Fig. 5

Partitioning algorithm computed by the R package “partykit” (version 1.2–20) according to HHSAF score and ABHR consumption. To ensure the presence of an appropriate sample size, a minimum number of 20 observations was required in each group (the same model without size constraints is shown in Supplementary Fig. S3). The best partition was obtained for an ABHR consumption threshold slightly above 57 ml/PD.