Health equity at a crossroads: the rise of AI
The United States (US) healthcare system is beset by significant disparities, spanning race, ethnicity, gender, age, and geography.1 Tackling these disparities calls for innovative solutions. One promising approach is the growing use of Artificial Intelligence (AI) in healthcare. Economist Enterprise, supported by EMD Serono, set out to assess how AI can mitigate and/or exacerbate health disparities in the US.
Supported by
AI usage, expected impact and challenges
According to our survey of 600 US healthcare providers, executives, payers, and patient advocates:
of respondents use AI technologies several times a week
use it every day
never use it
When asked if the use of AI in health and medicine will change patient outcomes in the next five years:
In the next five years, AI is expected to have the biggest positive impact* on:
Objective data analysis
Diagnostic accuracy and reliability
Optimization of resource allocation
In the next five years, AI is expected to have the least positive impact* on:
Patient engagement and experience
Bias in clinical decision-making
Patient-provider relationship
*Note: frequency selected
This suggests that while AI has the potential to enhance clinical decision-making and healthcare efficiency, patients will continue to value the personal, emotional, and ethical aspects of patient care.
The leading uses of AI among healthcare providers and hospital administrators are:
Diagnostic assistance
Predictive analytics
Personalized medicine/ treatment plans
While AI is also increasingly used to streamline administrative tasks such as scheduling, billing, and patient data management,2 fewer than 1 in 5 of surveyed healthcare providers and executives use AI for ‘scheduling and resource management’ and ‘patient flow management’.
Trust AI...
Distrust AI...
When it comes to trusting AI to make accurate decisions:
Payers and patient advocates:
Trust AI...
Distrust AI...
Healthcare providers:
Trust AI...
Distrust AI...
Healthcare executives:
This suggests overall trust in AI for administrative and diagnostic tasks with a preference for human involvement in areas requiring emotional intelligence, as well as concerns about data privacy and security.
The relatively recent and rapid rise of AI means that its adoption will also face challenges. Our survey identified top* five perceived barriers:
Lack of uniform access to advanced technology
Inconsistent incorporation into routine clinical practice and workflow
Lack of adequate digital infrastructure
Lack of public awareness of AI applications
Data privacy and security
*Note: frequency selected
This suggests key concerns involve implementation, access, and ethics.
The role of AI in advancing health equity
If individuals aren’t well represented ... then we shouldn’t expect the decisions [regarding] individuals to always be the best ones.
Dr. Colin G Walsh, Associate Professor, Department of Biomedical Informatics, Vanderbilt University
AI has the potential to help reduce disparities in healthcare
Improve diagnostic accuracy and reduce disparities in healthcare outcomes
Provide personalized (precision) treatment
Bridge health inequities
Address language barriers
Ensure equitable health outcomes across ethnic and racial groups
Survey respondents overwhelmingly agree that AI can:
Or amplify existing inequities and create new disadvantages
AI systems are only as reliable as the data and algorithms upon which they are built. Racial and ethnic minorities are frequently under-represented in datasets, largely due to variations in access to care and healthcare-seeking behaviors.3 These data gaps can worsen health disparities and affect the generalizability of the data.4 If not carefully implemented, AI could amplify existing inequities and create new ones, particularly for vulnerable populations.
Most survey respondents are not confident in AI’s ability to reduce bias in healthcare decision-making (only 24% of respondents believe it will) or improve trust in the healthcare system (20% say it will).
The biggest risk that we have cited or we have ascribed to AI is the fact that AI is trained on historical data.
Leo Anthony Celi, Senior Research Scientist, Massachusetts Institute of Technology (MIT); Clinical Research Director, Laboratory of Computational Physiology, MIT
The top* five ethical concerns raised by respondents include:
Ensuring AI works for all—policymaking priorities
AI can transform healthcare, but it must be designed with diverse input, patient trust, and transparency to ensure fairness and equity. Collaboration among stakeholders is crucial to prevent bias, safeguard patient rights, and ensure equitable access. Our research highlights the key priorities for US policymakers:
Combat bias in data and algorithmic design
Algorithmic vigilance is crucial for ensuring AI systems are designed to prevent adverse effects and remain fair and inclusive.
