comprisk 0.4.0

Python toolkit for competing risks: forest (RSF) + Fine-Gray subdistribution regression + Aalen-Johansen cumulative incidence + Gray's K-sample test + cause-specific Cox. Scales to n=10⁶ in ~1 min, 10–22× faster than randomForestSRC on real EHR data, scikit-learn-compatible.

comprisk

comprisk — a Python toolkit for competing risks. Ships a scalable, scikit-learn-compatible competing-risks random survival forest plus the three canonical regression / non-parametric methods clinical researchers actually need: Fine-Gray subdistribution-hazard regression, a stand-alone Aalen-Johansen cumulative-incidence estimator with cmprsk-parity variance, and cause-specific Cox PH (see Roadmap). Designed to remove the Python → R workflow split that applied researchers currently endure for competing-risks survival analysis.

Status: alpha. API and internals may change before v1.0. Renamed from crforest in 0.3.1 — pip install comprisk, from comprisk import CompetingRiskForest.

Highlights

• The four canonical CR methods, native Python. FineGrayRegression matches R cmprsk::crr() β̂ to floating-point noise (max |Δβ| = 1.4e-15 on three reference datasets); robust_se=True returns the Geskus cluster sandwich agreeing with cmprsk's IPCW-corrected SE to ~3 digits. CumulativeIncidence reproduces cmprsk::cuminc() to 1e-9 across CIF and variance. gray_test reproduces cmprsk::cuminc()$Tests to 1e-14. CauseSpecificCox matches survival::coxph(method="breslow") to 1e-9.
• Only native-Python competing-risks RSF. Cause-specific log-rank splitting + composite CR log-rank, Aalen-Johansen CIF, Nelson-Aalen CHF, Wolbers + Uno IPCW concordance, OOB Breiman VIMP, Ishwaran minimal-depth variable selection, exact TreeSHAP.
• CR-aware model evaluation. score_cr reports IPCW time-dependent AUC and Brier score under competing risks, plus integrated AUC / Brier (iAUC, IBS) with bootstrap CIs; calibration_cr returns tidy quantile- decile calibration data with per-bin Wilson intervals — one-call replacements for the CR-mode riskRegression::Score() / plotCalibration() blocks, taking a dict of named candidate models.
• 10–22× faster than randomForestSRC on real EHR data (CHF 14–22×, SEER 11.6×; full tables in docs/benchmarks.md), with C ≈ 0.85 on both libraries. ~95× faster than rfSRC built without OpenMP (default R-on-macOS).
• Order-of-magnitude faster than scikit-survival (16.6× at n = 5k, 544× at n = 50k), without disabling CIF/CHF outputs.
• Bit-identical to randomForestSRC with equivalence="rfsrc" — reproduces the per-tree mtry/nsplit RNG stream for paper-grade reproducibility, sensitivity checks, and rfSRC-baseline migrations.

comprisk vs alternatives

	comprisk	randomForestSRC	scikit-survival
Language	Python	R	Python
Native competing risks	✓	✓	✗ (single-event only)
Aalen–Johansen CIF output	✓	✓	n/a
Cumulative hazard at scale	✓	✓	✗¹
OOB permutation VIMP	✓	✓	✗
Bit-identical reproducibility mode	✓ (equivalence="rfsrc")	—	n/a
Scales to n = 10⁶	✓ (63 s on i7)	memory-bound past n ≈ 500 000 on consumer hardware	✗¹ / OOM²
Default parallelism	✓ (n_jobs=-1)	OpenMP (build-dependent; macOS Apple clang lacks it)	✓
GPU preview	✓ (CUDA 12)	✗	✗

¹ sksurv RandomSurvivalForest(low_memory=True) is the only mode that scales beyond ~10k samples, but it disables predict_cumulative_hazard_function and predict_survival_function (raises NotImplementedError). ² sksurv low_memory=False exposes CHF / survival outputs but stores per-leaf full CHF arrays; peak RSS reaches 16.8 GB at n = 5k on synthetic, OOMs (> 21.5 GB) at n = 10k on a 24 GB host.

Install

pip install comprisk          # or:  uv add comprisk
pip install "comprisk[gpu]"   # or:  uv add 'comprisk[gpu]'

Requires Python ≥ 3.10. Core dependencies: numpy, scipy, pandas, joblib, numba, scikit-learn. GPU extra adds cupy + CUDA 12 runtime libs (preview; faster only at low feature count today, full rewrite scheduled for v1.1).

Quickstart

import numpy as np
from comprisk import CompetingRiskForest

# Toy competing-risks data: 500 subjects, 6 features, 2 causes (+ censoring).
rng = np.random.default_rng(42)
n = 500
X = rng.normal(size=(n, 6))
time = rng.exponential(2.0, size=n) + 0.1
event = rng.choice([0, 1, 2], size=n, p=[0.4, 0.4, 0.2])  # 0 = censored

# Fit. Defaults: n_estimators=100, max_features="sqrt", logrankCR, n_jobs=-1.
forest = CompetingRiskForest(n_estimators=100, random_state=42).fit(X, time, event)

# Aalen-Johansen cumulative incidence over the forest's chosen time grid.
cif = forest.predict_cif(X[:5])                       # (5, n_causes, n_times)

# Cause-specific Wolbers concordance.
print("C-index, cause 1:", forest.score(X, time, event, cause=1))

Explainability and feature selection

# OOB permutation importance (Uno IPCW-scored).
vimp = forest.compute_importance(random_state=42)

# Ishwaran minimal-depth variable selection.
selected = forest.minimal_depth().query("selected")["feature"].tolist()

# Exact TreeSHAP attributions (Lundberg 2018, Algorithm 2).
shap, base = forest.shap_values(X[:10])               # (n, p, n_times, n_causes)

