Accepted in PLOS One
Full paper

This work introduces a multi-event survival model that predicts when ALS patients are likely to experience functional decline across five key abilities. Using the PRO-ACT dataset (N=3220), the method provides individualized survival curves for each function and enables counterfactual predictions to assess how changes in clinical variables affect expected decline.

To run the models, please refer to Requirements.txt. The code was tested in a virtual environment with Python 3.9.

1. Install requirements by running: pip install -r requirements.txt
2. Download the PRO-ACT dataset from the PRO-ACT website
3. Make directories mkdir data and mkdir results
4. Put the PRO-ACT files inside the data directory and run python src/make_dataset.py
5. Train the survival models by running python src/train_models.py.

If you find this paper useful in your work, please consider citing it.

@article{lillelund_meaningful_2025,
    doi = {10.1371/journal.pone.0336476},
    author = {Lillelund, Christian Marius and Kalra, Sanjay and Greiner, Russell and {The Pooled Resource Open-Access ALS Clinical Trials Consortium (PRO-ACT)}},
    journal = {PLOS ONE},
    title = {A meaningful prediction of functional decline in amyotrophic lateral sclerosis based on multi-event survival analysis},
    year = {2025},
    volume = {20},
    pages = {1-16},
    number = {11},
}