Assure patients a seat at the table
The needs, concerns, perspectives, and rights of patients must be prioritized and considered at every step, from conception to implementation.
Place patients at the center of AI implementation
Regulators and policymakers must work to ensure that standards are developed to ensure AI transparency and explainability—and, ultimately, trust. Access is also key.
Scroll
Combat bias in data and algorithmic design
Algorithmic vigilance is crucial for ensuring AI systems are designed to prevent adverse effects and remain fair and inclusive.
Assure patients a seat at the table
The needs, concerns, perspectives, and rights of patients must be prioritized and considered at every step, from conception to implementation.
Place patients at the center of AI implementation
Regulators and policymakers must work to ensure that standards are developed to ensure AI transparency and explainability—and, ultimately, trust. Access is also key.
Scroll
Anticipate and manage unintended or negative consequences
To mitigate the unintended consequences of AI use, collaboration between all stakeholders is required.
Establish regulation and oversight suited to the new challenges created by AI
Policymakers and regulators must work with developers, healthcare providers, insurers, and patient groups to ensure that patients’ needs and rights are safeguarded and protected.
Implement a fair, impact-based pricing model
A pricing model is needed that ensures both value and equitable access, without creating additional cost-related barriers for patients.
Implementing a transparent, ethical, equitable system operated by trusted partners
All stakeholders must work together to ensure that AI solutions are developed and applied responsibly and ethically, emphasizing transparency and equity.
It starts with patients. For a very long time, we have ignored patients… We have created solutions for patients without [their] involvement... As we build new tools, it’s critically important—no matter who the stakeholder is—if they think about a ‘trusted partner’, they should think about patients.
Smit Patel, Director of Digital Health & AI, Digital Medicine Society
Scroll down
Anticipate and manage unintended or negative consequences
To mitigate the unintended consequences of AI use, collaboration between all stakeholders is required.
Establish regulation and oversight suited to the new challenges created by AI
Policymakers and regulators must work with developers, healthcare providers, insurers, and patient groups to ensure that patients’ needs and rights are safeguarded and protected.
Implement a fair, impact-based pricing model
A pricing model is needed that ensures both value and equitable access, without creating additional cost-related barriers for patients.
Scroll
Scroll
Implementing a transparent, ethical, equitable system operated by trusted partners
All stakeholders must work together to ensure that AI solutions are developed and applied responsibly and ethically, emphasizing transparency and equity.
It starts with patients. For a very long time, we have ignored patients… We have created solutions for patients without [their] involvement... As we build new tools, it’s critically important—no matter who the stakeholder is—if they think about a ‘trusted partner’, they should think about patients.
Smit Patel, Director of Digital Health & AI, Digital Medicine Society
Scroll down
Combat bias in data and algorithmic design
Algorithmic vigilance is crucial for ensuring AI systems are designed to prevent adverse effects and remain fair and inclusive.
AI can transform healthcare, but it must be designed with diverse input, patient trust, and transparency to ensure fairness and equity. Collaboration among stakeholders is crucial to prevent bias, safeguard patient rights, and ensure equitable access. Our research highlights the key priorities for US policymakers:
Assure patients a seat at the table
The needs, concerns, perspectives, and rights of patients must be prioritized and considered at every step, from conception to implementation.
Scroll
Place patients at the center of AI implementation
Regulators and policymakers must work to ensure that standards are developed to ensure AI transparency and explainability—and, ultimately, trust. Access is also key.
Anticipate and manage unintended or negative consequences
To mitigate the unintended consequences of AI use, collaboration between all stakeholders is required.
Scroll
Establish regulation and oversight suited to the new challenges created by AI
Policymakers and regulators must work with developers, healthcare providers, insurers, and patient groups to ensure that patients’ needs and rights are safeguarded and protected.
