PD145 - IMPROVING ACCESS TO THE NEONATAL EARLY-ONSET SEPSIS CALCULATOR: AN OPEN-SOURCE SCRIPT TO FACILITATE LOCAL USE, ELECTRONIC INTEGRATION AND FURTHER RESEARCH (ID 797)
Abstract
Backgrounds:
Kaiser Permanente Perinatal Research Division developed the neonatal early-onset sepsis (EOS) calculator – a risk prediction tool that has led to drastic reductions in unnecessary empiric antibiotics. A central, peer-reviewed publication documenting the exact mechanisms and allowing precise replication and integration of the EOS calculator tool is missing, making neonatal clinics worldwide dependent on the website of Kaiser Permanente.
Methods
After detailed technical assessment of the EOS calculator tool, we constructed a script for R statistical software, using previously published intercepts, coefficients (to provide a pre-examination risk) and likelihood ratios for the neonatal clinical status categories (to construct the post-examination risk) and incorporated the algorithm described by the EOS calculator publications. so that the script also provides users with right EOS calculator recommendation as an output. Validation of the script was done using with extreme input variables at the end of input ranges and a previously established clinical database of 234 verified EOS cases.
Results:
Components of the script reproducing exact results of the current online EOS calculator tool are shown in Figure 1. Preliminary validation of the tool by comparing results of the online tool and our script using maximum and minimum input values and a previously established database of neonatal EOS cases showed perfect agreement between the Kaiser Permanente online EOS calculator and our final script (234 of 234 cases, 100%).
Conclusions/Learning Points:
Success and widespread use of the EOS calculator warrant detailed documentation and access beyond the current online tool on the website of Kaiser Permanente. We present a validated open-source script providing the same functionality that will be publicly available, which may help facilitate electronic integration, reduce clinicians’ dependency, and improve scientific evaluation of the tool.
