The TimeMetric (formerly PAmeasure) R package offers a comprehensive framework for evaluating prediction performance in survival models, including scenarios with right-censoring, competing risks, as well as under different designs such as nested case control and case-cohort designs. It provides a comprehensive suites of metrics such as pseudo $R$-squared, concordance indices, Brier Score, and time-dependent AUC, unified under a single platform. This paper presents an overview of the mathematical definitions of these metrics, implementation details, and application examples of these measures. Demonstrations using simulated and real-world datasets validate the utility and robustness of TimeMetric, making it a valuable tool for researchers and practitioners in survival analysis.

The package focuses on four key categories of performance evaluation:

• R²-related metrics: quantify the proportion of variability in survival times explained by the model.
• Concordance indices (C-indices): measure the discriminatory power of the model.
• Time-dependent AUC: evaluate model discrimination at different time points.
• Brier Score: assess the accuracy of probabilistic survival predictions.

The TimeMetrics package is organized into six main components:

• pam.coxph_restricted and pam.survreg_restricted: functions for generating predicted survival times from Cox and parametric survival models.
• pam.predicted_survival_eval: the primary function for computing survival performance metrics.
• pam.predict_subject_cif: calculates cumulative incidence function (CIF) predictions for individual subjects across multiple types of competing risks models.
• pam.predicted_survival_eval_cr: extended version for evaluating predictions in the presence of competing risks.
• pam.predicted_survial_eval_casecohort: evaluates survival model performance in a case–cohort setting using predicted survival probabilities and case–cohort sampling weights
• pam.predicted_survial_eval_ncc: evaluates survival model performance under a nested case–control (NCC) design using predicted survival probabilities and NCC sampling weights.

# 1. install.packages("remotes")     # if not already installed
remotes::install_github("toz015/TimeMetric", build_vignettes = TRUE)

Developed by Li’s Lab (UCLA), with coding and intellectual contributions from: Tong Zhu, Zian Zhuang, Wen Su, Xiaowu Dai, and Gang Li.