Clinical risk stratification in necrotizing fasciitis | NPC statistics | Bootstrap inference | Python and R packages

NF Risk Stratification

nf-risk-stratification provides Python and R tooling for reproducible clinical risk stratification in necrotizing fasciitis. It supports descriptive clinical summaries, partial tests, Fisher/Tippett/Liptak nonparametric combination methods, bootstrap confidence intervals, and patient severity ranking.

The package is aligned with the methodology described in:

Permutation-Based Analysis of Clinical Variables in Necrotizing Fasciitis Using NPC and Bootstrap. Piscopo, G., Bandaru, S. T., Giacalone, M., & Longobardi, M. (2025). Mathematics, 13(17), 2869. https://doi.org/10.3390/math13172869

Installation

Install the Python package from the GitHub source tree until the first PyPI release is published:

python -m pip install "git+https://github.com/saitejabandaru-in/nf-risk-stratification.git"

Install the R development release from GitHub:

install.packages("remotes", repos = "https://cloud.r-project.org")
remotes::install_github("saitejabandaru-in/nf-risk-stratification")

You can also build locally from a clone:

R CMD build .
R CMD INSTALL nfRiskStratification_0.1.0.tar.gz

Project Links

Resource	Link
GitHub repository	https://github.com/saitejabandaru-in/nf-risk-stratification
Issues	https://github.com/saitejabandaru-in/nf-risk-stratification/issues

PyPI and GitHub release pages will be linked here after the first public release is published.

Required Data

The real clinical dataset is not included due to ethical and privacy restrictions. Place an anonymized CSV at:

data/nf_clinical_data.csv

Required columns:

Column	Meaning
HBA1C	Glycemic control
ALBUMINA	Nutritional status

Optional columns:

Column	Meaning
MORTO	Mortality outcome
AMPUTAZIONE	Major amputation outcome
PATIENT_ID	Patient identifier

An anonymized toy dataset is included for examples:

example_path <- system.file(
  "extdata",
  "nf_clinical_data_example.csv",
  package = "nfRiskStratification"
)

Quick Start

Python:

from nf_risk_stratification import read_nf_data, run_nf_pipeline

data = read_nf_data("data/nf_clinical_data.csv")
results = run_nf_pipeline(data)
print(results["npc"]["global_p_values"])
print(results["ranking"][:10])

R:

library(nfRiskStratification)

data <- read_nf_data("data/nf_clinical_data.csv")

descriptive_stats(data)
npc_analysis(data)
bootstrap_ci(data, reps = 1000)
head(rank_patients(data), 10)

Run the full pipeline in one call:

results <- run_nf_pipeline(path = "data/nf_clinical_data.csv")

results$descriptive
results$npc$global_p_values
results$bootstrap
head(results$ranking, 10)

Legacy Scripts

The original top-level scripts still work and now call the package functions:

source("descriptive_stats.R")
source("npc_analysis.R")
source("bootstrap_analysis.R")
source("ranking_analysis.R")

Package API

Function	Purpose
read_nf_data()	Read and validate an anonymized CSV dataset
validate_nf_data()	Validate required columns and normalize numeric fields
descriptive_stats()	Summarize clinical variables
npc_analysis()	Compute partial p-values and Fisher, Tippett, and Liptak global p-values
bootstrap_ci()	Compute percentile bootstrap confidence intervals for means
rank_patients()	Rank patients by aggregate clinical severity
run_nf_pipeline()	Run the complete workflow

Release Workflow

Tagged releases build and publish a source package automatically:

git tag v0.1.0
git push origin v0.1.0

GitHub Actions then runs package checks, builds nfRiskStratification_0.1.0.tar.gz, and attaches it to the GitHub release.

PyPI distributions are prepared from:

python -m build
python -m twine check dist/*

After the first upload succeeds, update this README with the public PyPI and GitHub release URLs.

Never commit PyPI API tokens or credentials. Use environment variables or a trusted local keyring only at upload time.

Repository Layout

R/                         Package source
python/                    Python package source for PyPI
man/                       Function documentation
tests/testthat/            Unit tests
tests_python/              Python smoke tests
inst/extdata/              Example anonymized dataset
data/                      Private local dataset location, not packaged
.github/workflows/         R package checks and release automation

Author

Sai Teja Bandaru
Data Scientist & Clinical Analytics Researcher
Universita degli Studi della Campania Luigi Vanvitelli