Identifying hospital antimicrobial resistance targets via robust ranking

ABSTRACT

We develop a robust ranking procedure to uncover trends in variation in antibiotic resistance (AR) rates across hospitals for some antibiotic-bacterium pairs over several years. We illustrate how the method can be used to detect potentially dangerous trends and to direct attention to hospitals’ management practices. A robust method is indicated because some unusual reported resistance rates may be due to measurement protocol differences and not any real difference in AR rates. Our proposed method is less sensitive to outlier observations than other robust methods. The application on real AR data shows how a dangerous trend in a particular AR rate would be detected. Our results indicate the potential benefits of systematic AR rate collection and AR reporting systems across hospitals.

KEYWORDS:

• Hospital management
• antimicrobial resistance
• robust ranking
• L1-norm line-fitting

Acknowledgment

We are grateful to reviewers who contributed generously to improving this work.