This analysis code relates to the study: Mathiszig-Lee JF*, Catling FJR*, Moonesinghe SR, Brett SJ (*equal contribution). Highlighting uncertainty in clinical risk prediction using a model of emergency laparotomy mortality risk. npj Digital Medicine 2022;5:1–8.
An online calculator and API for the production version of the RUNE model is available at laparotomy-risk.com. This model has not yet been certified as a medical device and therefore should only be used for educational and research purposes and should not be used to inform patient care.
Please note that RUNE is referred to as 'the novel model' throughout this codebase.
The NELA dataset is required to run the full analysis, and is not included in this repository. Under the terms of the data sharing agreement for this study we are unable to share the source data directly. Requests for anonymous patient-level data can be made directly to the NELA Project Team.
Without these data, it is still possible to set up the project environment and run the tests.
This codebase is written in Python 3.8.6 and R 4.0.3. We manage dependencies with pipenv and renv. To set up your project environment, you'll need to:
- Install Python and pip
- Install pipenv
- Install R and renv
If working in the Imperial BDAU, pipenv isn't in path. We can define a convenient alias for it in this case:
alias pipenv='/opt/python/3.8.6/bin/pipenv'Then install dependencies with pipenv install and renv::restore()
The shell script run_full_analysis_on_bdau.sh runs the full analysis (not the initial enironmental setup) on the Imperial BDAU. It runs all scripts in sequence inside our pipenv and renv environments, restricting itself to 8 cores.
Enter the project's python environment:
pipenv shellThen run the python scripts in the root directory in numerical order, i.e. starting with:
python 00_initial_data_wrangling.pyInstall development packages with:
pipenv install --dev Run tests for the utils module and check coverage:
pytest --cov=utils tests/