Outcome and statistical analyses

The outcome of our study was the response to the first randomised question, which was expressed as a percentage from 1% to 99% (converted to a 0.01–0.99 scale). To explore the effect of each factor on the probability of choosing surgery compared with the situation in which woman receive no additional input, we modelled unadjusted and adjusted beta regression models. We used the probability of choice as the dependent variable, and the first randomised factor as the main independent variable.

Adjustment was performed using all the available sociodemographic data, and missing data were handled through chained equation multiple imputation, generating 10 imputed data sets (see table 1 and online supplemental table S2 for further information). As a subanalysis of the study, we also ran multilevel beta regression models for repeated measures to investigate the effect of each factor on the probability of choice within each single respondent, using the woman’s anonymous identifier as the higher-level grouping variable. Statistical analyses were carried out on Stata Software V.17.0 (StataCorp). Statistical significance was set at a p<0.05.

Missing data were handled by using chained equation multiple imputation, generating 10 imputed data sets. Multiple imputation is an iterative form of stochastic imputation that – —instead of filling in a single value – —uses the distribution of the observed data to estimate multiple values that reflect the uncertainty around the true (missing) value. We included in the multiple imputation model the dependent and independent variables of the regression models, and all women’s sociodemographic data.