Table 2 Overview of the survival prediction methods used in the benchmark study

Random survival forests (rsf)

ranger (0.13.1)

For C-index calculation: Sum of the values of the bootstrap ensemble cumulative hazard function [27] \(\:{H}_{e}^{*}\left(t\right|{\varvec{x}}_{i})\) calculated at all unique death times.

For integrated Brier score: Survival function estimated using \(\:{\text{e}\text{x}\text{p}(-H}_{e}^{*}\left(t\right|{\varvec{x}}_{i}\left)\right)\)

Block forests (bf)

blockForest (0.2.4)

See rsf above.

Lasso (lasso)

glmnet (4.1-3)

For C-index calculation: Linear predictor \(\:{\varvec{x}}_{i}^{T}\widehat{\varvec{\beta\:}}\)

For integrated Brier score: Survival function estimated as follows: \(\:{\text{e}\text{x}\text{p}(-\widehat{{\Lambda\:}}}_{0}\left(t\right)\:\text{e}\text{x}\text{p}\left({\varvec{x}}_{i}^{T}\widehat{\varvec{\beta\:}}\right))\), where \(\:{\widehat{{\Lambda\:}}}_{0}\left(t\right)\) is an estimate of the baseline cumulative hazard function obtained using the Efron estimator

IPF-LASSO (ipflasso)

ipflasso (1.1)

See lasso above.

Priority-Lasso (prioritylasso)

prioritylasso (0.2.5)

See lasso above.