Random forest using smartphone GPS in first wave of COVID-19 in the Maule region, Chile

Abstract

Background: The COVID-19 pandemic has had a global impact. Knowing the variables that affect the increase in infection is crucial for public health decision-making. Mobility and socio-demographic conditions of the population are important factors in the transmission of the SARS-CoV-2. The objective of this study is to analyze the relationship between people mobility, social determinants of health and COVID-19 cases using a Random Forest (RF) method. Methods: The COVID-19 cases were analyzed in the Maule Region, Chile. Spearman rank was performed to analyze the total mobility index for each municipality. RF regression was used to create a model between COVID-19 infections, mobility index and sociodemographic variables. P-value <0.05 was considered statistically significant. Results: Total mobility was highly correlated with new COVID-19 cases, adjusted for total population, in each municipality (ρ: 0.52-0.92). An upward trend is observed for mobility and COVID-19 cases for the 30 municipalities analyzed. For the RF model, COVID-19 active cases, total mobility, and external mobility are obtained as VIM. The most relevant demographic variables were overcrowding, density and area of municipality. The R-Squared was 0.68 for the performed RF model. Conclusions: Artificial Intelligence methodologies are increasingly used for their excellent performance. RF Regression offers a clear solution for the design of predictor variables on the number of new cases per week. Mobility is a powerful predictor variable for the number of COVID-19 new cases.