Comparison of Methods for Adjustment of Confounding by Indication in Assessing Respiratory Syncytial Virus (RSV) Immunoprophylaxis Impact on Childhood Asthma

• Activity Number: 554
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Tebeb Gebretsadik* and Kecia Carroll and Gabriel Escobar and Pingsheng Wu and Sherian Xu Li and Eileen M. Walsh and Edward F. Mitchel and Chantel Sloan and William D. Dupont and Tina V. Hartert
• Companies: Vanderbilt University School of Medicine and Vanderbilt University and Kaiser Permanente and Vanderbilt University and Kaiser Permanente Division of Research and Kaiser Permanente Division of Research and and Brigham Young University and Vanderbilt University and Vanderbilt University
• Keywords: Confounding by indication adjustment ; propensity score (PS) ; PS matching with replacement ; electronic health care databases ; RSV prophylaxis ; asthma

Abstract

We compared adjustment methods to overcome confounding by indication resulting in differential receipt of RSV prophylaxis to eligible infants, including: (1) covariate-adjustment regression (2) propensity (PS) adjusted regression and (3) PS matching with replacement with a defined caliper width to assess the effect of adherence to RSV prophylaxis on asthma in a retrospective birth cohort of 6751 children. The first two methods yielded similar results with higher adherence to RSV prophylaxis being associated with increased asthma risk. However in PS matched analysis among treated infants, infants with >=70% adherence were less likely to develop asthma than infants with < 20% adherence. Matching may provide the closest resemblance to a randomized clinical trial by reducing imbalances in confounding variables between the two treatment groups. We evaluate how PS score matching with replacement and optimal caliper width estimation methods affect group balance, bias, variance and generalizability in assessing the effect of RSV prophylaxis on asthma risk. Our findings provide guidance on adjusting for confounding by indication from observational data and for future methods development.