Fig 1.
Clustering and Treatment completeness. (A). Relative euclidean distances calculated for the hierarchical clustering. The first data division in two big groups of different heterogeneity and amount is observed at a distance of 0.8. (B). Treatment completeness frequency. Treatment completeness frequency (1) versus not completeness (0) according to the cluster. Chi square was used to establish cluster and treatment completeness relations, showing a significant association with a p value of 2.4−28.
Fig 2.
Plots showing patient age distributions, living with aggressor, educational level and legal derivation in each cluster. T test analysis shows no significant differences between means of age: p = 0.756 (A). Chi square association shows no significance (p = 0.21) between living with aggressor state and cluster (B). Educational level frequencies in the two clusters, with 0 indicating no studies, 1: medium, 3: university, and 4: postgraduate studies show significant difference between both clusters, with a p value of 0.037 (C). Distribution of legal derivation (0 = NO, 1 = YES) with respect to the cluster shows significant association between both variables with a p value of 0.0063 (D).
Fig 3.
SHAP classification model explicability.
This beeswarm graph shows the specific impact on the 0 exit (towards negatives) or 1 (towards positives) of treatment completeness and depicts each parameter value measured in either discrete or continuous variables in color codes. The X-Axis shows each input parameter of the classification model organized according to its absolute impact.