Heart-Transplant-Rankings

Code files corresponding to the manuscript The Accuracy of Initial US Heart Candidate Rankings by Pezler, Zhang, et al., published in JACC: Heart Failure. Full manuscript available at https://pubmed.ncbi.nlm.nih.gov/37052549/.

All files are tested in R 4.2.2 and Python 3.10.0. Files should be run in this order:

1. Survival_Data_Initial_Processing_UC.ipynb (Python 3.8.8)

Uses raw SRTR data and performs initial cleaning procedures: inclusion-exclusion criteria, derived variables. Splits data into the early cohort/training data from 2010-2017 and the late cohort/test data from 11/2018 to 3/2020.

2. Survival_Data_Descriptive_Stats_UC.ipynb (Python 3.8.8)

Derives descriptive statistics for Table 1. Log-rank and chi-squared tests are used to determine differences. Optional file.

3. Survival_Data_Output_for_R_UC.ipynb (Python 3.8.8)

Normalizes data, adds status ranking, saves data to csv files for processing in R.

4. Waitlist_Survival_Data_Processing.Rmd (R 4.2.2)

Reads in csv files, imputes missing data using mice.

5. CPH_and_RF_Calculations.Rmd (R 4.2.2)

Runs the primary models, Cox Proportional Hazards and Random Survival Forest, with and without treatments. Includes hyperparameter tuning.

Updated 10/2022: Adds the second rcorrp.cens method.

6. Variable_importance.Rmd (R 4.2.2)

Calculates single and joint variable importance in the Random Survival Forest models.

7. 6statusC.Rmd (R 4.2.2)

Calculates Harrell's c-indices (Harrell et al, Evaluating the yield of medical tests, JAMA, 1982) based on the 6-status ranking at listing for the late cohort, treating lower status as equivalent to a higher risk prediction. Bootstrapping is used to calculate the confidence intervals.

8. Brier_scores.Rmd (R 4.2.2)

Calculates Brier scores (Brier, Verification of forecasts expressed in terms of probability, Mon Weather Rev, 1982).

9. Bootstrapping_c_indices_Harrells.Rmd (R 4.2.2)

Uses bootstrapping to find confidence intervals for Harrell's c-indices.

10. Bootstrapping_c_differences.Rmd (R 4.2.2)

Uses bootstrapping to find confidence intervals for Wald tests of differences.