processes which are posited to prognosticate the outcome. The literature on joint modeling is diverse; however, current approaches can typically jointly analyze one or a few longitudinal processes. The motivating application for this work comes from a study on Age-related Macular Degeneration (AMD), a disease that affects 12.6% of the US adults aged ≥ 40 years (19.6 million) and more than 190 million people globally. We propose a nonparametric Bayesian joint model for the time-to-event and high-dimensional longitudinal processes that uses flexible low-dimensional structures. We evaluate performance of the proposed approach in simulation studies. 

Keywords

AMD

Interaction 

Speaker

Sanjib Basu, University of Illinois At Chicago

A Joint Model of Longitudinal CVD Risk Factors, Time-varying Medication Use, and Time-to-Terminal Events

We introduce a novel Bayesian approach for jointly modeling longitudinal cardiovascular disease (CVD) risk factor trajectories, medication use, and time-to-events. Through this integrated framework, we connect models of longitudinal CVD risk factors, medication history, and CVD events. A novel component of our joint model is that the model for medication history accommodates uncertainty due to missing medication status as well as the age at which subjects switch off (or on) medication between visits. This history forms a key feature in the time to event model. Our research aims to provide a comprehensive understanding of CVD progression and the role of medications, thus enhancing predictive accuracy and informing personalized intervention strategies. 

Keywords

Bayesian methods

Longitudinal outcomes

Missing data

Multiple Imputation

Medication patterns 

Co-Author(s)

Juned Siddique, Northwestern University
Michael Daniels, University of Florida

Speaker

Zeynab Aghabazaz, Northwestern University

Personalized feature threshold estimation in joint modelling of longitudinal and time-to-event data

Cardiovascular disease (CVD) cohort studies record longitudinal data on numerous CVD risk factors including body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), glucose, and total cholesterol. The commonly used threshold values for identifying subjects at high risk are 30 kg/m^2 for BMI, 120 mmHg for SBP, 80 mmHg for DBP, 126 mg/dL for glucose, and 230 mg/dL for total cholesterol. When studying the association between features of longitudinal risk factors and time to a CVD event, an important research question is whether these CVD risk factor thresholds should vary based on individual characteristics as well as the type of longitudinal feature being estimated. Using data from the Atherosclerosis Risk in Communities (ARIC) Study, we develop methods to estimate risk factor thresholds in joint models with multiple features for each longitudinal risk factor. These thresholds are allowed to vary by sex, race, and baseline smoking status. Our methods have the potential for personalized CVD prevention strategies as well as more precise estimates of CVD risk. 

Keywords

Joint modelling

Threshold estimation 

Speaker

Mirajul Islam, University of Florida

Incorporating Cross-Sectional Information into a Joint Model of Longitudinal and Survival Data using Informative Priors

We discuss Bayesian approaches for incorporating cross-sectional cardiovascular risk factor data and its association with events into a joint model of longitudinal data and survival outcomes through the use of informative priors. We use longitudinal data from the Coronary Artery Risk Development in Young Adults (CARDIA) cohort study and cross-sectional data from the Third National Health and Nutrition Examination Survey (NHANES) Linked Mortality File.  

Keywords

power prior

commensurate prior 

Co-Author(s)

Zeynab Aghabazaz, Northwestern University
Michael Daniels, University of Florida

Speaker

Juned Siddique, Northwestern University