[Paper Review] ModelChain: Decentralized Privacy-Preserving Healthcare Predictive Modeling Framework on Private Blockchain Networks

Tsung-Ting Kuo, Lucila Ohno‐Machado|arXiv (Cornell University)|Feb 6, 2018

Privacy-Preserving Technologies in Data14 references201 citations

TL;DR

The paper proposes ModelChain, a decentralized framework that uses private blockchain networks to enable privacy-preserving, cross-institutional predictive modeling without sharing raw patient data.

ABSTRACT

Cross-institutional healthcare predictive modeling can accelerate research and facilitate quality improvement initiatives, and thus is important for national healthcare delivery priorities. For example, a model that predicts risk of re-admission for a particular set of patients will be more generalizable if developed with data from multiple institutions. While privacy-protecting methods to build predictive models exist, most are based on a centralized architecture, which presents security and robustness vulnerabilities such as single-point-of-failure (and single-point-of-breach) and accidental or malicious modification of records. In this article, we describe a new framework, ModelChain, to adapt Blockchain technology for privacy-preserving machine learning. Each participating site contributes to model parameter estimation without revealing any patient health information (i.e., only model data, no observation-level data, are exchanged across institutions). We integrate privacy-preserving online machine learning with a private Blockchain network, apply transaction metadata to disseminate partial models, and design a new proof-of-information algorithm to determine the order of the online learning process. We also discuss the benefits and potential issues of applying Blockchain technology to solve the privacy-preserving healthcare predictive modeling task and to increase interoperability between institutions, to support the Nationwide Interoperability Roadmap and national healthcare delivery priorities such as Patient-Centered Outcomes Research (PCOR).

Motivation & Objective

• Motivate cross-institutional predictive modeling to improve generalizability of healthcare models.
• Address privacy, security, and robustness concerns in centralized architectures for collaborative modeling.
• Propose a decentralized approach that preserves patient privacy while enabling model parameter sharing.
• Integrate privacy-preserving online machine learning with a private blockchain and a novel information-sharing protocol.

Proposed method

• Introduce ModelChain architecture that uses a private blockchain for disseminating partial model information.
• Integrate privacy-preserving online machine learning techniques so only model data, not observations, are exchanged.
• Develop a new proof-of-information mechanism to determine the order of the online learning process.
• Leverage transaction metadata to propagate partial models across participating sites.
• Discuss benefits, limitations, and interoperability implications for national healthcare priorities.

Experimental results

Research questions

• RQ1 How can private blockchain networks be used to enable privacy-preserving, cross-institutional predictive modeling?
• RQ2 What mechanisms ensure secure, robust, and privacy-protective exchange of model information across sites?
• RQ3 How does a proof-of-information algorithm influence the order of online learning in a decentralized setting?
• RQ4 What are the interoperability and policy implications of applying blockchain-based privacy-preserving modeling in healthcare?

Key findings

• ModelChain enables collaborative predictive modeling without exposing patient-level data across institutions.
• The framework uses transaction metadata to disseminate partial models rather than raw data.
• A novel proof-of-information algorithm is introduced to regulate the online learning order.
• The approach targets enhanced interoperability and supports national healthcare priorities such as PCOR.
• The paper discusses benefits and potential issues of applying blockchain to privacy-preserving healthcare modeling.