An Interrupted Time Series Analysis of Parental Antibiotic Use in Colombia

• Activity Number: 17
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section*
• Author(s): Adriana Perez*+ and Amparo Castro and Rodolfo Dennis
• Affiliation(s): University of Texas, Houston/Pontificia Universidad Javeriana and Universidad Nacional de Colombia and Pontificia Universidad Javeriana
• Keywords: ARIMA model ; Intervention ; Pharmacoepidemiology ; Modelling ; Diagnosis ; Analysis

Abstract

A University-based hospital developed and implemented an educational intervention, as well as a structured antibiotic order form. This intervention was performed after assessing the appropriateness of the observed antibiotic-prescribing practices, using a quasi-experimental study. An application of interrupted time-series intervention analysis was conducted in three antibiotic groups of the hospital weekly rate of incorrect prescriptions. Pre-intervention ARIMA models were identified, estimated, and diagnosed for four time series. The first time series was an ARIMA (0,1,2) with corresponding estimates and standard error (SE) as theta1=0.36 (SE=0.102) and theta2=0.49 (SE=0.101), respectively. The second time series was an ARIMA(0,1,1) with corresponding estimate =0.82 and SE=0.07. The third time series was an ARIMA(0,0,1) with corresponding estimate=-0.72 and SE=0.08. Surgery antibiotic prophylaxis series was an ARIMA (0,1,1). These models were used in the post-intervention series to test pre-post series level differences. An abrupt constant change was significant in the 1st, 3rd and 4th time series indicating a 47%, 7.3% and 20% reduction on incorrect prescriptions after intervention