RadaR (Rapid analysis of diagnostic and antimicrobial patterns in R) - an interactive open source software tool

Christian Luz, Matthias Berends, Jan-Willem Dik, Nienke Beerlage-de Jong, Mariëtte Lokate, Corinna Glasner, Bhanu Sinha

Research output: Contribution to conferencePosterAcademic

111 Downloads (Pure)


Background: Analysing outcome and quality of care indicators for infectious patients in an entire hospital requires processing large datasets, accounting for numerous patient parameters and treatment guidelines. Rapid, reproducible and adaptable analyses usually require substantial technical expertise. We describe a dashboard tool (RadaR) allowing user-friendly, intuitive and interactive analysis of large datasets without prior in-depth knowledge. This tool was developed for studying the effect of taking blood cultures on length of stay (LOS) and antibiotic consumption in patients receiving intravenous (IV) antibiotics at an academic tertiary referral hospital. RadaR handled a modelling dataset of more than 80,000 patients (eight years, 59 sub-specialties, 35 different antibiotic agents).

Materials/methods: RadaR was built in R (version 3.4.2), an open source programming language using Shiny package (version 1.0.5), a web application framework for R. Analytical graphs are generated with ggplot2 and survminer packages. The source code and additional required R packages for RadaR can be found at github.com/ceefluz/radar with a running example at ceefluz.shinyapps.io/radar.

Results: RadaR visualizes analytical graphs in an interactive manner within seconds. Users can control different input variables: time of blood culture taken, study year, patient age, specialty, admission route and antibiotic agents. For a predefined grouping variable (e.g. blood cultures taken vs. not taken) in the selected patient population RadaR automatically calculates the following: LOS distribution, animated LOS distribution over time, Kaplan-Meier estimates for hospital discharge, frequencies and ratios in antibiotic prescriptions, antibiotic consumption (in DDD) and mortality. Stratification can be done for (sub-)specialties, admission route, age, gender, admissions per quarter and antibiotic agent. Moreover, multiple logistic and Cox regression analysis in RadaR allows to investigate the grouping variable further. Finally, datasets of identified groups can easily be downloaded for further analysis.

Conclusions: This tool enables intuitive, rapid and reproducible quality of care pattern analysis of infectious patients without prior software experience. Hence, it facilitates understanding and communication of important trends, performances and patient outcome. We have started using RadaR to investigate blood culture use at our institution. However, due to its open source nature this tool can be easily adapted to different objectives and settings.
Original languageEnglish
Publication statusPublished - 20-Jun-2018
EventEuropean Congress of Clinical Microbiology and Infectious Diseases (ECCMID) 2018 - IFEMA, Madrid, Spain
Duration: 21-Apr-201824-Apr-2018


ConferenceEuropean Congress of Clinical Microbiology and Infectious Diseases (ECCMID) 2018
Abbreviated titleECCMID
Internet address

Cite this