RDU Webinar Series: Estimating Neighbourhood-level Behavioural Risk Factor Prevalence from Large Population-Based Surveys: a Bayesian Approach

Recorded On: 11/02/2016

Complex survey-based behavioural risk factor estimates are available at the regional level, but public health insights are limited due to unobserved heterogeneity that may exist in these regions. Advanced spatial analysis techniques can produce sensible small-area estimates of risk factors and identify areas of high prevalence. This presentation describes a spatial Bayesian hierarchical model which estimates small-area prevalence of current smoking and excess bodyweight (body mass index ≥ 25 kg/m2) by pooling 5 cycles of the Canadian Community Health Survey for a region in southwestern Ontario. This study demonstrates the feasibility of a full Bayesian model for complex survey data to identify areas with elevated risk factor prevalence. These spatial analysis techniques provide small-area estimates to inform surveillance activities and may be used to explain cancer incidence at small-area levels of geography.

Laura Seliske, PhD

Research Associate-Surveillance Research, Analytics & Informatics

Dr. Seliske has been a Research Associate with Cancer Care Ontario since 2012, and is involved in the small-area analysis of cancer and its determinants. Prior to joining Cancer Care Ontario, Dr. Seliske completed her training in epidemiology at Queen's University, in Kingston Ontario.