Distribution of the environmental and socioeconomic risk factors on COVID-19 death rate across continental USA: a spatial nonlinear analysis

Abstract

The COVID-19 outbreak has become a global pandemic. The spatial variation in the environmental, health, socioeconomic, and demographic risk factors of COVID-19 death rate is not well understood. Global models and local linear models were used to estimate the impact of risk factors of the COVID-19, but these do not account for the nonlinear relationships between the risk factors and the COVID-19 death rate at various geographical locations. We proposed a local nonlinear nonparametric regression model named geographically weighted random forest (GW-RF) to estimate the nonlinear relationship between COVID-19 death rate and 47 risk factors derived from the US Environmental Protection Agency, National Center for Environmental Information, Centers for Disease Control and the US census. The COVID-19 data were employed to a global regression model random forest (RF) and a local model GW-RF. The adjusted R 2 of the RF is 0.69. The adjusted R 2 of the proposed GW-RF is 0.78. The result of GW-RF showed that the risk factors (i.e. going to work by walking, airborne benzene concentration, householder with a mortgage, unemployment, airborne PM 2.5 concentration and per cent of the black or African American) have a high correlation with the spatial distribution of the COVID-19 death rate, and these key factors driven from the GW-RF were mapped, which could provide useful implications for controlling the spread of the COVID-19 pandemic.

Geographically weighted machine learning model for untangling spatial heterogeneity of type 2 diabetes mellitus (T2D) prevalence in the USA

“…Although the GW-RF model in this study used only six well-known risk factors for exploring spatial heterogeneity of T2D prevalence, the focus of this study is not understanding the causation of T2D prevalence across US counties. Instead, this study is intended as a demonstration of how the recently developed GW-RF model 23 , 24 , 76 , 77 can be used as both a predictive and exploratory tool to explore spatial heterogeneity of T2D considering the non-linear relationship between risk factors and T2D prevalence. Thus, this method is applicable in many instances where there is an issue about selecting significantly correlated variables at various geographical locations.…”

“…The linear model is susceptible to outliers, and strong assumptions are required regarding the relationships between predictors and target variables (linearity) and the relationships among the predictors (collinearity). The nonlinear non-parametric models such as random forest (RF) do not need to consider multicollinearity and can analyze all independent variables without screening 24 . The geographically-weighted random forest (GW-RF) model may address the limitations of the linear GW-OLS model and can improve predictive performance relative to a non-geographically-weighted random forest model, which is unable to resolve heterogeneous spatial processes 23 .…”

“…The geographically-weighted random forest (GW-RF) model may address the limitations of the linear GW-OLS model and can improve predictive performance relative to a non-geographically-weighted random forest model, which is unable to resolve heterogeneous spatial processes 23 . The main idea of GW-RF is similar to that of the traditional GW-OLS model, in which the model is calibrated locally rather than globally 17 by integrating spatial weight matrix (SWM) and RF into a local regression analysis framework 24 . The local feature importance represents the mean increase in Mean Squared Error (incMSE) if a predictor would be randomly permuted or the decrease in node impurities (IncNodePurity) from splitting on the variable, averaged over all trees.…”

“…The main idea of GW-RF is similar to that of the GW-OLS model 17 , in which the model is calibrated locally rather than globally. The GW-RF does not need to consider multicollinearity and can analyze all independent variables without screening 24 and may provide superior predictive power and evaluation of associations between independent and dependent spatial variables compared to the GW-OLS 23 . However, applying a non-parametric GW-RF model to explain spatial heterogeneity between disease outcome and risk factors is still lacking, and further research comparing GW-OLS and GW-RF is therefore warranted.…”

Exploring spatiotemporal effects of the driving factors on COVID-19 incidences in the contiguous United States

“…The spatial variability and clustered concentration of both COVID-19 mortality and morbidity in many countries have demonstrated a strong spatial dependency of the confounding factors ( Desmet & Wacziarg, 2020 ; Ren et al, 2020 ; Zhang & Schwartz, 2020 ). Although several timely efforts (e.g., Luo, Yan, & McClure, 2020 ) have analyzed spatial heterogeneous patterns and uneven distributions of COVID-19 casualties, few studies have utilized the spatial time-varying dimension in spatial econometric modeling for analyzing geographic disparities in COVID-19 casualties in the United States ( Sun, Matthews, Yang, & Hu, 2020 ). The present research, therefore, has made an effort to examine how spatial analysis can help with identifying the hotspots and vulnerable locations as well as exploring the spatial dependency of confounding factors that explain the overall casualties caused by COVID-19.…”