A Genetic Algorithm-Optimized Federated Averaging Framework for Adaptive Client Selection in Diabetic Readmission Prediction Across Distributed Hospital Data
Keywords:
diabetes mellitus, federated learning, GA-FedAvg, genetic algorithm, privacy-preserving machine learning, electronic health records, clinical risk prediction, non-IID data.Abstract
The unplanned re-hospitalization of adult patients with diabetes mellitus within 30-days is an important clinical and operational performance issue. Models built from centralised dataset can identify patients at risk of readmission within 30 days. Nonetheless, challenges to cooperation between hospitals and the use of shared datasets revolve around privacy, governance, and patient-data-sharing restrictions. This paper proposes a GA FedAvg framework, which is a Genetic Algorithm-enabled Federated Averaging framework, for predicting unplanned 30-day hospital readmission among adult patients with diabetes mellitus in various hospitals. They share model updates instead of patient-level data. Expert researchers utilized a genetic algorithm in order to enhance federated aggregation and client-selection procedures.Then, they utilized this research in order to tackle cross-hospital heterogeneity as well as non-independently and identically distributed (non-IID) data. The model included routinely available demographics, clinical, hospitalization, laboratory, comorbidity and discharge variables. The area under the receiver operating characteristic curve (AUC-ROC) was used in evaluating the performance of GA-FedAvg and baselines used in centralized and conventional federated learning. In centralized mode, an AUC-ROC of 0.6567 was achieved whereas across federated settings, AUC-ROC values ranging from 0.6514 to 0.6547 were achieved with GA-FedAvg. The proposed method achieves around 99.2 99.7% of the performance of a centralized model, without transferring patient-level data among hospitals. The GA-FedAvg approach could enable collaboration among institutions for privacy-preserving prediction of diabetic patient readmissions, with only a slight drop in discrimative performance compared to centralized training. This technique might help identify diabetic individuals at an early stage who might be at risk. Thus, allowing for targeted discharge planning and enhancing post-discharge care.
Downloads
Downloads
Published
Issue
Section
License
You are free to:
- Share — copy and redistribute the material in any medium or format for any purpose, even commercially.
- Adapt — remix, transform, and build upon the material for any purpose, even commercially.
- The licensor cannot revoke these freedoms as long as you follow the license terms.
Under the following terms:
- Attribution — You must give appropriate credit , provide a link to the license, and indicate if changes were made . You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
- No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.
Notices:
You do not have to comply with the license for elements of the material in the public domain or where your use is permitted by an applicable exception or limitation .
No warranties are given. The license may not give you all of the permissions necessary for your intended use. For example, other rights such as publicity, privacy, or moral rights may limit how you use the material.

