JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) is increasingly used in health applications despite concerns about biased decision-making and health equity. For example, algorithms may perpetuate biases and discriminatory behaviors encoded in health data.

Objective:

In this study, we aimed to provide a comprehensive view of the health equity issues related to the use of AI and the strategies proposed to address them.

Methods:

We reviewed academic and grey literature sources to identify 18 issues and 15 strategies and document a many-to-many mapping of which strategies are proposed to address each issue.

Results:

Equity issues arising from data characteristics and model design dominate the literature, many of which are best addressed by multiple complementary strategies. The optimal set of strategies depends on the issues facing the application, which stakeholders are engaged, and the resources available.

Conclusions:

Many AI health equity efforts will use some of the four most commonly proposed strategies: improving the quantity and quality of data, evaluating the disparities introduced by an application, involving the broader community in AI development, and improving governance. In addition, some low-overhead strategies can be considered for every application, such as considering whether the use of AI should be reduced or adopting equity-focused checklists. This scoping review aids stakeholders in creating sets of strategies that both address a broad range of equity concerns and focus on the most significant issues in their application.