Structural Nested Mean Modeling of Response-Maximized Breastfeeding Strategies

• Activity Number: 581
• Type: Invited
• Date/Time: Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Erica E.M. Moodie and Susan Shortreed*+
• Companies: McGill University
• Keywords: Causal inference ; Confounding ; Longitudinal data ; G-estimation ; Structural nested mean models ; Optimal dynamic treatment regimes

Abstract

To estimate the sequence of actions that optimizes response in a longitudinal setting, it is important to study sets of decision rules rather than series of single-action comparisons. We take as our motivating example the duration of breastfeeding with a view to maximizing infant growth, using the 17,046 women recruited in the Promotion of Breastfeeding Intervention Trial (PROBIT). Breastfeeding has many recognized beneficial effects on development, however observational evidence has suggested that breastfeeding is associated with reduced infant growth. Structural nested mean models are proposed and their parameters found by G-estimation in order to obtain unbiased estimates of the effect of continued breastfeeding on infant growth at one year of age. The decision rules given by the G-estimates are compared with the decisions implied by a myopic optimization approach.