Some Measures of Comparing Risk in Clustered Epidemiological Data

• Activity Number: 466
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Risk Analysis
• Author(s): Tasneem Zaihra*+
• Companies: UNB-SJ
• Keywords: Cluster ; Risk Difference ; Risk Ratio ; Relative Risk Difference ; Overdispersion ; Epidemiological Indices

Abstract

In this study, we compare risk between two groups, where each group consists of some clusters and in each group there is a risk. The response of members of each cluster is binary and also members of the same cluster tend to have similar (correlated) responses. We discuss interval estimation of three risk measures used to compare the risk (proportion of response) of binary outcomes between the two groups namely, the risk difference (RD), the risk ratio (RR), the relative risk (RED). We further discuss situations where one risk measure is preferred over other. For instance, RR quantifies the strength of the association between a given disease and suspected risk factor and is widely used in toxicological studies. RD is preferred when we want to determine the relative order of importance of disease in terms public health issues as the level of rates can be lost in using RR and RED is preferred to see effectiveness of some intervention dealing with public health issues where one has to allocate limited resources while making public health policies.