Title: Optimizing spatiotemporal modeling for public health risk analysis

Project Team: Helen Brown, Christopher Murphy, Chris Williams, Erich Seamon, Mohamed Megheib

Start Date: September 2021

This is a continuation of small area estimate (SAE) modeling of Idaho Behavioral Risk Factor Surveillance System (BRFSS) health indicators. In 2019 the Pl sought MAG support to model obesity indicators. A model was created but the person modeling the work left the UI unexpectedly and did not leave behind reproducible code. Brown, Pl, sought funding support from IDHW/BRFSS to engage IMCI modelers Seamon and Megheib to develop a methodology to create county level measures for three BRFSS health indicators. This work was completed and a publication is in process.

This project will continue the work initiated by Seamon and Megheib to optimize the modeling techniques used to arrive at county level health estimates. This is critically important as the BRFSS data is currently only available at a Public Health District (PHO) level. Each PHO comprises 5-7 counties and the lack of data granularity does not allow for clear assessment of critical public health concerns, nor targeted and strategic public health interventions and evaluation. This work also has the potential to inform BRFSS sampling strategies, increase the capacity of the state to generate county level estimates, and initiate new techniques for combining BRFSS with other existing data sets.

The prime areas of model optimization include: 1) expanding upon existing BRFSS model development, by extending the approach to a larger range of health indicators; 2) automating the transformation and constraint selection via more sophisticated algorithmic evaluation; and 3) modularizing the model pipeline to allow for dynamic running and model parameter selection thru the development of an R-specific dashboard (only accessible to researchers and other policy staff – not public)