Point and Interval Estimation of Average Population Attributable Fraction in a Matched Case Control Design

• Activity Number: 145
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Lin H. Tian*+ and Laura A. Schieve and Owen Devine
• Companies: CDC and CDC and CDC
• Keywords: Average population attributable fraction ; Matched case control design ; Bootstrap method

Abstract

The average population attributable fraction (PAF) is a potentially useful measure for quantifying the contribution of a risk to a particular outcome accounting for the co-occurrence risk factors in addition to adjusting for other confounders. However, estimation of the average PAF can be complex and literature on point and, especially, interval estimation of this value in a matched case control design is scarce. The talk illustrates point and interval estimation of average PAF focused on three perinatal risk factors for autism spectrum disorders. First, the PAFs of the individual and combined effects of the three risk factors were estimated with the stratum-specific odds ratios from the conditional logistic regression models and the corresponding proportions of exposed cases in the strata. Second, the sequential PAFs were calculated based on the possible orders for eliminating risk factors. Third, the average PAFs were calculated by averaging all PAFs for a specific risk factor among all possible sequences. Finally, a bootstrap method was used to calculate the interval estimates. SAS macro codes created by the authors to carry out the calculations will also be presented.