Table 4.

Results from multi compound logistic regression presented in terms of mean area under the curve (AUC) and its standard deviation (SD) as calculated by Monte Carlo cross validation

In addition, mean balanced accuracy is presented to aid interpretation (with threshold of 50%). A model based on existing clinical data predicts severity very well mainly because FiO2 is key to determining severity. The same model in COVID-19 outcome provides a reasonable performance. In both cases the addition of KYNA/TRP ratio improves the model and in COVID-19 outcome KYNA/TRP offers more predictive power than clinical data