Domain Generalization and Adaptation in Intensive Care with Anchor Regression
This work addresses the challenge of deploying predictive models across different hospitals in intensive care units, which is an incremental improvement with domain-specific impact.
This paper tackles the problem of predictive model performance degradation in clinical settings due to distribution shifts by applying causality-inspired domain generalization methods, including anchor regression and a novel anchor boosting extension, to a large dataset of 400,000 patients from nine ICU databases, resulting in improved out-of-distribution performance, especially for dissimilar target domains, and proposing a framework to quantify the utility of external data across different regimes.
The performance of predictive models in clinical settings often degrades when deployed in new hospitals due to distribution shifts. This paper presents a large-scale study of causality-inspired domain generalization on heterogeneous multi-center intensive care unit (ICU) data. We apply anchor regression and introduce anchor boosting, a novel, tree-based nonlinear extension, to a large dataset comprising 400,000 patients from nine distinct ICU databases. The anchor regularization consistently improves out-of-distribution performance, particularly for the most dissimilar target domains. The methods appear robust to violations of theoretical assumptions, such as anchor exogeneity. Furthermore, we propose a novel conceptual framework to quantify the utility of large external data datasets. By evaluating performance as a function of available target-domain data, we identify three regimes: (i) a domain generalization regime, where only the external model should be used, (ii) a domain adaptation regime, where refitting the external model is optimal, and (iii) a data-rich regime, where external data provides no additional value.