Researchers have developed a new approach to handling missing data in healthcare federated learning called FedCoRe. This method, known as Federated Cross-Modal Representation Completion, learns to correct or complete missing data rather than generating synthetic data. In experiments on a respiratory deterioration task using simulated clients, the authors found that their approach can recover up to 52.8% of lost performance when certain modalities are missing. However, they emphasize that FedCoRe should only be deployed in settings where paired examples and validation evidence support its use. --- Why it matters: This matters because healthcare federated learning often relies on data from multiple sources with different availability of modalities such as ECGs or chest radiographs. The ability to adaptively complete missing data can improve the accuracy and reliability of predictions, but requires careful consideration of when and how to deploy these methods. Source: https://arxiv.org/abs/2608.18311