Navigating Ethical Frontiers of AI-Driven Clinical Decision Support Systems: Exploring Explainability and Bias

NIH RePORTER · NIH · R21 · $144,672 · view on reporter.nih.gov ↗

Abstract

PROJECT SUMMARY Note: This project is specific to bioethics research. Existing National Comprehensive Cancer Network (NCCN) guidelines to screen patients for genetic testing that are based on family history and personal risk factors remain highly inaccurate with a positive predictive value (PPV) of less than 10%. Emerging AI models based on imaging data already show great potential to improve the effectiveness of screening patients (with improved PPV and reduced missed detections). However, the bioethics of AI-driven models and their implementation in patient care remain unregulated and unexplored. This research will research the bioethical concerns arising from using black-box AI models for stratifying newly diagnosed breast cancer patients for genetic mutation and communicating test results. NCCN has established guidelines on who is eligible for genetic testing. However, due to the inconsistent testing guidelines, more than 90% of the one million women in the U.S. who are estimated to have a BRCA mutation remain undiagnosed. Universal testing has been proposed as a possible solution. However, it is not economically feasible as it may incur up to $400 billion to perform genetic testing for all U.S. women, not to mention the lack of genetic testing for low- income populations--a large portion of population-at-risk. There are already ethical challenges that surround genetic testing, and without timely intervention, AI might reinforce or even exacerbate the outcome of patients who are already vulnerable due to existing disparities in the current healthcare system. To this end, we will focus on the following specific aims: SA1: Analyze Ethical Guidelines and Concerns with AI in Genetic Screening to assess the ethical concerns of integrating AI in recommending genetic testing and communicating the results with the patients. SA2: Managing AI Ethics–Generating Explanations, Minimizing Bias by investigating how different methods of generating explanations and best practices to minimize bias towards certain racial/ethnic groups could mitigate the ethical concerns of AI. Understanding the ethical consequences of AI in genetic screening and steps to mitigating them will be crucial in enhancing the transparency and trust in emerging healthcare technologies and guiding future policies toward effectively integrating AI in medical decision-making.

Key facts

NIH application ID
11064608
Project number
3R21EB033923-01S1
Recipient
ARIZONA STATE UNIVERSITY-TEMPE CAMPUS
Principal Investigator
Ashif Iquebal
Activity code
R21
Funding institute
NIH
Fiscal year
2024
Award amount
$144,672
Award type
3
Project period
2023-08-01 → 2026-07-31