Health Equity and Ethical Considerations in Using Artificial Intelligence in Public Health and Medicine

Irene Dankwa-Mullan, MD, MPH¹ (View author affiliations)

Suggested citation for this article: Dankwa-Mullan I. Health Equity and Ethical Considerations in Using Artificial Intelligence in Public Health and Medicine. Prev Chronic Dis 2024;21:240245. DOI: http://dx.doi.org/10.5888/pcd21.240245.

PEER REVIEWED

Abstract

This commentary explores the critical roles of health equity and ethical considerations in the deployment of artificial intelligence (AI) in public health and medicine. As AI increasingly permeates these fields, it promises substantial benefits but also poses risks that could exacerbate existing disparities and ethical challenges. This commentary delves into the current integration of AI technologies, underscores the importance of ethical social responsibility, and discusses the implications for practice and policy. Recommendations are provided to ensure AI advancements are leveraged responsibly, promoting equitable health outcomes and adhering to rigorous ethical standards across all populations.

Top

Introduction

The integration of artificial intelligence (AI) in public health and medicine is revolutionizing how health care and public health professionals approach health care delivery, disease prediction, population health, and patient care management (1). As these technologies evolve, they offer unprecedented opportunities for expanding precision health, enhancing efficiency, and optimizing effectiveness in health services (2). However, this integration also prompts critical discussions of the ethical use of AI and the imperative to ensure health equity. This commentary explores how AI is reshaping public health and medicine, concerns about bias, ethical challenges, and the importance of incorporating an equity lens in its deployment.

AI’s potential to transform health is immense, from improving diagnostic accuracy to personalizing treatment plans and predicting disease trends (2). Yet, as we stand on the brink of this technological revolution, it is crucial to address the ethical implications and ensure that these advancements benefit all sections of society equitably. The misuse or unethical application of AI can lead to increased disparities and further exacerbate adverse outcomes for socially and economically disadvantaged populations.

This commentary not only discusses the current applications and benefits of AI but also emphasizes the critical need to maintain a balance between innovation and ethical responsibilities. The commentary explores the historical context of technological transitions in health, examines the effect of AI on health equity, and provides actionable insights and recommendations to guide practitioners, policymakers, researchers, and developers. The aim is to foster a health care environment that not only embraces technological advancements but also upholds the highest standards of equity and ethical practice.

Top

Background on AI in Public Health and Medicine

Historical perspective

The integration of technology in health care is not a novel concept. AI was initially described in the 1950s as expert computer systems that could mimic human intelligence (2). These systems were followed in the 2000s by the emergence of computer vision and machine learning (2). Even though researchers continued to explore AI technologies with the evolution of data, the rapid advancement and adoption of AI has come to represent a transformative shift in the landscape. Technological innovations such as the electronic health record (EHR) and medical imaging revolutionized medical diagnostics and patient record management (2). Today, AI builds on these foundational advancements by offering more sophisticated tools for data analysis and clinical decision-making.

Current trends

AI is now being used across various facets of public health and medicine, substantially altering how health professionals engage with their patients, communities, and health data. Two key areas where AI is making a mark are diagnostic algorithms and predictive analytics (2). For example, AI algorithms are being increasingly used to diagnose diseases from imaging scans — with higher accuracy and speed than human radiologists (3). In predictive analytics, AI can forecast outbreaks of diseases (4), hospital readmission rates (5), and a patient’s risk of developing chronic illnesses (6) by analyzing vast datasets. In this era of precision medicine, AI can help in tailoring medical treatments to individual genetic profiles, potentially improving outcomes and minimizing side effects (7). Public health surveillance, disease forecasting, and epidemic modeling are increasingly becoming important areas for integration of AI-based tools (6). These applications showcase a few of AI’s potential to enhance the efficacy and precision of public health and clinical decision-making. However, they also bring to light the need for a robust framework to manage these technologies responsibly.

