The pervasiveness of health disparities around the world continues to be a pervasive problem. Factors such as limited access to healthcare, varied clinical treatment, and inconsistencies in diagnostic capabilities feed into the difficulties in achieving health equity globally. The introduction of artificial intelligence (AI) into healthcare has the potential to tackle these challenges, but careful precautions must be in place to ensure that it does not exacerbate existing inequalities. As a result, there are increasing calls from academic, clinical, and regulatory bodies for a thorough evaluation and subsequent mitigation of potential negative impacts of AI deployment on health equity.
Health equity is based on the premise of providing all individuals with equal opportunities to attain optimal health outcomes. This definition, given by public health bodies, underlines the need for different approaches or additional efforts for those facing significant obstacles in achieving better health. Unlike fairness in AI, which leans towards equal performance across all patient groups, health equity focuses on bridging health disparity gaps.
In light of the necessity to take health equity into account while developing AI technologies for healthcare, Google Health researchers have proposed a model dubbed the Health Equity Assessment for Machine Learning Performance (HEAL) framework. The HEAL framework is essentially a quantitative means to assess whether a particular AI tool is equitable in terms of its performance. More specifically, it examines whether the AI model functions better for groups with poorer average health outcomes than others. The focus here lies in equating model performance with varying health outcomes as impacted by structural inequities, hence placing health equity considerations at the forefront of AI development processes.
To exemplify its use, the HEAL framework was applied to a dermatology AI model. This demonstrated the capacity of the framework to evaluate the considerations for health equity in AI technologies and how these technologies could influence different patient demographics. This pilot application of the HEAL framework demonstrated its potential in identifying and addressing health equity issues in AI development.
There’s an emerging emphasis on actively assessing health equity implications during the development of AI tools. The focus is on targeting efforts to redress health inequities for subpopulations that are unduly burdened by structural obstacles. While the HEAL framework may not decipher causative relationships or quantify the direct influence of novel AI technologies on reducing health outcome disparities, it’s a powerful tool for pinpointing instances where model efficiency might not align with the objective of addressing pre-existing health disparities.
In summary, the HEAL framework marks an important advancement in grappling with health equity considerations in AI technologies. Ongoing research and development are crucial to honing and broadening the application of this methodology across various healthcare domains. Its integration into AI model development processes is anticipated to curb the exacerbation of health disparities and foster more equitable health outcomes for all individuals.
References:
Paper and Blog
Credit: Google Health Researchers
Twitter, Telegram Channel, Discord Channel, and LinkedIn Group
38k+ ML SubReddit
MarkTechPost
