P145: The role of audit and feedback in the ED setting: are physicians able to accurately predict their own practice?

Abstract

Introduction: Prior research has shown that audit and feedback (A &F) can be an effective tool for practice change. However, questions remain about how to optimize A&F. The objectives of this project were to determine if: 1) there are differences in practice between physicians who do, and do not, consent to receive a confidential report on their practice and; 2) if there is a relationship between consenting physicians self-predicted and actual practice. Methods: This was a prospective, cross-sectional study embedded in a larger quality improvement (QI) initiative to align physician practice with best evidence in the emergency department (ED) care of infants with bronchiolitis. All physicians practicing in the ED of a tertiary care pediatric hospital were offered the opportunity to consent to receive an individual, confidential data report on their practice. Prior to receiving their data, consenting physicians completed a survey which asked them to predict the proportion of bronchiolitic patients for whom they ordered diagnostic tests or treatments. We used chi-squared testing to compare the proportion of consenting and non-consenting physicians whose diagnostic test (Chest X-ray (CXR), viral study) and treatment (steroid, Ventolin) ordering was above the median for all ED physicians. We used Pearsons correlation to assess the relationship between consenting physicians self -predicted and actual practice. Results: 56% (37/66) of physicians consented to receive a data report. The median proportion of patients with an x-ray ordered was 20%, 63% of non-consenters were above the median, compared to 36% of consenters (X2 (1, N=66)=4.91 p=0.03). For viral testing, 31% of patients had a test ordered, with 50% of non-consenters and 50% consenters above the median (( X2 (1, N=66) =0 p=1); 11% of patients had steroids ordered, with 53% of non-consenters and 47% of consenters above the median ( X2 (1, N=66)=0.24 p-0.621); and 18% of patients had Ventolin ordered, with 60% of non-consenters and 42% of consenters above the median ( X2 (1, N=66) =2.2 p=0.138). There was a moderate correlation between physicians predicted and actual practice with respect to viral testing (r=0.67), but minimal correlation for CXR (0.05), steroids (r=0.17) or Ventolin (r=0.33) ordering. Conclusion: The finding that physicians have a limited ability to accurately predict their own performance emphasizes the importance of providing physicians with feedback. However, our results suggest that the consent process may be a potential barrier to effective A &F.