Specific methods

The prevalence of each cause and the distribution by severity were primarily estimated with the Bayesian meta-regression method DisMod-MR 2.1 [7,11]. Regression models were used to adjust data that did not follow the standard definition for each cause in the direction of case definition–based data [12]. The all-cause and cause-specific death rates were estimated by the Cause of Death Ensemble model [11]. YLDs, YLLs, and DALYs were estimated with the methods in the published literature [7,10,11,13–17]. Briefly, YLDs for a specific cause was calculated by multiplying its prevalence with the corresponding disability weight, which had been estimated in several previous worldwide surveys and adjusted for comorbidity. YLLs was calculated by summing up the remaining life expectancy of people dying in each age group. DALYs equaled to the sum of YLLs and YLDs by any corresponding sub-population of a specific cause [7,10,11,13–17]. Age-standardized mortality rates, YLLs rates, YLDs rates, and DALYs rates were computed based on the world standard population reference developed for GBD study, which was a time-in-variant standard [11,18].