Table 5

Triple-difference regression results

Difference-in-differences modelsLagged dependent variable models
All years (1)2011–12 onwards (2)All years (3)
Coefficient95% CICoefficient95% CICoefficient95% CI
Triple differences−0.617−0.941 to  −0.297−0.259−0.573 to  0.052−0.238−0.446 to  −0.031
Triple differences(alternative grouping of CCGs)−0.774−1.134 to  −0.420−0.440−0.786 to  −0.099−0.234−0.461 to  −0.009
Number of observations105 74575 09944 472
Test of parallel trendsF(3, 7672)P valueF(1, 7667)P valueN/A
 4.890.0020.330.567
  • Values are regression estimations from weighted least squares models on the empirical logit transformation of the rate including practice condition-specific fixed effects, full interaction of year with condition; and full interaction of year with the dummy for (N of LAs). Weighted by the denominators used to calculate the rates. Robust SEs, clustered by practice.

  • The triple difference represents (the change over time in cervical screening rates for CCGs working with only one LA minus the change over time in cervical screening rates for CCGs working with more than one LA) minus (the change over time in unassisted birth rates for CCGs working with only one LA minus the change over time in unassisted birth rates for CCGs working with more than one LA).

  • This model uses an alternative grouping of CCGs based on the number of LAs they work with ([1 or 2] vs [>2]).

  • LDV also contains values of the dependent variable in each of the pre intervention years. Estimated only on post intervention years.

  • CCG, Clinical Commissioning Group; LA, local authority; N/A, not available.