Analysis of Binary Data Arising from a Prospective Cluster Randomized Study on the Diagnosis of Chronic Obstructive Pulmonary Disease Using Overdispersed Binomial Models

• Activity Number: 321
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Santosh Sutradhar*+ and Valentina Bayer Zubek
• Companies: Novartis and Boehringer Ingelheim Pharmaceuticals, Inc.
• Keywords: Beta-binomial ; finite-mixture ; zero-inflated-binomial ; Parametric bootstrapping ; goodness-of-fit ; overdispersed binomial model

Abstract

Chronic Obstructive Pulmonary Disease (COPD) is a lung disease that is often under- or misdiagnosed. SEARCHI (Screening, Evaluating and Assessing Rate CHanges of diagnosing respiratory conditions in primary care) was a prospective cluster randomized study that explored the impact of a screening tool and a spirometric device on COPD diagnosis. Cluster binary data on COPD diagnosis were collected. It is reasonable to assume that binary data within a cluster are positively correlated; therefore, the cluster binary counts are expected to exhibit larger variances than those permitted by the binomial model. The focus of this presentation is to analyse the data from this study using available overdispersion models (beta-binomial, finite-mixture, and zero-inflated-binomial). In order to choose the best fitted model for this dataset, we will apply a goodness-of-fit (GOF) test for testing model adequacy of an overdispersed binomial model (Sutradhar et al. J Stat Plan Inference 2008;138:1459-71). A simulation will be conducted in order to understand the distribution of the calculated GOF-statistic. The results of the SEARCHI study will be summarized based on the best fitted model.