Introduction to secure data sharing in primary care using the federated causal learning models.

Area of Science:

• Primary healthcare research
• Data science
• Causal inference

Background:

• Primary healthcare research faces challenges with data silos and missing data.
• High technical barriers hinder effective cross-regional data analysis.

Purpose of the Study:

• To apply the federated causal learning framework to primary healthcare.
• To estimate cross-regional causal effects without sharing raw data.
• To evaluate framework performance under various missing data scenarios.

Main Methods:

• Applied federated causal learning framework in two primary healthcare case studies.
• Developed a step-by-step protocol for estimating cross-regional causal effects.
• Conducted a simulation study to assess performance with missing data.

Main Results:

• Successfully applied the framework to chronic non-communicable and infectious diseases.
• Federated model's average treatment effect (ATE) closely matched centralized model.
• Stable model achieved near-perfect coverage rates with up to 20% missing data.

Conclusions:

• Federated causal learning is effective and practical for decentralized primary healthcare data.
• This framework offers a safe solution for federated causal inference in primary care.
• The approach advances data-driven precision decision-making in primary care.