This repository contains code to generate the results reported in Development and Validation of an Electronic Health Record-Based Machine Learning Model to Estimate Delirium Risk in Newly Hospitalized Patients.
The data have been redacted due to HIPAA requirements, but model training and validation are reported in caret_tuning.R.
The models reported in the manuscript are in /rda/models. They are named with the suffix _nudesc1 if the outcome was calculated using Nu-DESC≥1, else they correspond to the outcome calculated using Nu-DESC≥2.
To generate ROC curves and confusion matrices reported in the paper, run:
Rscript -e "rmarkdown::render('caret_analysis.Rmd')"
The results of this command are in caret_analysis.html
