We appreciate the response to our paper1 from Craig Ramsay and colleagues; we reached a conclusion similar to theirs when we initially analysed the data, recognizing that induction rates were already declining when we implemented the intervention. We did include statistical testing in early drafts of the paper but this information was omitted at the editors' request. Table 1 shows how logistic regression was used to calculate odds ratios for induction with the first time period (January–June 1994) as the reference category. The odds of having an induction in 1997 and 1998 were significantly less than in the reference time period; in contrast, the odds of having an induction in the time periods before 1997 did not differ from those in the reference time period.
Table 1.
We then created a logistic regression model with time periods 1, 2, … 10 entered as a continuous variable, which allowed us to compare the slope of the change in the induction rate for different groups of time periods. The slope of the change in induction rate for the period January 1994 to June 1996 is not different from that for July 1996 to December 1998. However, the induction rate appears to have stabilized after December 1997. If the slope for January 1994 to June 1996 is compared with that for July 1996 to December 1997, the p value for the difference in slopes is between 0.05 and 0.1. This finding is in agreement with the analysis presented by Ramsay and colleagues. What is perhaps most interesting is that 3 years post-intervention the rate has not returned to pre-intervention levels.
We did not undertake a cost–benefit analysis. We hope that other investigators will build upon our work by mounting a randomized controlled trial of an accelerated quality improvement process to reduce hospital induction rates.2
References
- 1.
- 2.