Skip to content
Open access

Privacy-Preserving Federated AI for Early Disease Risk Prediction Across Distributed U.S. Healthcare Systems

Oct 2026 · Mathematica Applicanda · Vol 15, pp. 31-51 · 0 citations · 51 references

TL;DR

A framework in which hospitals train a shared risk-prediction model without exchanging raw records is presented, which combines federated averaging, record-level differential privacy (DP), secure aggregation, and site-specific model calibration, and it is evaluated on four dimensions: predictive accuracy, privacy leakage, performance across hospital populations, and communication overhead.

Abstract

Privacy law, institutional policy, and re-identification risk make it difficult to pool hospital records for early disease risk prediction. This paper presents a framework in which hospitals train a shared risk-prediction model without exchanging raw records. It combines federated averaging, record-level differential privacy (DP) enforced at each hospital, secure aggregation, and site-specific model calibration, and it is evaluated on four dimensions: predictive accuracy, privacy leakage, performance across hospital populations, and communication overhead. Experiments use a synthetic network of ten hospitals (eight training, two held out) with heterogeneous populations in a simulated three-year type 2 diabetes risk task; the results demonstrate the method and are not findings about real patients. Federated training reached a mean hospital AUC of 0.801, compared with 0.766 for local-only models and 0.811 for centralized pooling, and raised AUC in the smallest hospital from 0.703 to 0.810. With DP, AUC was 0.791 at ε = 4 per hospital (δ = 10⁻⁵), fell to 0.768 at ε = 2, and dropped below local-only performance at ε = 1. DP degraded calibration (expected calibration error 0.082 versus 0.040 without DP), but local recalibration restored it to 0.038, including at two hospitals that never took part in training. A membership-inference attack performed no better than chance against the federated models, whereas local-only models in the smallest hospitals leaked measurably (advantage up to 13.4% against a 6.2% null). Federated traffic was 4.5% of the volume of the raw training data. The results indicate a practical budget near ε = 4 for this task.

Read PDF

Similar papers

Open access 2026

Privacy-Preserving Federated Learning for Healthcare Analytics: Integrating Differential Privacy and Secure Aggregation for Predictive Modeling

Healthcare data is often unique in that privacy laws often restrict the ability to share patient information centrately, and as a result, there's an increasing need for predictive healthcare analytics often limited by privacy regulations. We propose a federated learning based distributed system that exploits both the d...

Govinda Sahu · 0 citations
Open access 2026

An Adaptive Federated Machine Learning Framework for Privacy-Preserving Prediction of Type 2 Diabetes Mellitus across Healthcare Institutions.

There are challenges in collaboration between health care institutions in the areas of confidentiality, governance considerations and information system integration that hinder the accurate prediction of Type 2 diabetes mellitus to enable timely preventive interventions. This is further complicated by differences in th...

Olasunkanmi Joshua Olaoye, Olufemi Olukayode Folarin, Omisore Ademola Saheed · 0 citations
Sep 2026

Privacy Noise Damages Calibration in Federated Clinical Models

Hospitals cannot pool patient records, and federated learning is widely proposed as the remedy. Most evidence comes from simulated federations cut from one dataset, which understates the difficulty because every shard shares one measurement process. We instead used a cardiology cohort of 920 patients collected independ...

Abdul Ahad · 0 citations
Review Open access Oct 2026

Enhancing Privacy and Collaborative Efforts in Healthcare through Federated Machine Learning

- Healthcare institutions create data from complementary electronic health records, medical images, laboratory measurements, genomic profiles, and connected devices. While aggregating this data can enhance machine-learning models, centralized collection increases exposure risks, conflicts with institutional data sovere...

Dharamdas Kumhar, Anil Kewat · 0 citations
Nov 2026

FedGATE: A Reputation-Aware Federated Learning Framework for Robust Healthcare Fraud Detection under Heterogeneous Institutional Data

Cross-institutional healthcare fraud detection is constrained by privacy regulations that prevent medical and insurance organizations from pooling raw claim records. Federated learning (FL) enables collaborative training without exchanging raw healthcare records, but practical deployments remain vulnerable to non-ind...

Zhi-Gang Xu, Man Zhong, Hong-Mu Han et al. · 0 citations
Conference Sep 2026

Secure Explainable AI Framework for Privacy-Aware Digital Healthcare Marketing Data Analytics

Artificial Intelligence in Digital Healthcare Marketing offers great promise for personalized patient engagement, however, data privacy, model interpretability, and regulatory compliance are important challenges. In this paper, we propose an optimized secure and explainable AI framework for healthcare marketing, combin...

Urvashi Dhirajlal Gohil, Mohanraju Muppala · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.