Personalized Preoperative Prediction of the Length of Hospital Stay after TAVI Using a Dedicated Decision Tree Algorithm

Figures

• 

  Figure 1. Decision Tree Algorithm for LoS > 6 days. Legend Figure 1. A decision tree is an algorithm that recursively divides the training data, based on certain splitting criteria, to predict a given target. This decision tree starts with all the participants in the root node (Node 0), then divides these participants into those with GFR less than 33 mL/min/1.73 m2, and those with GFR greater than or equal to 33 mL/min/1.73 m2; for all participants with GFR less than 33 mL/min/1.73 m2, an additional separation was made between participants with KCCQ-scores greater than 30 and participants with KCCQ-scores lower than or equal to 30. LoS_gt_6_days = LoS greater than 6 days; GFR_lt_33_mL = GFR lower than 33 mL; KCCQle30 = KCCQ score lower than or equal to 30. In each node, 0.0 (zero) means that the participants within this node do not have the respective feature (e.g., GFR_lt_33_mL) and 1.0 means that the participants do have this feature. Depicted with blue (or 0.000) are participants with LoS equal or lower than 6 days and depicted with red (or 1.000) are participants with LoS > 6 days.

• 

  Figure 2. Distribution of LoS as a function of the calculated RS. Legend of Figure 2. Box plots depicting LoS (in days after TAVI), depending on the cumulative number of the aforementioned four predictors (RS_LoS range = 0−4). Within the box on the upper left corner of the figure the RS Score (0−4) is provided on the first column and the respective median numerical values, including the interquartile range within parentheses, are provided on the second column.