Univariant (n=137) | Multivariable* (n=133) | Multivariable (n=117) | |||||||

Coefficient | 95% CI | P value | Coefficient | 95% CI | P value | Coefficient | 95% CI | P value | |

Consumption versus number of opioids in EMLs | 0.17 | 0.09 to 0.6 | 0 | 0.05 | −0.02 to 0.11 | 0.19 | 0.01 | −0.009 to 0.03 | 0.27 |

GDP/100 per capita | 0.006 | 0.003 to 0.009 | 0 | 0.00005 | −0.0012 to 0.0013 | 0.93 | |||

Healthcare expenditure per capita | 0.0001 | 0.0007 to 0.002 | 0 | 0.0004 | 0.0002 to 0.0005 | 0 | |||

Population | −1.35e-10 | −5.05e-10 to 2.35e-10 | 0.47 | ||||||

Life expectancy | −0.009 | −0.03 to 0.01 | 0.38 | ||||||

Human development index | 1.33 | 0.012 to 2.64 | 0.48 | ||||||

Corruption perception score | 0.007 | 0.0007 to 0.01 | 0.03 | ||||||

Region (Africa) | |||||||||

America Asia Europe Oceania | −0.065 0.12 0.32 0.16 | −0.3 to 0.17 −0.09 to 0.33 0.04 to 0.59 −0.3 to 0.6 | 0.59 0.27 0.03 0.48 |

The assumptions for untransformed linear regression were not met. Thus, we used a square root transformation of the dependent variable (ie, opioid consumption in mg/person), which improved the model.

*we conducted this multivariable analysis first as it had the least amount of missing data and the variables had the strongest predictors of opioid consumption.

GDP, gross domestic product.