11 Automated Generation of Cumulative Antibiograms Using a Novel R-based Approach. (11th January 2018)

• Record Type: Journal Article
• Authors: Nowak, Michael; Hargrave, Judith; Jenkins, Stephen; Westblade, Lars
• Is Part Of: American journal of clinical pathology. Volume 149(2018)Supplement 1
• Page Start: S168
• Page End: S169
• Publication Date: 2018-01-11
• Subjects: Diagnosis, Laboratory -- Periodicals; Pathology -- Periodicals; 616.07
• DOI: 10.1093/ajcp/aqx149.380
• Languages: English
• ISSNs: 0002-9173

Abstract

Abstract: It is imperative for health care facilities to publish a cumulative local antibiogram at regular intervals in order to support clinical decision-making and development of evidence-based empiric antibiotic therapy guidelines. The Clinical and Laboratory Standards Institute (CLSI) provides detailed information on the preparation of an accurate and consistent cumulative antibiogram. However, numerous challenges are associated with the collection of data and implementation of CLSI guidelines. Manual methods of data processing can be time consuming and prone to error, while automated commercial systems may not comply with the most current guidelines or may lack the flexibility to adapt to changes in laboratory practice. To address this, we sought to develop a fully automated method for the generation of cumulative antibiograms that adheres to CLSI guidelines. First, a complete source of antimicrobial susceptibility test (AST) data was needed. A generic Cerner Millennium report was identified that contained the required data, including results that were not reported in the electronic medical record (EMR) due to cascade reporting rules, which suppress AST results of secondary agents when an isolate is susceptible to primary agents. Next, a script was developed using the open source R programming language to parse data and apply a series of filters to generate an organized database of AST results appropriate for inclusion in the cumulative antibiogram. Filters included elimination of duplicate isolates of the same organism from the same patient during a defined time period, as well as removal of inappropriate specimen types such as screening, epidemiological, or laboratory proficiency cultures. Using Microsoft Excel, we developed an interface to interactively explore the frequency of organisms isolated in specific subsets as defined by specimen source, collection date, patient location, or patient age. A cumulative antibiogram can be generated for specific subsets of interest, and antibiotic susceptibility data easily compared to another subset, with statistically significant differences calculated using the χ 2 test. User-defined antibiotic combinations can be included in the results, which are calculated using a logical function in Excel. Comparison of our fully automated method to a fully manual data analysis from Cerner Millennium for a 1-month period showed only rare differences, predominantly caused by human error with the manual method. Compared to our hospital's previous semiautomated method, which drew data from the EMR for a 1-year period, our fully automated approach exhibited notable differences that were predominately due to the inclusion of results suppressed by cascade reporting, and resulted in superior accuracy with our fully automated method. In addition, our fully automated method allows subset analyses that were previously unavailable, requires no manual data processing, and can be accomplished rapidly.