Predicting Partially Observed Long-Term Outcomes with Adversarial Positive-Unlabeled Domain Adaptation

Mengying Yan, Meng Xia, Wei Angel Huang, Chuan Hong, Benjamin Goldstein, Matthew M. Engelhard
Proceedings of the sixth Conference on Health, Inference, and Learning, PMLR 287:672-690, 2025.

Abstract

Predicting long-term clinical outcomes often requires large-scale training data with sufficiently long follow-up. However, in electronic health records (EHR) data, long-term labels may not be available for contemporary patient cohorts. Given the dynamic nature of clinical practice, models that rely on historical training data may not perform optimally. In this work, we frame the problem as a positive–unlabeled domain adaptation task, where we seek to adapt from a fully labeled source domain (e.g., historical data) to a partially labeled target domain (e.g., contemporary data). We propose an adversarial framework that includes three core components: (1) Overall Alignment, to match feature distributions between source and target domains; (2) Partial Alignment, to map source negatives to unlabeled target samples; and (3) Conditional Alignment, to address conditional shift using available positive labels in the target domain. We evaluate our method on a benchmark digit classification task (SVHN-MNIST), and two real-world EHR applications: prediction of one-year mortality post COVID-19, and long-term prediction of neurodevelopmental conditions (NDC) in children. In all settings, our approach consistently outperforms baseline models and, in most cases, achieves performance close to an oracle model trained with fully observed labels.

Cite this Paper

BibTeX
@InProceedings{pmlr-v287-yan25a, title = {Predicting Partially Observed Long-Term Outcomes with Adversarial Positive-Unlabeled Domain Adaptation}, author = {Yan, Mengying and Xia, Meng and Huang, Wei Angel and Hong, Chuan and Goldstein, Benjamin and Engelhard, Matthew M.}, booktitle = {Proceedings of the sixth Conference on Health, Inference, and Learning}, pages = {672--690}, year = {2025}, editor = {Xu, Xuhai Orson and Choi, Edward and Singhal, Pankhuri and Gerych, Walter and Tang, Shengpu and Agrawal, Monica and Subbaswamy, Adarsh and Sizikova, Elena and Dunn, Jessilyn and Daneshjou, Roxana and Sarker, Tasmie and McDermott, Matthew and Chen, Irene}, volume = {287}, series = {Proceedings of Machine Learning Research}, month = {25--27 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v287/main/assets/yan25a/yan25a.pdf}, url = {https://proceedings.mlr.press/v287/yan25a.html}, abstract = {Predicting long-term clinical outcomes often requires large-scale training data with sufficiently long follow-up. However, in electronic health records (EHR) data, long-term labels may not be available for contemporary patient cohorts. Given the dynamic nature of clinical practice, models that rely on historical training data may not perform optimally. In this work, we frame the problem as a positive–unlabeled domain adaptation task, where we seek to adapt from a fully labeled source domain (e.g., historical data) to a partially labeled target domain (e.g., contemporary data). We propose an adversarial framework that includes three core components: (1) Overall Alignment, to match feature distributions between source and target domains; (2) Partial Alignment, to map source negatives to unlabeled target samples; and (3) Conditional Alignment, to address conditional shift using available positive labels in the target domain. We evaluate our method on a benchmark digit classification task (SVHN-MNIST), and two real-world EHR applications: prediction of one-year mortality post COVID-19, and long-term prediction of neurodevelopmental conditions (NDC) in children. In all settings, our approach consistently outperforms baseline models and, in most cases, achieves performance close to an oracle model trained with fully observed labels.} }
Endnote
%0 Conference Paper %T Predicting Partially Observed Long-Term Outcomes with Adversarial Positive-Unlabeled Domain Adaptation %A Mengying Yan %A Meng Xia %A Wei Angel Huang %A Chuan Hong %A Benjamin Goldstein %A Matthew M. Engelhard %B Proceedings of the sixth Conference on Health, Inference, and Learning %C Proceedings of Machine Learning Research %D 2025 %E Xuhai Orson Xu %E Edward Choi %E Pankhuri Singhal %E Walter Gerych %E Shengpu Tang %E Monica Agrawal %E Adarsh Subbaswamy %E Elena Sizikova %E Jessilyn Dunn %E Roxana Daneshjou %E Tasmie Sarker %E Matthew McDermott %E Irene Chen %F pmlr-v287-yan25a %I PMLR %P 672--690 %U https://proceedings.mlr.press/v287/yan25a.html %V 287 %X Predicting long-term clinical outcomes often requires large-scale training data with sufficiently long follow-up. However, in electronic health records (EHR) data, long-term labels may not be available for contemporary patient cohorts. Given the dynamic nature of clinical practice, models that rely on historical training data may not perform optimally. In this work, we frame the problem as a positive–unlabeled domain adaptation task, where we seek to adapt from a fully labeled source domain (e.g., historical data) to a partially labeled target domain (e.g., contemporary data). We propose an adversarial framework that includes three core components: (1) Overall Alignment, to match feature distributions between source and target domains; (2) Partial Alignment, to map source negatives to unlabeled target samples; and (3) Conditional Alignment, to address conditional shift using available positive labels in the target domain. We evaluate our method on a benchmark digit classification task (SVHN-MNIST), and two real-world EHR applications: prediction of one-year mortality post COVID-19, and long-term prediction of neurodevelopmental conditions (NDC) in children. In all settings, our approach consistently outperforms baseline models and, in most cases, achieves performance close to an oracle model trained with fully observed labels.
APA
Yan, M., Xia, M., Huang, W.A., Hong, C., Goldstein, B. & Engelhard, M.M.. (2025). Predicting Partially Observed Long-Term Outcomes with Adversarial Positive-Unlabeled Domain Adaptation. Proceedings of the sixth Conference on Health, Inference, and Learning, in Proceedings of Machine Learning Research 287:672-690 Available from https://proceedings.mlr.press/v287/yan25a.html.

Related Material