Abstract
We thank Johannes C van der Wouden and Ivo Smeele for their interest in our study. The differences in the reported estimates of effect can be explained by the statistical methods used in the analysis.
The primary analysis that was specified in our study protocol was based on aggregated weekly rates of catheter-associated asymptomatic bacteriuria and catheter-associated urinary tract infection (UTI).1 This is analogous to performing piecemeal interrupted time-series analyses using Poisson regression on multiple time series, with the intervention effect estimated by the step at the time of crossover. In this statistical model, calendar time was included as a covariate and hospital and intervention as fixed effects. As noted in figures 2 and 3 of the Article, an increasing pre-intervention trend might have exaggerated the impact of the intervention.1
The primary analysis that was specified in our study protocol was based on aggregated weekly rates of catheter-associated asymptomatic bacteriuria and catheter-associated urinary tract infection (UTI).1 This is analogous to performing piecemeal interrupted time-series analyses using Poisson regression on multiple time series, with the intervention effect estimated by the step at the time of crossover. In this statistical model, calendar time was included as a covariate and hospital and intervention as fixed effects. As noted in figures 2 and 3 of the Article, an increasing pre-intervention trend might have exaggerated the impact of the intervention.1
| Original language | English |
|---|---|
| Pages (from-to) | 1165-1165 |
| Number of pages | 1 |
| Journal | The Lancet Infectious Diseases |
| Volume | 19 |
| Issue number | 11 |
| DOIs | |
| Publication status | Published - Nov 2019 |
