Effect of Electronic Prescribing on Medication Error Rates: A Comparison of Poisson Regression Methods

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Yolanda Barrón*+ and Lisa Kern and Erika Abramson and Alison Edwards and Rainu Kaushal
• Companies: Weill Medical College of Cornell University and and and and
• Keywords: Poisson regression ; Zero Inflated Poisson Regression ; Clustered Data ; Overdispersion ; Count and Rate Data ; Medication errors

Abstract

Medication errors are an outcome best captured as counts or rates. Poisson regression is increasingly being used to determine the impact of interventions, such as electronic prescribing, on reducing medication errors. However, there are several features of medication error studies that may require advanced applications of Poisson regression. We will compare traditional Poisson regression to two types of modified Poisson regression. Zero-inflated Poisson regression can account for the high proportion of prescriptions that have no errors. Cluster-adjusted Poisson regression can account for including multiple prescriptions per patient and multiple prescriptions per provider. We will use a data set from a community-based study on electronic prescribing to illustrate the impact of these methods.