Table 2.
Results of the univariate and multivariate GLM of height of COVID-19 deaths' peak, with normal distribution and logit link function.
| Variable | Univariate analysis |
Multivariate analysis |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Base case GLM (full model) |
Base case GLM + selection algorithms (final model) |
|||||||||||
| Estimate | Wald 95% confidence limit |
P-value | Estimate | Wald 95% confidence limit |
P-value | Estimate | Wald 95% confidence limit |
P-value | ||||
| Lower | Upper | Lower | Upper | Lower | Upper | |||||||
| All-bed capacity (per 1 mln inhabitants) | −0.0002 | −0.001 | 0.0001 | 0.109 | – | – | – | – | – | – | – | – |
| ICU bed capacity (per 1 mln inhabitants) | −0.002 | −0.008 | 0.003 | 0.404 | – | – | – | – | – | – | – | – |
| Total no. of tests up to the time of the peak (per 1 mln inhabitants) | 0.000 | −0.001 | 0.001 | 0.961 | – | – | – | – | – | – | – | – |
| Stay-at-home order day | 0.027 | −0.001 | 0.053 | 0.057∗∗ | – | – | – | – | – | – | – | – |
| Day of closure of educational facilities | 0.067 | 0.028 | 0.107 | 0.001∗ | – | – | – | – | – | – | – | – |
| Day of imposing restrictions on gatherings | 0.045 | 0.006 | 0.084 | 0.023∗ | – | – | – | – | – | – | – | – |
| Business closure day | 0.029 | 0.009 | 0.049 | 0.005∗ | −0.001 | −0.022 | 0.020 | 0.944 | – | – | – | – |
| Population size (mln) | 0.001 | −0.009 | 0.011 | 0.877 | – | – | – | – | – | – | – | – |
| Proportion living in urban areas | 8.117 | 3.695 | 12.539 | <0.001∗ | 7.127 | 3.642 | 10.611 | <0.001∗ | 6.848 | 4.016 | 9.680 | <0.001∗ |
| Proportion living in metropolitan cities with more than 1 mln inhabitants | 2.428 | −0.464 | 5.319 | 0.099∗∗ | 1.063 | −1.331 | 3.457 | 0.384 | – | – | – | – |
| Median age | −0.0003 | −0.120 | 0.120 | 0.996 | – | – | – | – | – | – | – | – |
| Arrivals at airports in 2018 (per 1 inhabitant) | 0.016 | −0.097 | 0.129 | 0.784 | – | – | – | – | – | – | – | – |
| No. of foreign tourists in 2018 (per 1 inhabitant) | −0.076 | −0.421 | 0.270 | 0.668 | – | – | – | – | – | – | – | – |
| Mobility score at the day of the first reported death | 0.048 | 0.001 | 0.096 | 0.046∗ | 0.041 | −0.0004 | 0.083 | 0.052∗ | 0.049 | 0.022 | 0.077 | <0.001∗ |
| Border closure day | 0.025 | 0.010 | 0.041 | 0.002∗ | −0.005 | −0.030 | 0.021 | 0.729 | – | – | – | – |
| No. of COVID-19 infections when borders were closed (per 1 mln inhabitants) | 0.0002 | 0.0000 | 0.004 | 0.022∗ | 0.0003 | −0.0001 | 0.001 | 0.113 | 0.0002 | 0.0001 | 0.0003 | 0.016∗ |
| No. of COVID-19 deaths when borders were closed (per 1 mln inhabitants) |
0.004 |
0.001 |
0.007 |
0.002∗ |
– |
– |
– |
– |
– |
– |
– |
– |
| Multivariate model statistics | ||||||||||||
| Scale | – | – | – | – | 5.020 | 3.958 | 6.367 | – | 5.092 | 4.015 | 6.458 | – |
| AIC | – | – | – | – | 222.203 | – | – | – | 217.169 | – | – | – |
| AICC | – | – | – | – | 227.963 | – | – | – | 219.312 | – | – | – |
∗P-value <0.05; ∗∗P-value <0.1.
AIC, Akaike Information Criterion; AICC, AIC corrected for small sample sizes; GLM, generalised linear model; ICU, intensive care unit; mln, million; COVID-19, coronavirus disease 2019.
Univariate and multivariate GLMs with normal distribution and logit link function were used to explore factors associated with height of COVID-19 deaths' peak as of June 3, 2020. Each model was run using 34 observations. Variables significant in univariate models were included in the multivariate base case model, avoiding highly correlated pairs. The final multivariate model was selected based on the use of selection algorithms (backward, forward, stepwise and genetic algorithm).