Liver transplant adherence analysis

This repository contains the R code used to reconstruct questionnaire scores and conduct the inferential, sensitivity, and exploratory prediction analyses for the liver transplant adherence study.

Repository contents

.
├── analysis.R
├── README.md
├── data/
│   └── README_data_access.md
└── results/                 # generated locally; not committed

The participant-level dataset is not included in the public repository.

Required input

The script expects the analytic SPSS file at:

data/Nazari_Liver_Transplant_Clean.sav

A different location can be supplied without editing the script:

Sys.setenv(ANALYSIS_DATA_FILE = "path/to/analytic_file.sav")
Sys.setenv(ANALYSIS_OUTPUT_DIR = "path/to/results")
source("analysis.R")

Reproducibility

Run the analysis from the repository root. The script records the random seed, input-file SHA-256 hash, package versions, session information, resampling configuration, and generated output inventory.

Before the final repository release:

1. verify the questionnaire scoring keys and the ATQ cutoff;
2. run the script on the locked final analytic dataset;
3. reconcile every regenerated estimate, table, and figure with the manuscript;
4. inspect model and resampling error logs;
5. create or update renv.lock from the successful final run;
6. confirm that no participant-level data or fitted model objects are tracked by Git.

Data confidentiality

The repository must not include the SPSS dataset, participant identifiers, source-row numbers, duplicate-cluster details, row-level predictions, or fitted workflow objects. Public repository runs use:

EXPORT_ROW_LEVEL_OUTPUTS <- FALSE
SAVE_FITTED_MODELS <- FALSE

Scope of the prediction analyses

The machine-learning analyses are exploratory and use grouped nested repeated cross-validation. Potential duplicate clusters remain in the same resampling partition. The repository does not provide a clinically deployable model and does not constitute external validation.