Article Text

Download PDFPDF

6ER-025 Importance of appropriate before-and-after quasi-experimental design to evaluate the impact of antimicrobial stewardship programmes: comparative results using statistical hypothesis testing or interrupted time series analysis
  1. T Lopez Viñau Lopez1,
  2. M Saez-Torres De Vicente2,
  3. L Garcia Martinez1,
  4. J Hernandez Parada1,
  5. G Ruiz Arca1,
  6. A Perea Perez1
  1. 1Hospital Reina Sofía, Pharmacy Unit, Córdoba, Spain
  2. 2Hospital Universitario De Badajoz, Pharmacy Unit, Badajoz, Spain


Background and importance Most antimicrobial stewardship programmes (ASP) use a before-and-after research design, which has a high risk of bias. Efforts to enhance the conduct of these quasi-experimental studies are urgently needed to more rigorously evaluate interventions.

Aim and objectives The aim was to compare the results of an interrupted time series analysis (ITS) versus statistical hypothesis testing in a before-and-after study to evaluate the impact of ASP on cephalosporin consumption in a tertiary university hospital.

Material and methods A quasi-experimental study was designed before (January 2013–January 2014) and during the intervention (February 2014–February 2016). We recorded the impact of ASP on cephalosporin consumption in defined daily dose (DDD)/1000 hospital stays according to the anatomical therapeutic chemical classification system. For this task, all patients prescribed cephalosporins were identified daily through the prescription system (Farmatools). Statistical hypothesis testing was conducted using the Mann–Whitney U test, evaluating means (SD). The null hypothesis assumed both periods had the same averages (p>0.05). In contrast, ITS regression analysis was carried out to compare time trends before and after the intervention. It was performed using a longitudinal segmented regression with a generalised least squares approach to estimate changes in level and/or trend after the intervention. Autocorrelation was considered using moving average autoregressive models. Normality of residuals was verified, and the autocorrelation structures were validated. We also calculated, for a time point equivalent to 2 years after ASP, relative differences between observed changes and estimated values expected in the absence of the intervention. Data analyses were performed with R software, V.3.6.1. A p value <0.05 (two tailed) was considered significant.

Results Results of statistical hypothesis testing showed a significant increase in cephalosporin consumption (83.12 (SD 12.35) vs 104.87 (SD 10.48); p<0.001) in the intervention period. However, ITS regression analysis showed that the intervention led to a significant change in trend of −1.90 DDD/1000E, moving from a pre-intervention upward slope of 2.17 DDD/1000E to an almost horizontal slope of 0.27 DDD/1000E. Therefore, 2 years after the intervention, there was a significant decrease in measured consumption compared with that expected of −28.07%.

Conclusion and relevance Although both quasi-experimental designs showed significant changes in cephalosporin consumption after the intervention, the interpretation of results was contradictory. While hypothesis testing showed an increase after the intervention, ITS analysis revealed that this consumption was even less than expected. This suggests the programme may have been useful in reducing the consumption of these antimicrobials. Therefore, a robust design is essential in ASP, enabling appropriate interpretation of the results.

Conflict of interest No conflict of interest

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.