cghr-toronto/wen-et-al-2025-cava

Computer Assisted Verbal Autopsy: Comparing Large Language Models to Physicians for Assigning Causes to 6939 Deaths in Sierra Leone from 2019-2022 # Computer Assisted Verbal Autopsy: Comparing Large Language Models to Physicians for Assigning Causes to 6939 Deaths in Sierra Leone from 2019-2022 - Publication - Public Data ## Overview This repository contains code to reproduce findings from the paper. The code requires R to run InterVA-5 and InSilicoVA models, as well as Python and an OpenAI API key to run GPT models. Plots are generated using R, while diagrams are created using draw.io. Folders in this repository are organized in the following manner: - archive: past versions of files with large differences - bin: scripts meant to be run on the command line - data: anonymized public datasets, and their data dictionaries if available - src: code and files for models, plots, and diagrams - figures: draw.io files for diagrams - models: code to run each model - manuscript: files for manuscript with related files and figures - figures: figure image files generated from R and draw.io - files: additional files to submit with the manuscript Files of interest: - data/healsl_rd1to2_cod_v1.csv: Data of cause of death coding results from models and physicians for all verbal autopsy records - src/plots.Rmd: Code for generating figure plots using R ## Install 1. Install git 2. Install Python 3 3. Install R and RStudio 4. Install draw.io (to edit diagrams under src/figures) 5. If on Windows, install RTools 6. Open a command line interface and clone this repository `git clone github.com` 7. Navigate into the cloned folder `cd wen-et-al-2025-cava` 8. Create a tmp folder for temporary files not tracked by git `mkdir tmp` * This folder is used for model input and output data * Intermediate files from running code may also be stored here * The python environment also exists under `tmp/venv` 9. Setup the python environment using scripts in the bin folder with the following command: * MacOS/Linux: `source bin/setup_python.sh` * Windows: `bin\setup_python` 10. Install pycrossva v0.97 …