CATCH-HF Hospitalization Predictor

Clinical calculation support · Verify results before clinical use

Patient variables and model settings

Enter patient values and the exact coefficients from a validated model publication. No validated CATCH-HF coefficients are preloaded.

Configurable model
Enter a JSON object with all nine numeric keys. Example format only—not validated CATCH-HF coefficients: {"intercept":-3.2,"age":0.02,"nyha":0.6,"prior_hf_adm":1.1,"ntprobnp":0.0008,"egfr":-0.01,"sbp":-0.005,"hr":0.01,"hgb":-0.12} The model assumes the linear predictor is intercept + Σ(beta × variable), followed by the logistic transformation. Confirm the predictor scale, units, endpoint, and time horizon against the source publication.

Model overview

This page implements a configurable logistic equation using age, NYHA class, prior HF hospitalization, NT-proBNP, eGFR, systolic blood pressure, heart rate, and hemoglobin. It does not embed a published coefficient set, baseline risk, risk categories, or a prediction time horizon.

Calculation method

LP = intercept + Σ(βᵢ × xᵢ)

Probability = 1 / [1 + exp(−LP)]

The formula is a generic logistic transformation. Its mathematical implementation alone does not establish that this set of variables or coefficients represents a validated CATCH-HF model.

Before clinical use

  • Verify the original peer-reviewed derivation paper and external validation, including exact coefficients and any transformations or centering.
  • Confirm population, hospitalization endpoint, prediction horizon, and units match the intended use.
  • Assess calibration and discrimination in the target population; consider recalibration only with appropriate evidence.
  • Use the output only as an adjunct to clinical judgement, not as a stand-alone decision rule.

Evidence and references

Prospective development and validation of a model to predict heart failure hospitalisationPublished prediction model with its own population and model specification. It does not validate the coefficients or variable set in this configurable tool.
Read article — Heart (2014)
Machine Learning Prediction of Mortality and Hospitalization in Heart Failure With Preserved Ejection FractionAngraal S, et al. JACC: Heart Failure. 2020;8(1):12–21. DOI: 10.1016/j.jchf.2019.06.013. The study concerns models developed in the TOPCAT HFpEF trial, not the coefficients used here.
PubMed record

The supplied code and targeted search did not establish a peer-reviewed derivation/validation publication explicitly named “CATCH-HF” with this exact predictor and coefficient set. The model therefore remains unverified pending a source publication or DOI.

Clinical disclaimer: This configurable calculator is for educational and development use until the exact model specification is verified. A displayed probability must not be interpreted as validated clinical risk without confirmed coefficients, population, endpoint, time horizon, and performance. Do not use it as the sole basis for diagnosis, treatment, admission, or discharge decisions.