Bootstrapping-based Regularisation for Reducing Individual Prediction Instability in Clinical Risk Prediction Models

Dimensions overall score 8.0

• The proposed bootstrapping-based regularisation framework embeds bootstrapping directly into deep neural network training to constrain prediction variability across resampled datasets.method 95%
  Evidencepartial

  we propose a novel bootstrapping-based regularisation framework that embeds the bootstrapping process directly into the training of deep neural networks

  Implicationpartial

  Directly stated in abstract as the core methodological contribution

  Verificationpartial

  partial

• Models using the proposed regularisation approach exhibit improved prediction stability compared to conventional models across all evaluated datasets.result 92%
  Evidencepartial

  Across all datasets, our model exhibited improved prediction stability, with lower mean absolute differences (e.g., 0.019 vs. 0.059 in GUSTO-I; 0.057 vs. 0.088 in Framingham)

  Implicationpartial

  Explicitly stated with specific numeric evidence from multiple datasets

  Verificationpartial

  partial

• The proposed approach produces a single model with inherent stability properties rather than requiring ensemble methods.technical 88%
  Evidencepartial

  This approach constrains prediction variability across resampled datasets, producing a single model with inherent stability properties

  Implicationpartial

  Directly stated in abstract as a key advantage of the method

  Verificationpartial

  partial

• The proposed method maintains discriminative performance and feature importance consistency while improving stability.result 90%
  Evidencepartial

  discriminative performance and feature importance consistency were maintained, with high SHAP correlations between models (e.g., 0.894 for GUSTO-I; 0.965 for Framingham)

  Implicationpartial

  Explicitly stated with supporting evidence of high SHAP correlations

  Verificationpartial

  partial

• Ensemble models achieve greater stability but at the expense of interpretability due to inconsistent predictor usage across constituent models.limitation 85%
  Evidencepartial

  While ensemble models achieved greater stability, we show that this came at the expense of interpretability, as each constituent model used predictors in different ways

  Implicationpartial

  Directly stated as a limitation of ensemble approaches

  Verificationpartial

  partial

• The method provides a practical route toward more reliable and clinically trustworthy deep learning models, particularly valuable in data-limited healthcare settings.market 80%
  Evidencepartial

  This method provides a practical route toward more reliable and clinically trustworthy deep learning models, particularly valuable in data-limited healthcare settings

  Implicationpartial

  Directly stated as a conclusion but represents an application claim rather than a demonstrated result

  Verificationpartial

  partial

• The proposed approach reduces the number of significantly deviating predictions compared to conventional models.result 87%
  Evidencepartial

  markedly fewer significantly deviating predictions

  Implicationpartial

  Explicitly stated in abstract with qualitative evidence

  Verificationpartial

  partial

• The method achieves greater robustness and reproducibility without sacrificing interpretability.technical 83%
  Evidencepartial

  our approach allows prediction models to be developed that achieve greater robustness and reproducibility without sacrificing interpretability

  Implicationpartial

  Directly stated as a conclusion but represents a comparative claim that requires inference from the full evaluation

  Verificationpartial

  partial

Competitive landscape

Develop a regularisation framework for stabilizing clinical prediction models' outputs, enhancing reliability and interpretability in healthcare.