[Paper Review] AI-Driven Healthcare: A Review on Ensuring Fairness and Mitigating Bias

Sribala Vidyadhari Chinta, Zichong Wang+6 more|arXiv (Cornell University)|Jul 29, 2024

Ethics and Social Impacts of AI20 citations

TL;DR

A survey of AI in healthcare focusing on bias sources, fairness concerns, and strategies to detect and mitigate unfairness across diverse clinical domains and regulatory contexts.

ABSTRACT

Artificial intelligence (AI) is rapidly advancing in healthcare, enhancing the efficiency and effectiveness of services across various specialties, including cardiology, ophthalmology, dermatology, emergency medicine, etc. AI applications have significantly improved diagnostic accuracy, treatment personalization, and patient outcome predictions by leveraging technologies such as machine learning, neural networks, and natural language processing. However, these advancements also introduce substantial ethical and fairness challenges, particularly related to biases in data and algorithms. These biases can lead to disparities in healthcare delivery, affecting diagnostic accuracy and treatment outcomes across different demographic groups. This review paper examines the integration of AI in healthcare, highlighting critical challenges related to bias and exploring strategies for mitigation. We emphasize the necessity of diverse datasets, fairness-aware algorithms, and regulatory frameworks to ensure equitable healthcare delivery. The paper concludes with recommendations for future research, advocating for interdisciplinary approaches, transparency in AI decision-making, and the development of innovative and inclusive AI applications.

Motivation & Objective

• Motivate the study by highlighting the growth of AI in healthcare and the fairness challenges it introduces.
• Identify and classify sources of bias in data, algorithms, and deployment contexts.
• Review consequences of biased AI on diagnostics, treatment, and health equity.
• Survey bias detection and mitigation approaches and discuss ethical and regulatory considerations.
• Provide recommendations for future interdisciplinary research and transparent AI deployment.

Proposed method

• Synthesize existing literature on AI applications across cardiology, ophthalmology, dermatology, neurology, radiology, emergency care, and public health.
• Categorize bias types into data bias, algorithm bias, explicit bias, implicit bias, and selection bias with illustrative examples.
• Discuss consequences of bias including misdiagnosis, inequitable outcomes, trust erosion, legal/ethical implications, resource misallocation, and stifled innovation.
• Summarize bias detection methods such as statistical analysis and auditing tools, and propose mitigation strategies and governance considerations.

Experimental results

Research questions

• RQ1 How does bias manifest in AI systems used in healthcare?
• RQ2 What are the sources and drivers of bias in healthcare AI, and what are their potential consequences?
• RQ3 How can bias be detected and mitigated in AI healthcare applications?
• RQ4 What ethical, legal, and governance considerations are needed to promote fair AI in healthcare?

Key findings

• Bias in healthcare AI stems from data, algorithms, explicit/implicit biases, and selection effects, affecting diverse patient groups.
• Biased AI can lead to misdiagnosis, inequitable outcomes, reduced trust, legal/ethical challenges, and misallocation of resources.
• Biases have been observed in dermatology, pneumonia diagnosis, and resource allocation, underscoring the need for diverse datasets and transparency.
• Bias detection methods include statistical parity, equal opportunity, predictive equity, and auditing frameworks like DIS, PROGRESS, IAT, and dashboards.
• Mitigation strategies emphasize diverse data, fairness-aware algorithms, transparency, and governance to support equitable AI-driven care.