All analyses were carried out with the R software (39), using the packages metaphor (40), meta (41), & weight (42). Meta-analysis of proportions, using random-effects models were carried out to estimate the pooled prevalence of each comorbidity. Only studies that explicitly reported rates of a particular comorbidity were included in the meta-analysis of that particular comorbidity. As most of the comorbidities had a prevalence rate of <20%, we used a double-arcsine transformation to ensure normality of the variance estimates. The pooled prevalence was then calculated using the DerSimionian-Laird inverse variance approach (43). This approach ensures that a study with a larger sample size is given more weight compared to a study with a smaller sample. The pooled prevalence estimates, along with the 95% confidence and prediction intervals are reported after back-transformation to percentages. The heterogeneity of the pooled estimates are reported using the I2 statistic and its p-value.
