B0456
Title: A Bayesian semi-parametric approach for incremental intervention effects in mortal cohorts
Authors: Maria Josefsson - Umea School of Business, Economics and Statistics (Sweden) [presenting]
Abstract: Bayesian semi-parametric inference is proposed on a hypothetical incremental intervention that depends on the natural value of treatment from longitudinal data with attrition. In particular, two types of mortal cohort inferences are proposed using the extended G-formula. It is argued that a combination of the two approaches is to be preferred, especially if the population is older and death is thought to be a post-randomization event, such that conditioning on survival introduces bias. Using data from a longitudinal prospective cohort study on ageing, cognition and dementia, the approach is applied to estimate the cognitive effects of a hypothetical intervention where systolic blood pressure was monitored at more optimal levels over 15 years.
