Ic v 13

Summary statistics of independent variables by surgical outcomes and country are first presented, followed by the regression results. Logistic regression analysis was used to examine the factors correlated with the two binary surgical outcomes. Robust SEs were clustered at the hospital level to capture within-hospital correlation, and country-fixed characteristics were also included in the regression. The penalised maximum likelihood logistic regression based on the Firth’s method was selected to reduce the bias caused by the rarity of events.22 23 (link) The ORs of independent variables using two models were shown for comparing association rather than causal interpretation. The statistical analysis was conducted using Stata IC V.13. Significance levels at p<0.05 and 95% CIs were selected.

Descriptive and multivariable analyses were conducted in Stata IC V.13. First, we estimated prevalence with 95% CIs accounting for sampling design by using the svyset function (weights, clusters=schools/classes and stratification by region). Crude ORs and corresponding p values were then computed to estimate the effect of each characteristic on the outcome (self-reported diagnosed diabetes). Multivariable logistic regression analysis was then used to examine the relationship between key variables under study after adjusting for confounding from other predictors.