decision-curve-analysis

Decision Curve Analysis

Python package for Andrew Vickers' Decision Curve Analysis method to evaluate prediction models and diagnostic tests

Reproduce the results from the paper: "Decision Curve Analysis for Personalized Treatment Choice between Multiple Options. K. Chalkou, et. al. Med Decis Making . 2023 Apr;43(3):337-349. doi: 10.1177/0272989X221143058." This methodological paper describes the methods for evaluating the clinical utility of prediction model with multiple treatments.

Clinical threshold evaluation for oncology AI models

Clinical prediction model for breast fine-needle aspirates: bootstrap validation, calibration and decision curve analysis on the WDBC data

R and Python packages for decision curve analysis when the outcome is recorded with error: bias formulas, bounds and tipping points, probabilistic bias analysis, and chart review estimation and design. Includes all code and open data to reproduce the paper.

Predicting diabetes risk from BRFSS health indicators with interpretable ML, threshold optimization, and bias/subgroup analysis.

Pre-registered reanalysis of gut-microbiota-derived TMAO in 750 post-PCI patients. Tests whether TMAO adds information beyond choline, betaine and carnitine for recurrent angina, and whether whole-panel discrimination survives honest out-of-fold validation.

Statistical audit & benchmark of ML models for Chronic Kidney Disease (CKD) diagnosis: leakage audit, corrected repeated CV significance testing, explainability, and external validation (UCI #857). Paper ID 183.

Certification and figure code for a secondary analysis of nonprescription statin self-selection studies (SELECT, CREST, TACTiC): 57 asserted reproducibility checks.

MCMC Bayesian logistic regression with prior sensitivity analysis for thyroid cancer recurrence prediction

R pipeline for a prospective cohort study of baseline BMI category and incident treated hypertension in Australian men (Ten to Men linked with PBS/MBS; n=12,742, 11-year follow-up). Cox models, multiple imputation, bootstrap C-index, decision curve analysis, E-values, competing-risks sensitivity. Manuscript under revision (Chron Dis Transl Med).

Discrimination is not calibration: from-scratch ROC, calibration, and decision-curve evaluation validated against a synthetic oracle.

Drug–drug interaction alert prioritisation under clinician attention constraints. A pre-registered negative result: patient context lost to a simple severity rule by 59 percentage points at 11× the burden.

30-day hospital readmission model on 99k real encounters: leaks removed (AUC 0.805 -> 0.671 honest), calibration, decision curves and a fairness audit