Skip to main content
Fig. 6 | Critical Care

Fig. 6

From: Machine learning derivation of four computable 24-h pediatric sepsis phenotypes to facilitate enrollment in early personalized anti-inflammatory clinical trials

Fig. 6

Heterogeneous treatment interactions and mortality risks among the phenotypes. Heatmap of Elastic Net Regression analysis shows the association between 14 individual therapies (diagonal values) and their 91 combination interactions (total cells = 105) with mortality in PedSep-B, C, and D among children who received anti-inflammatory therapies. The PedSep-A phenotype is not presented due to limited number of deaths. Blank cells have no patients. Values in each cell represent odds ratios of mortality, where 1 represents no association with mortality. Color in each cell represents direction of effect, where red represents mortality direction, green represents survival direction. Cells located at the diagonal are odds ratio of association from the 14 individual therapies. The other cells represent the mortality odds ratio of combinations of these therapies compared to all other combinations. For example, survivors in PedSep-D phenotype are less likely to be treated with IVIG than non-survivors (red), whereas survivors in PedSep-D are more likely to be treated with combined IVIG + methylprednisolone (green). This machine analysis method does not allow calculation of confidence intervals

Back to article page