Meta-regression
We conducted univariate and multivariable random-effects meta-regression
analyses to assess the association between log-transformed
seroprevalence and pre-decided factors. All variables in the univariate
model whose p-value was less than 0.1 were included in the multivariable
analyses. A P -value < 0.05 (two sides) in the
multivariable analysis indicated a statistically meaningful association.