Table 1.
Transfer patient characteristics, and mortality rates for the development phase of SafeNET.
Table 2.
Completeness of data and respective values for patient variables that have the most important variables predictive of mortality by mortality status.
Fig 1.
Comparison of discrimination and calibration during model development.
Calibration plots and discrimination outcomes of guided machine learning are displayed for models using all 54 patient variables (A-D) vs. the top 14 most important variables (E-H) to predict in-hospital, 30-day, and 90-day mortality and CMO/hospice transition among hospital transfer patients.
Fig 2.
Comparison of receiver operating characteristic (ROC) curves for SafeNET and qSOFA.
The ROC curves and discrimination of the SafeNET (Safe Nonelective Emergent Transfers) and qSOFA (quick Sequential Organ Failure Assessment) models are displayed.
Fig 3.
SafeNET compliance over time by hospital facility.
Compliance rates are displayed during a 7-month pilot period (A) and broken down by participating facility (B).
Fig 4.
Discrimination and calibration of SafeNET models for predicting mortality.
Validation of SafeNET for predicting in-hospital (A,D), 30-day (B,E), and 90-day (C,F) mortality among transfer patients is shown.
Table 3.
Sensitivity, specificity, and predictive values of prospectively measured SafeNET score at a threshold of 20% mortality.