Online Program
|
Funnel Plots as a Vehicle for Policy Relevant Analysis: Gaining Focus Keywords: Health Surveillance, Small Area Analysis, Data Visualization, Over Dispersion Background: Public health surveillance systems regularly inform health policy by providing data and information to decision makers at various levels. Variation is health status across small geographies is often of interest and funnel plots have been recommended for this purpose. Purpose: The interplay between information from surveillance analyses and policy development, public health interventions, and research is examined. Methods: As an example of the inductive, exploratory process of developing information for decision makers, administrative data on health status is combined with risk factor information from surveys in a sequential modeling process. The funnel plot is the vehicle used for visualization and presentation. The role of overdispersion in model formation for policy is critically examined. Results: Examining small area rates of motor vehicle injury and diabetes prevalence, substantial overdispersion was noted. Funnel plots could not be sensibly interpreted on their own. Risk adjusted rates were then created through the use of survey data. The risk adjusted funnel plots lead to the identification of small areas with unexpectedly high rates. This information was passed on to front line public health staff for further investigation. The variation in risk factors was also examined and this information was passed on to policy development staff. The amount of variation in rates due to the risk factors was then provided to decision makers. In the case where overdispersion persists in risk adjusted rates, this was communicated to university researchers. Conclusions: Each step in the analysis of surveillance data can provide insights to surveillance experts, front line public health staff, policy setters, and researchers. The use of funnel plots and the critical examination of overdispersion provide a framework for understanding the differences between statistical model building for explanation and for policy development.
|