Incidence difference estimates: pre- vs post-implementation periods. The 95% credible intervals represent the 2.5th and 97.5th percentiles of 1 000 000 bootstrapped incidence difference estimates. Negative incidence estimates are set to zero in the bootstrapping procedure, because negative incidence is not possible, and CIs are therefore not necessarily symmetrical around the point estimate. Given the multiple comparisons taking place, a critical value of P = .05 may not be appropriate. See supplemental Figure 2 for incidence difference credible intervals based on the Bonferroni correction.