In-sample | Out-of-sample | |

Hospital-IV-days | ||

AUC for the probability of a zero count | 0.809 (0.802 to 0.815) | 0.807 (0.798 to 0.815) |

Bias (days) | 0.089 (0.041 to 0.133) | 0.078 (−0.447 to 0.650) |

RMSE (days) | 17.13 (16.34 to 17.87) | 17.18 (16.26 to 18.53) |

Observed total (days) | 191 143 (185 875 to 196 754) | 70 325 (67 495 to 73 321) |

Predicted total (days) | 192 745 (187 173 to 198 218) | 70 841 (68 249 to 73 514) |

Home-IV-days | ||

AUC for the probability of a zero count | 0.822 (0.815 to 0.828) | 0.820 (0.811 to 0.828) |

Bias (days) | 0.092 (0.038 to 0.151) | 0.063 (–0.424 to 0.589) |

RMSE (days) | 15.83 (14.51 to 17.85) | 15.92 (14.40 to 19.78) |

Observed total (days) | 149 940 (144 790 to 154 897) | 55 215 (52 810 to 57 748) |

Predicted total (days) | 151 585 (146 256 to 156 826) | 55 627 (53 320 to 58 318) |

The model fitted in each bootstrap sample was evaluated in-sample and out-of-sample. Estimated 95% CI are given in parentheses and were obtained using the 2.5th and 97.5th percentiles across the 1000 bootstrap samples.

AUC, area under the curve; RMSE, root mean squared error.