Nicole M White1
1Australian Centre for Health Services Innovation, Queensland University of Technology, 60 Musk Av Kelvin Grove, Qld, 4059, firstname.lastname@example.org
Infection control studies collect data from a variety of sources and study designs, from retrospective case-control or ‘quasi-experimental’ studies that use existing data sources to cluster randomised trials where data are collected across one or more locations over time. For a chosen outcome (eg. healthcare-associated infection rates), data collected pose a number of challenges for statistical analysis, requiring one to consider how best to account for features of the data that are attributable to the study design and potential confounders. Confusingly, studies reporting on the same outcome will often vary in terms of the study design and/or statistical methods used. The interpretation of study findings reported in the literature therefore requires considerable care, as the assumptions made as part of data analysis may fail to account for this complexity and, in some cases, impact study conclusions.
This talk outlines common statistical methods used in infection control studies and how failure to account for underlying features or ‘pitfalls’ in the data collected can affect study conclusions. We focus on current statistical approaches used for the analysis of intervention studies and estimating excess length of stay attributable to healthcare-associated infections. The emerging role of ‘big data’ analytics in infection control and potential analysis pitfalls will also be discussed.