Skip to main content

Table 1 Analyses of the potential predictors of spatial heterogeneity in hospitalisation incidence

From: Investigating the drivers of the spatio-temporal heterogeneity in COVID-19 hospital incidence—Belgium as a study case

  HI (01/03–31/05/2020) HI (01/09–30/11/2020) HI (01/03–30/11/2020)
  MLR R2 = 0.18*, BRT cor. = 0.69 MLR R2 = 0.24*, BRT cor. = 0.76 MLR R2 = 0.23*, BRT cor. = 0.75
Spatial covariate ULR R2 MLR β BRT RI ULR R2 MLR β BRT RI ULR R2 MLR β BRT RI
Population density 0.00 0.8% 0.00 1.1% 0.00 0.9%
Median age 0.00 2.6% 0.02 6.2% 0.00 2.4%
Prop. > 65 years old 0.00 6.7% 0.01 1.4% 0.00 1.3%
Ratio NH beds/population 0.15* 0.22* 48.1% 0.22* 0.41* 55.3% 0.21* 0.31* 56.8%
Median income 0.00 3.2% 0.05* -0.07 5.5% 0.00 3.9%
% in primary sector 0.04 5.7% 0.04* 0.00 14.3% 0.05* 0.02 14.4%
% in secondary sector 0.02 2.2% 0.02 1.8% 0.00 2.4%
% in tertiary sector 0.05* 0.06 3.0% 0.01 2.5% 0.00 1.5%
PM10 concentration 0.01 1.6% 0.00 3.1% 0.00 1.7%
PM2.5 concentration 0.01 0.9% 0.00 1.8% 0.00 0.9%
Prop. urban areas 0.05* 0.05 25.4% 0.05* 0.09 7.1% 0.06* 0.12 13.7%
  1. This table summarises the results of univariate linear regression (ULR), multivariate linear regression (MLR), and boosted regression trees (BRT) analyses performed to investigate the association between measures of hospitalisation incidence (HI) and various spatial covariates associated with hospital catchment areas (HCAs). We report the following metrics: the coefficient of determination (R2) for the ULR analyses, the regression coefficient (β) for the MLR analyses, and the relative influence (RI) associated with each spatial covariate for the BRT analyses. We also report the overall R2 and Spearman correlation (“cor.”) for each distinct MLR and BRT analysis, respectively. (*) indicates if a given R2 or β is significant (p-value < 0.05). See also Additional file 4: Table S1 for the results of the same analyses performed when only considering HI values based on nursing home (NH) residents, and Additional file 4: Table S2 for the results of the same analyses performed after having discarded HCAs of Brussels-Capital region as well as a potential outlier HCA (highlighted in Fig. 4)