Fine-Gray, Aalen-Johansen, Gray's test, and cause-specific Cox

from comprisk import (
    FineGrayRegression, CumulativeIncidence, CauseSpecificCox, gray_test,
)

# Fine-Gray subdistribution-hazard regression — matches R cmprsk::crr()
# β̂ to floating-point noise. robust_se=True gives the Geskus cluster
# sandwich (matches cmprsk's IPCW-corrected SE to ~3 digits).
fg = FineGrayRegression(cause=1, robust_se=True).fit(X, time=time, event=event)
print(fg.coef_, fg.se_)
F = fg.predict_cumulative_incidence(X[:5])            # (5, n_event_times)

# Non-parametric Aalen-Johansen CIF (cmprsk::cuminc parity, optional groups).
ci = CumulativeIncidence().fit(time=time, event=event, group=group_var)
est, var = ci.timepoints([1.0, 5.0, 10.0])            # (n_curves, 3)

# Gray's K-sample test for CIFs — matches cmprsk::cuminc()$Tests to 1e-14.
result = gray_test(time, event, group_var, cause=1)
print(result.stat, result.pvalue, result.df)

# Cause-specific Cox PH — competing events censored at t_j.
# Matches survival::coxph(method="breslow") to 1e-9.
cs = CauseSpecificCox(cause=1).fit(X, time=time, event=event)

Detailed walkthroughs — additivity checks, global SHAP importance, sklearn- compatible slicing, performance caveats, rfSRC threshold compatibility — in docs/quickstart.md, which also covers data format, prediction shapes, cross-validation, GPU, and rfSRC migration.

scikit-learn drop-in. CompetingRiskForest is a real sklearn estimator (BaseEstimator, clone()-friendly, picklable). cross_val_score, KFold, Pipeline work without a wrapper — pass Surv.from_arrays(event, time) as the y argument, or use the legacy 3-arg fit(X, time, event) form. Full example in docs/quickstart.md § Cross-validation.

Roadmap

comprisk is intentionally CR-focused. For non-CR survival methods (general Cox PH, AFT, parametric, deep-survival, Kaplan-Meier as a standalone API), use lifelines or scikit-survival.

Version	Module	Status
v0.3	CompetingRiskForest (CR-RSF)	Shipped
v0.4	FineGrayRegression (subdistribution hazard)	Shipped
v0.4	CumulativeIncidence (stand-alone Aalen-Johansen)	Shipped
v0.4	gray_test (Gray's K-sample log-rank)	Shipped
v0.4	CauseSpecificCox (CR-aware censoring)	Shipped
v0.4	score_cr / calibration_cr (CR-aware evaluation)	Shipped
v1.0	API freeze + JMLR MLOSS submission	Planned
v1.1	Full GPU rewrite	Planned

Benchmarks

Headline numbers — full tables, methodology, and reproducibility scripts in docs/benchmarks.md.

vs randomForestSRC, matched-pair on real EHR data:

Cohort	n × p	Hardware	comprisk	rfSRC OMP-on	Speedup
CHF (cardio)	75k × 58	Apple M4 / i7-14700K / HPC	5.6–9.4 s	84.8–207.3 s	14–22×
SEER breast (oncology)	238k × 17	HPC Xeon Gold 6148	7.0 s	81.6 s	11.6×

Both libraries fit similarly well at every tested workload (HF / cancer-specific C ≈ 0.85). The 10–22× cross-dataset band tracks feature count: rfSRC's per-split exhaustive scan scales with p, so the gap narrows on lower-p cohorts. ~95× speedup vs rfSRC built without OpenMP (default R-on-macOS install).

vs scikit-survival, paired on i7-14700K — synthetic 2-cause Weibull, p = 58, both libraries at their best config:

n	sksurv low_memory=True	comprisk	speedup
5 000	18.2 s	1.10 s	16.6×
50 000	2935 s (49 min)	5.40 s	544×

The gap widens super-linearly (sksurv ≈ n^2.2; comprisk ≈ n^0.7). comprisk also provides Aalen-Johansen CIF + Nelson-Aalen CHF that sksurv low_memory=True raises NotImplementedError for.

Scaling on a consumer desktop: n = 10⁶ in 63 s on i7-14700K, 14.5 GB RSS. Reproducible via validation/spikes/lambda/exp5_paper_scale_bench.py.

API

Full parameter list in src/comprisk/forest.py; usage by task in docs/quickstart.md. Two splitrules are available: logrankCR (composite competing-risks log-rank, default) and logrank (cause-specific).

Documentation

• Quickstart — common tasks with runnable code
• PRD — what comprisk aims to be at v1.0
• Equivalence vs rfSRC — cross-library validation methodology
• References — algorithmic provenance (Park-Miller, Bays-Durham, Wolbers 2009, Uno 2011, Cole & Hernán 2008, Breiman 2001, Ishwaran 2008/2014, etc.)

Development

Requires uv.

uv venv
uv pip install -e ".[dev]"
uv run pre-commit install
uv run pytest
uv run ruff check .
uv run ruff format --check .

License

Apache-2.0. See LICENSE and NOTICE.

Citation

@software{yang_comprisk_2026,
  author    = {Yang, Sunny and Zhao, Wanqi},
  title     = {{comprisk: a Python toolkit for competing risks}},
  year      = {2026},
  publisher = {Zenodo},
  version   = {0.3.1},
  doi       = {10.5281/zenodo.19876282},
  url       = {https://doi.org/10.5281/zenodo.19876282},
}

DOI is concept-level (always resolves to the latest version). GitHub's "Cite this repository" button generates a version-specific record from CITATION.cff. Algorithmic references in docs/REFERENCES.md.