Distribution Shift-Aware Clinical Prediction across Heterogeneous Hospital Populations
- Abstract
-
Clinical prediction models trained at a single institution often experience performance degradation when deployed across hospitals with different patient populations, clinical practices, and data distributions. This study presents a distribution shift-aware learning framework for robust clinical prediction across heterogeneous healthcare environments. The proposed framework combines representation alignment, temporal distribution monitoring, and uncertainty-aware prediction to identify and mitigate changes in clinical data distributions. Patient representations are learned from demographics, laboratory measurements, vital signs, and diagnostic histories, while a distribution discrepancy module quantifies shifts between training and deployment populations. An adaptive correction mechanism subsequently adjusts representations according to the estimated degree of distribution shift. Experiments simulate cross-hospital and cross-year deployment scenarios and evaluate mortality, readmission, and complication prediction. The framework achieves more stable AUROC and calibration performance than conventional empirical risk minimization and domain adaptation baselines, with the largest improvements observed under substantial demographic and laboratory distribution shifts. These findings demonstrate the importance of explicitly modeling dataset shift when developing generalizable clinical AI systems.
- References
-
[1] A. Rajkomar et al., “Scalable and accurate deep learning with electronic health records,” npj Digital Medicine, vol. 1, Art. no. 18, 2018.
[2] X. Yan, J. Du, X. Li, X. Wang, X. Sun, P. Li, and H. Zheng, “A Hierarchical Feature Fusion and Dynamic Collaboration Framework for Robust Small Target Detection,” IEEE Access, vol. 13, pp. 92953–92964, 2025.
[3] J. Wei, Y. Liu, X. Huang, X. Zhang, W. Liu, and X. Yan, “Self-Supervised Graph Neural Networks for Enhanced Feature Extraction in Heterogeneous Information Networks,” ICMLCA, pp. 272–276, 2024.
[4] J. N. Acosta, G. J. Falcone, P. Rajpurkar, and E. J. Topol, “Multimodal biomedical AI,” Nature Medicine, vol. 28, pp. 1773–1784, 2022.
[5] H. Zheng, L. Zhu, W. Cui, R. Pan, X. Yan, and Y. Xing, “Selective knowledge injection via adapter modules in large-scale language models,” ICAIDE, pp. 373–377, 2025.
[6] X. Yan, W. Wang, M. Xiao, Y. Li, and M. Gao, “Survival prediction across diverse cancer types using neural networks,” pp. 134–138, 2024.
[7] J. G. Moreno-Torres, T. Raeder, R. Alaiz-Rodríguez, N. V. Chawla, and F. Herrera, “A unifying view on dataset shift in classification,” Pattern Recognition, vol. 45, no. 1, pp. 521–530, 2012.
[8] Y. Li, W. Zhao, B. Dang, X. Yan, M. Gao, W. Wang, and M. Xiao, “Research on adverse drug reaction prediction model combining knowledge graph embedding and deep learning,” MLISE, pp. 322–329, 2024.
[9] J. Quiñonero-Candela et al., Eds., Dataset Shift in Machine Learning. MIT Press, 2009.
[10] X. Yan, Y. Jiang, W. Liu, D. Yi, and J. Wei, “Transforming Multidimensional Time Series into Interpretable Event Sequences for Advanced Data Mining,” ICHCI, pp. 126–130, 2024.
[11] M. Xiao, Y. Li, X. Yan, M. Gao, and W. Wang, “Convolutional neural network classification of cancer cytopathology images: Taking breast cancer as an example,” pp. 145–149, 2024.
[12] H. Zheng, Y. Ma, Y. Wang, G. Liu, Z. Qi, and X. Yan, “Structuring low-rank adaptation with semantic guidance for model fine-tuning,” ICECAI, pp. 731–735, 2025.
[13] J. C. Beam and I. S. Kohane, “Big data and machine learning in health care,” JAMA, vol. 319, no. 13, pp. 1317–1318, 2018.
[14] W. Wang, Y. Li, X. Yan, M. Xiao, and M. Gao, “Breast cancer image classification method based on deep transfer learning,” pp. 190–197, 2024.
[15] X. Yan, J. Du, L. Wang, Y. Liang, J. Hu, and B. Wang, “The Synergistic Role of Deep Learning and Neural Architecture Search in Advancing Artificial Intelligence,” ICEDCS, pp. 452–456, 2024.
[16] Y. Li, X. Yan, M. Xiao, W. Wang, and F. Zhang, “Investigation of Creating Accessibility Linked Data Based on Publicly Available Accessibility Datasets,” pp. 77–81, 2024.
[17] B. Shickel et al., “Deep EHR: A survey of recent advances in deep learning techniques for electronic health record analysis,” IEEE Journal of Biomedical and Health Informatics, vol. 22, no. 5, pp. 1589–1604, 2018.
[18] H. Harutyunyan et al., “Multitask learning and benchmarking with clinical time series data,” Scientific Data, vol. 6, Art. no. 96, 2019.
- Downloads
- Published
- 2026-03-01
- Issue
- Vol. 6 No. 3 (2026)
- Section
- Articles