Fig. 4. Ranking of the ten most important features driving the physicians’ predictions.

n =  400 patients and nine transplant physicians. We ranked the features importance for each physician based on the mean decrease accuracy from each physicians’ Random Survival Forest. Each plot expresses how much accuracy is lost in the prediction by excluding each parameter. The features are presented from descending importance. The higher the value of mean decrease accuracy, the higher the importance of the feature in the model. The cg and i Banff scores were not found in any physicians; the t Banff score was found in 3 (33%) physicians and the MFI of the immunodominant anti-HLA DSA was found in 6 (66%) physicians. ABMR antibody mediated rejection, i interstitial inflammation Banff score, t tubulitis Banff score, g glomerulitis Banff score, ptc peritubular capillaritis Banff score, cv arterial intimal fibrosis Banff score, ah arteriolar hyalinosis Banff score, c4d c4d staining of peritubular capillaritis Banff score, IFTA interstitial fibrosis and tubular atrophy, MFI mean fluoresence intensity, DSA donor specific antibody, ECD expanded criteria donor, eGFR estimated Glomerular Filtration Rate.