Can we share models if sharing data is not an option?

Cross-border data sharing through the lens of research ethics committee members in sub-Saharan Africa

BACKGROUND

Several factors thwart successful data sharing-ambiguous or fragmented regulatory landscapes, conflicting institutional/researcher interests and varying levels of data science-related expertise are among these. Traditional ethics oversight mechanisms and practices may not be well placed to guarantee adequate research oversight given the unique challenges presented by digital technologies and artificial intelligence (AI). Data-intensive research has raised new, contextual ethics and legal challenges that are particularly relevant in an African research setting. Yet, no empirical research has been conducted to explore these challenges.

MATERIALS AND METHODS

We explored REC members' views and experiences on data sharing by conducting 20 semi-structured interviews online between June 2022 and February 2023. Using purposive sampling and snowballing, we recruited representatives across sub-Saharan Africa (SSA). We transcribed verbatim and thematically analysed the data with Atlas.ti V22.

RESULTS

Three dominant themes were identified: (i) experiences in reviewing data sharing protocols, (ii) perceptions of data transfer tools and (iii) ethical, legal and social challenges of data sharing. Several sub-themes emerged as: (i.a) frequency of and approaches used in reviewing data sharing protocols, (i.b) practical/technical challenges, (i.c) training, (ii.a) ideal structure of data transfer tools, (ii.b) key elements of data transfer tools, (ii.c) implementation level, (ii.d) key stakeholders in developing and reviewing a data transfer agreement (DTA), (iii.a) confidentiality and anonymity, (iii.b) consent, (iii.c) regulatory frameworks, and (iii.d) stigmatisation and discrimination.

CONCLUSIONS

Our results indicated variability in REC members' perceptions, suboptimal awareness of the existence of data protection laws and a unanimously expressed need for REC member training. To promote efficient data sharing within and across SSA, guidelines that incorporate ethical, legal and social elements need to be developed in consultation with relevant stakeholders and field experts, along with the training accreditation of REC members in the review of data-intensive protocols.

Unified fair federated learning for digital healthcare

Federated learning (FL) is a promising approach for healthcare institutions to train high-quality medical models collaboratively while protecting sensitive data privacy. However, FL models encounter fairness issues at diverse levels, leading to performance disparities across different subpopulations. To address this, we propose Federated Learning with Unified Fairness Objective (FedUFO), a unified framework consolidating diverse fairness levels within FL. By leveraging distributionally robust optimization and a unified uncertainty set, it ensures consistent performance across all subpopulations and enhances the overall efficacy of FL in healthcare and other domains while maintaining accuracy levels comparable with those of existing methods. Our model was validated by applying it to four digital healthcare tasks using real-world datasets in federated settings. Our collaborative machine learning paradigm not only promotes artificial intelligence in digital healthcare but also fosters social equity by embodying fairness.

Towards a framework for interoperability and reproducibility of predictive models

The development and deployment of machine learning (ML) models for biomedical research and healthcare currently lacks standard methodologies. Although tools for model replication are numerous, without a unifying blueprint it remains difficult to scientifically reproduce predictive ML models for any number of reasons (e.g., assumptions regarding data distributions and preprocessing, unclear test metrics, etc.) and ultimately, questions around generalizability and transportability are not readily answered. To facilitate scientific reproducibility, we built upon the Predictive Model Markup Language (PMML) to capture essential information. As a key component of the PREdictive Model Index and Exchange REpository (PREMIERE) platform, we present the Automated Metadata Pipeline (AMP) for conversion of a given predictive ML model into an extended PMML file that autocompletes an ML-based checklist, assessing model elements for interoperability and reproducibility. We demonstrate this pipeline on multiple test cases with three different ML algorithms and health-related datasets, providing a foundation for future predictive model reproducibility, sharing, and comparison.