Intended for healthcare professionals

Editorials

Improving outpatient antibiotic prescribing

BMJ 2019; 364 doi: https://doi.org/10.1136/bmj.l289 (Published 13 February 2019) Cite this as: BMJ 2019;364:l289

Linked Research

Effectiveness and safety of electronically delivered prescribing feedback and decision support on antibiotic use for respiratory illness in primary care

  1. Lauri A Hicks, director1,
  2. Laura M King, health research analyst1,
  3. Katherine E Fleming-Dutra, deputy director1
  1. 1Office of Antibiotic Stewardship, Division of Healthcare Quality Promotion, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, 1600 Clifton Road NE, Mailstop A-31 Atlanta, GA 30329, USA
  1. Correspondence to: L A Hicks auq3{at}cdc.gov

Harness the power of data

Outpatient prescriptions account for an estimated 85-95% of the volume of antibiotics used in people, and antibiotics are frequently overused and misused in outpatient settings.123 Optimizing antibiotic use in outpatient settings is increasingly recognized as an opportunity to improve patient safety.4 Two studies in The BMJ illustrate how to harness the power of outpatient antibiotic prescribing data to inform quality improvement.56 Data are critical to identify opportunities for improvement to inform action, track and report antibiotic use, and evaluate the impact of interventions.4

A first step to inform action is to assess the appropriateness of antibiotic prescribing. Chua and colleagues assigned antibiotic appropriateness categories—always, sometimes, or never—to 91 378 ICD-10-CM (international classification of diseases, 10th revision, clinical modification) codes, a commendable feat.5 They applied this scheme to a claims based dataset from a convenience sample of privately insured individuals under age 65 in the United States and found that 23% of antibiotic fills corresponded to visits with only “antibiotics never appropriate” diagnoses. This scheme can be applied by public health officials, health systems, clinicians, and researchers to datasets with ICD-10-CM diagnosis codes to identify opportunities to improve antibiotic prescribing. Notably, the authors found that 28% of prescriptions were not associated with a recent, captured medical encounter, highlighting a challenge of using administrative data to identify opportunities to improve antibiotic prescribing. Nonetheless, the study by Chua and colleagues provides a valuable roadmap for assessing antibiotic prescribing that can be applied from the national level down to the individual clinician depending on the data source.

Having a roadmap to follow is key, but taking action is an important next step. Gulliford and colleagues share findings from the REDUCE trial, a large cluster-randomized trial in 79 general practices in the United Kingdom examining the effect of a low cost, electronic health record based, antibiotic stewardship intervention leveraging antibiotic use data.6 The intervention included a short webinar, practice level monthly feedback reports on antibiotic prescribing for respiratory tract infections (RTIs), RTI decision support tools embedded in the electronic health record, and stewardship champions at each practice. After 12 months, this multifaceted intervention resulted in a 12% reduction (95% confidence interval 1-22%) in the antibiotic prescription rate for RTIs in intervention compared with control practices. However, the intervention did not result in any differences in the consultation rate for RTIs, the proportion of RTI visits with antibiotics prescribed, or the total antibiotic prescribing rate for all conditions.

The authors concluded that this intervention was moderately effective in reducing antibiotic prescribing for RTIs in adults, but did not affect overall antibiotic prescribing. Although RTIs are the most common diagnoses leading to outpatient antibiotic prescriptions, most outpatient antibiotic prescriptions are for other diagnoses. In order to drive antibiotic prescribing improvements and prevent diagnostic shifting (that is, clinicians changing a code to justify prescribing), antibiotic stewardship interventions should ideally incorporate prescribing audit and feedback for both high priority conditions and overall.4

Lessons learnt from behavioral science might shed some light on the modest effect of the REDUCE trial. Firstly, owing to data limitations, Gulliford and colleagues were only able to track and provide feedback on antibiotic prescribing data at the practice level. While audit and feedback at the practice level has been effective, especially when included in a package of interventions,7 variation in antibiotic prescribing is driven primarily by differences in prescribing patterns between individual clinicians.8 Therefore, when possible, providing clinician level feedback is preferable.

Secondly, the design and delivery of feedback is critical, and the incorporation of behavior change strategies into audit and feedback interventions appears to improve efficacy.910 Comparing clinicians’ antibiotic prescribing practices to those of their peers has been shown to be quite effective and is based on the idea that providing information about social norms (how others normally behave) drives individuals to bring their behavior in line with the norm.9 Additionally, comparing clinicians to top performing peers drives performance toward the goal rather than toward the mean.10 Intervention practices in the REDUCE trial were given comparisons of their current prescribing against their own baseline data. Providing social norm feedback with comparisons to top performing practices could have provided benefit.

Antibiotic stewardship leaders, clinicians, and researchers can apply the roadmap of Chua and colleagues to characterize antibiotic prescribing and lessons learnt from the REDUCE trial to leverage data for action and improve antibiotic use. Prescribing is as much a behavior as a rational clinical decision.1112 Thus, when leveraging data to improve antibiotic prescribing, the way in which antibiotic prescribing is measured and provided back to clinicians needs to be carefully considered.

Acknowledgments

The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Footnotes

References