Translation of Black-Box Clinical Prediction Models into Standalone Transparent Nomograms: Temporal External Validation in Heart Transplantation
Abstract
We convert black-box clinical prediction models for tabular data into standalone nomograms that can be audited term by term. PRiSM (Partial Responses in Structured Models) takes the shape of each effect and interaction from the source model, not merely which variables mattered, and lets the outcome select and weight them. We tested this in 50,356 heart transplant recipients, with validation in a later era than training. Nomograms from all 5 source models - a public clinical risk score, logistic regression, neural networks, random forests and extreme gradient boosting - met a prespecified noninferiority criterion for discrimination before any further simplification, and generally preserved calibration and clinical net benefit. Those from the 3 machine-learning models showed no detectable difference in discrimination from de novo generalized additive and explainable boosting models, exceeded neural additive models, and carried fewer terms than the explainable boosting model. PRiSM is released as an open-source Python package.
Related
- Prediction-Powered Smoothing and Validation for Disaggregated AI Evaluation
- Mutation-resolved ABL1 inhibitor resistance prediction under mutation-disjoint and temporal cross-study validation
- Evaluation of clinical prediction models: Decomposing performance uncertainty into model development and validation components
- Estimating Stability of Individual-Specific Predictions from AI-based prediction models using Bootstrap and Cross-validation Resampling
- Development and internal validation of prediction models for ICU-level intervention and 14-day mortality at Medical Emergency Team activation: A retrospective cohort study
BibTeX
@misc{pigot2026translation,
title = {{Translation of Black-Box Clinical Prediction Models into Standalone Transparent Nomograms: Temporal External Validation in Heart Transplantation}},
author = {Henry Pigot and Paulo J. G. Lisboa and Sandra Ortega-Martorell and Ivan Olier and Joseph Mahon and Johan Nilsson},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.07610},
eprint = {2609.07610},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.07610},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Translation of Black-Box Clinical Prediction Models into Standalone Transparent Nomograms: Temporal External Validation in Heart Transplantation
AU - Henry Pigot
AU - Paulo J. G. Lisboa
AU - Sandra Ortega-Martorell
AU - Ivan Olier
AU - Joseph Mahon
AU - Johan Nilsson
PY - 2026
DA - 2026/09/07
PB - arXiv
DO - 10.48550/arxiv.2609.07610
UR - https://arxiv.org/abs/2609.07610
AB - We convert black-box clinical prediction models for tabular data into standalone nomograms that can be audited term by term. PRiSM (Partial Responses in Structured Models) takes the shape of each effect and interaction from the source model, not merely which variables mattered, and lets the outcome select and weight them. We tested this in 50,356 heart transplant recipients, with validation in a later era than training. Nomograms from all 5 source models - a public clinical risk score, logistic regression, neural networks, random forests and extreme gradient boosting - met a prespecified noninferiority criterion for discrimination before any further simplification, and generally preserved calibration and clinical net benefit. Those from the 3 machine-learning models showed no detectable difference in discrimination from de novo generalized additive and explainable boosting models, exceeded neural additive models, and carried fewer terms than the explainable boosting model. PRiSM is released as an open-source Python package.
ER -