ICU COVID-19 Mortality Probability Predictions

SuperLearner ICU Mortality Predictions

The web application reports three Super Learner (SL) prediction tools developed and validated on the Intensive Care Unit (ICU) data.

The ICU death COVID-19 probabilities may be calculated defining the specific levels of predictors.

Select from the left side panel the prediction model. Three predictive models, trained on different sets of variables, are available:

- Model Set 1 (admission core set variables). The first model has been tuned considering only the variables collected at admission having at least the 85% of valid cases.

- Model Set 2 (admission variables). The second model has been tuned considering all the variables collected at baseline even though the valid cases were less than the 85% of the sample.

- Model Set 3 (admission + hospitalization variables). The third model has been tuned considering the variables collected during the hospitalization together with those measured at admission even though the valid cases were less than the 85% of the sample (Model Set 2).

Computations have been performed using the SuperLerner[1] package in R[2] (version 3.6.2).

References

1. Eric Polley, Erin LeDell, Chris Kennedy and Mark van der Laan (2021). SuperLearner: Super Learner Prediction. R package version 2.0-27-9000. https://github.com/ecpolley/SuperLearner

2. Team, R. C. (2015). R Foundation for Statistical Computing; Vienna, Austria: 2014. R: A language and environment for statistical computing, 2013.