Implement a fair, impact-based pricing model
A pricing model is needed that ensures both value and equitable access, without creating additional cost-related barriers for patients.
Scroll
Implementing a transparent, ethical, equitable system operated by trusted partners
All stakeholders must work together to ensure that AI solutions are developed and applied responsibly and ethically, emphasizing transparency and equity.
It starts with patients. For a very long time, we have ignored patients… We have created solutions for patients without [their] involvement... As we build new tools, it’s critically important—no matter who the stakeholder is—if they think about a ‘trusted partner’, they should think about patients.
Smit Patel, Director of Digital Health & AI, Digital Medicine Society
Scroll down
Combat bias in data and algorithmic design
Algorithmic vigilance is crucial for ensuring AI systems are designed to prevent adverse effects and remain fair and inclusive.
Assure patients a seat at the table
The needs, concerns, perspectives, and rights of patients must be prioritized and considered at every step, from conception to implementation.
AI can transform healthcare, but it must be designed with diverse input, patient trust, and transparency to ensure fairness and equity. Collaboration among stakeholders is crucial to prevent bias, safeguard patient rights, and ensure equitable access. Our research highlights the key priorities for US policymakers:
Place patients at the center of AI implementation
Regulators and policymakers must work to ensure that standards are developed to ensure AI transparency and explainability—and, ultimately, trust. Access is also key.
Anticipate and manage unintended or negative consequences
To mitigate the unintended consequences of AI use, collaboration between all stakeholders is required.
Scroll
Establish regulation and oversight suited to the new challenges created by AI
Policymakers and regulators must work with developers, healthcare providers, insurers, and patient groups to ensure that patients’ needs and rights are safeguarded and protected.
Implement a fair, impact-based pricing model
A pricing model is needed that ensures both value and equitable access, without creating additional cost-related barriers for patients.
Scroll
Implementing a transparent, ethical, equitable system operated by trusted partners
All stakeholders must work together to ensure that AI solutions are developed and applied responsibly and ethically, emphasizing transparency and equity.
It starts with patients. For a very long time, we have ignored patients… We have created solutions for patients without [their] involvement... As we build new tools, it’s critically important—no matter who the stakeholder is—if they think about a ‘trusted partner’, they should think about patients.
Smit Patel, Director of Digital Health & AI, Digital Medicine Society
Scroll down
References
- Nduga N, Hill L, Artiga S. Key data on health and health care by race and ethnicity. [Internet]. San Francisco [CA]:KFF;[updated 11 June 2024]. Available from: https://www.kff.org/key-data-on-health-and-health-care-by-race-and-ethnicity/?entry=racial-diversity-within-the-u-s-today-total-population-by-race-and-ethnicity.
- Spear J, Ehrenfeld JM, Miller BJ. Applications of Artificial Intelligence in Health Care Delivery. J Med Syst. 2023;47(1):121. doi: 10.1007/s10916-023-02018-y.
- Tipton K, Leas BF, Flores E, Jepson C, Aysola J, Cohen J, Harhay M, Schmidt H, Weissman G, Treadwell J, Mull NK, Siddique SM. Impact of healthcare algorithms on racial and ethnic disparities in health and healthcare. [Internet]. Rockville [MD]:Agency for Healthcare Research and Quality (US);2023. Report No.: 24-EHC004.
- Dulaney A, Virostko J. Disparities in the demographic composition of the cancer imaging archive. Radiol Imaging Cancer. 2024;6(1):e230100. doi: 10.1148/rycan.230100.
©2025 The Economist Group. All rights reserved. The Economist Group endeavors to provide accurate information; however, neither The Economist Group nor its affiliates can accept any responsibility or liability for reliance by any person on this information. Asset commissioned by EMD Serono. EMD Serono is the Healthcare business of Merck KGaA, Darmstadt, Germany in the U.S. and Canada. US-NONNI-02187 03/25.
