Data extraction Information from each included study was extracted into a database with predesignated fields, including metadata (e.g. title, authors), study features (e.g. study period, data sources, population), statistical methods (e.g. model type, analytical unit, covariates) and results (e.g. coefficient point estimates, standard errors); see Supporting information, Table S2. One reviewer extracted data from each study and entered them into the standardized form. A second reviewer checked all fields for accuracy and completeness; discrepancies were resolved by consensus. Quality assessment We assessed methodological quality by whether authors adjusted for potential confounders, accounted for potential collinearity in policy adoption and used models wellsuited for causal inference (i.e. inclusion of pre-post policy data, control/comparison groups and number of treated units). Methodological concerns for studies are described narratively in the Results. Effect size estimates Studies reported effects in various ways. To facilitate comparisons of the magnitude of effect sizes across studies, we calculated and presented incidence rate ratios (IRRs). Studies reporting the results from a negative binomial or Poisson regression model are directly reported as IRRs with their associated confidence intervals (CIs). For low probability outcomes, odds ratios (ORs) were interpreted and reported as IRRs with their associated CIs. For studies that used linear models with a logtransf