A fast approach for analyzing spatio-temporal patterns in ischemic heart disease mortality across US counties (1999-2021)

TL;DR

This paper introduces a scalable Bayesian approach using divide and conquer strategies to efficiently analyze spatio-temporal patterns of ischemic heart disease mortality across US counties from 1999 to 2021, revealing regional and urban-rural disparities.

Contribution

It presents a novel scalable Bayesian modeling framework with imputation for high-dimensional spatio-temporal disease data, enabling fast analysis of IHD mortality trends at multiple geographic levels.

Findings

Slowdown in IHD mortality decline after 2014

Slight increase in IHD mortality after 2019

Regional and urban-rural disparities in trends

Abstract

Ischaemic heart disease (IHD) remains the primary cause of mortality in the US. This study focuses on using spatio-temporal disease mapping models to explore the temporal trends of IHD at the county level from 1999 to 2021. To manage the computational burden arising from the high-dimensional data, we employ scalable Bayesian models using a "divide and conquer" strategy. This approach allows for fast model fitting and serves as an efficient procedure for screening spatio-temporal patterns. Additionally, we analyze trends in four regional subdivisions, West, Midwest, South and Northeast, and in urban and rural areas. The dataset on IHD contains missing data, and we propose a procedure to impute the omitted information. The results show a slowdown in the decrease of IHD mortality in the US after 2014 with a slight increase noted after 2019. However, differences exists among the counties, the four big geographical regions, and rural and urban areas.