Predicting Future Care Requirements Using Machine Learning for Pediatric Intensive and Routine Care Inpatients

Figures

• 

  Figure 1.. Model calibration for the 2017–2018 independent cohort. The y-axis shows the expected proportion of ICU care areas for the time periods based on the risk intervals, and the x-axis shows the observed proportion of ICU care areas in the time periods. The line of identity is the dashed line. The limits of the risk intervals were initially defined by 1,100 equally spaced empirical quantiles. Some intervals were combined to have a larger number of risk values but ensuring 99% of the intervals were smaller than 0.007 resulting in all intervals containing at least 1,770 time periods and 1,063, 1,026, 1,034, and 1,028 risk intervals in the four models. Within each interval, we compute the average expected risk of ICU admission and the observed risk of ICU admission. The circles indicate the observed and the expected proportions of ICU 6-hr time periods over ascending Criticality Index intervals. 94.26%, 89.01%, 90.81%, and 90.64% of the risk intervals within each plot have a Cohen’s h value less than 0.2, implying there are mostly small effect size differences between the observed and expected proportions. A linear regression is reported in each panel with their respective R2, and the…

• 

  Figure 2.. Precision-recall curves for the 2017–2018 independent cohort classifying care as routine or ICU for the respective future time periods. The area under the precision-recall curves and 95% CIs are included in each panel. The areas were computed with integral approximations, and the CIs were computed using a logit method (24).