Geostatistical software for spatial and multi-dimensional joinpoint regression analysis of time series of health outcomes

Goovaerts, Pierre

Abstract

Analyzing temporal trends in cancer incidence and mortality rates can provide a more comprehensive picture of the burden of the disease and generate new insights about the impact of various interventions. Join point regression developed by NCI Surveillance Research Program is increasingly used to identify the timing and extent of changes in time series of health outcomes and to project future cancer burden through the prediction of the future number of new cancer cases or deaths. The analysis of temporal trends outside a spatial framework is however unsatisfactory, since it has long been recognized that there is significant variation among U.S. counties and states with regard to the incidence of cancer. It is thus critical to implement join point regression within Geographical Information Systems (GIS), and develop interfaces offering user-friendly tools for pre-processing, modeling, visualizing and summarizing large ensembles of time series of health outcomes. This SBIR project is developing the first commercial software to offer tools for the geostatistical modeling and join point regression analysis of time series of health outcomes. The research product will be a stand-alone module into the desktop space-time visualization core developed by BioMedware, an Esri partner. This software package will provide a comprehensive suite for: 1) the computation and geostatistical noise-filtering (kriging) of time series of health outcomes at various spatial scales (e.g. ZIP codes, counties), 2) the visualization of how the parameters of the regression model (e.g. join point years, Average Annual Percent Change) change in space and across spatial scales, and 3) the analysis of similarities among time series and their aggregation through multi-dimensional scaling and clustering analysis. These tools will be suited for the analysis of data outside health sciences, such as in crime mapping, fish stock assessment or climate change, broadening significantly the commercial market for the end product. This project will accomplish three aims: ? Conduct simulation-based studies to assess the benefits of: 1) the application of join point regression to smoothed time series (kriging-based and Bayesian filters) for identifying temporal trends from unstable rates recorded in small geographical units, 2) multi-dimensional scaling to visualize differences among ensemble of time series, and 3) clustering analysis to group geographical units with similar temporal trend. Develop and test a prototype module that will guide users through the creation, join point regression modeling, visualization and multi-dimensional analysis of time series of health outcomes, based on BioMedware's space-time visualization and analysis technology. ? Conduct a usability study and identify additional methods and tools to consider in Phase II. These technologic, scientific and commercial innovations will revolutionize our ability to detect changes in cancer incidence and mortality across space and through time, bringing important information and knowledge that will benefit substantially cancer epidemiology, control and surveillance and help reducing these disparities.

Public Health Relevance

The economic burden of cancer in the US is substantial and expected to increase significantly in the future due to expected growth and aging of the population and improvements in survival as well as trends in treatment patterns and costs of care following cancer diagnosis. Many public health officials and policy-makers require quantitative prediction about the future cancer burden in order to assist in the planning and prioritization of prevention activities, allocation of health services, and evaluation of cancer control interventions or treatments. The analysis of temporal trends outside a spatial framework is however unsatisfactory, since it has long been recognized that there is significant variation among U.S. counties and states with regard to the incidence of cancer. Thus, many public health issues would greatly benefit from improved tools for detecting and visualizing changes in cancer incidence and mortality across space and through time.