Enhancing antimicrobial surveillance: an automated, dynamic and interactive approach

Abstract

Background: Antimicrobial resistance (AMR) is a global patient safety issue, which is threatening our ability to treat common infectious diseases. There is a growing interest in the potential secondary use of electronic clinical records to support antimicrobial stewardship programs to combat AMR. In addition to standard resistance measures, an antibiotic spectrum index and a novel method to quantify resistance trends are presented. A portable, dynamic and interactive visualization tool has been developed to facilitate data exploration and comprehensive reporting among clinicians and policy makers. Methods & Materials: Rather than limiting AMR evaluation to particular microorganisms, automated analysis of over 3.5 million susceptibility tests was undertaken. The resistance index indicates the ratio of laboratory cultures to which an antibiotic is resistant. Hence, an antibiotic is considered effective for resistance indexes lower than a threshold. The distribution of such indexes is used to define the antibiotic spectrum index. Furthermore, the sliding-time window technique is used to compute resistance time-series by defining a fixed region, which is moved across time. This method drops the dependence between granularity and accuracy of traditional surveillance systems and enhances further analysis. Consequently, weighted linear regression analysis has been applied on the time-series to numerically quantify resistance trends. Results: The use of sliding-time window surveillance produces visually more pleasant time-series and valid regression models for a larger amount of pathogen-antibiotic pairs (increase of 40%). Also, it captures seasonal variations (durbin-watson < 0.8) and measures resistance trends accurately (p-values < 0.05). Approximately half of the susceptibility tests correspond exclusively to Streptococcus Aureus and Escherichia Coli, revealing a clear over-testing tendency on current susceptibility guidelines. Their resistance indexes and trends have been thoroughly compared with existing research and public health reports and demonstrate to be congruent. Conclusion: Surveillance is the cornerstone for assessing the burden of AMR. Insights extracted from this system should be considered to enhance existing guidelines and strengthen AMR knowledge, especially in situations with sparse data where local directions are required. This system is being integrated into a clinical decision support system (Enhance Personalize and Integrated Care for Infection Management at Point of Care - EPIC IMPOC) with great potential to revamp prescription practices.