Transition challenges

As the public health and health care sectors navigate their way through digital transformation, several challenges emerge. These include technology challenges, widening knowledge gaps, and overall hesitance and resistance to change. For example, integrating AI into any existing public health or health care infrastructure requires substantial technology upgrades, a robust data architecture, and staff training. Apart from providing upgrades, gaps in understanding AI technologies among health care providers can hinder their effective implementation. In addition to that, adapting to AI-driven methods requires changes in established workflows and practices, which often meet with resistance from traditional health care providers. As AI continues to evolve, the health care industry must not only keep pace with these technological changes but also anticipate future developments. Addressing these challenges head-on will be essential for leveraging AI to improve health outcomes while ensuring that such technologies are used ethically and equitably.

Top

The Importance of Promoting Health Equity and Addressing Bias in AI Applications

The potential of AI to transform public health and medicine is immense. Yet, as health professionals harness these technologies, they must also consider the implications on health equity and ethical practices. Health equity in the context of AI applications refers to the fair and just distribution of health technologies and their benefits (8). It ensures that all individuals have access to the same high-quality health care services, regardless of their socioeconomic status, race, sex or gender, ethnicity, disability status, or geographic location (8). The deployment of AI diagnostic tools for diabetic retinopathy primarily in well-resourced health care settings or among populations with insurance coverage exemplifies an unfair distribution of technology. This approach disproportionately benefits people with greater economic means and access while potentially excluding socially or economically disadvantaged populations that may have a higher prevalence of disease but lack the resources or insurance necessary to access such advanced diagnostic tools. Equity is the absence of systematic disparities in health, or in the social determinants of health, between groups with different levels of underlying social advantage such as wealth, power, privilege, and prestige (9). For AI to be truly transformative, it must not only advance health care and outcomes but do so in a way that bridges existing health disparities rather than widening them.

Top

Sources and Risk of Bias

One of the most noteworthy concerns with AI is the risk of bias in algorithms, which can inadvertently perpetuate existing health disparities. AI bias is a general concept that refers to the fact that an AI system has been designed in a way that makes the system’s decisions or use unfair (10). These AI data biases often arise from various sources, including the processes of data access, collection, acquisition, preparation, processing, development, and validation (11). Bias can also arise from the processes through which scientific evidence is generated, from lack of research diversity and from inadequate data governance. AI models are typically trained on available data, which may not adequately represent racial and ethnic minority groups or other populations that are medically underserved (11). For example, Obermeyer et al discovered that commercial algorithms, which use cost as a proxy for illness, exhibit racial bias by inadequately identifying the health needs of Black patients compared with White patients despite similar levels of chronic illnesses (12). Training data can also reflect historical biases in treatment and access to care for socially disadvantaged populations, leading AI to replicate these injustices (12). Finally, many AI tools are so-called black boxes — in which decision-making processes are not transparent — making it difficult to assess and rectify biases (13). These are some of the problems that underscore the need for meticulous oversight and corrective measures in the development and deployment of AI technologies to ensure they serve all populations equitably.

Even though addressing AI biases has primarily focused on algorithms, external sources of AI bias exist. They include experience and expertise, exclusion, environment, empathy, and evidence (14).

Experience and expertise bias

Experience and expertise bias refers to the skew introduced by the varying levels of expertise among individuals involved in developing AI systems (14). This bias can manifest in several ways including:

• Training data quality: The quality of the training data can be influenced by the expertise of those who collect, label, and input the data. Inconsistent or incorrect labeling due to lack of expertise can lead to a biased model (2,11,15).
• Algorithm development: The design and tuning of algorithms require a high level of expertise. Inadequate expertise can result in models that do not generalize well across diverse populations (2).
• Clinical implementation: Varying levels of familiarity with AI tools among health care providers can affect how these tools are implemented and interpreted, potentially leading to biased outcomes (2,14,16).

Exclusion bias

Exclusion bias occurs when certain groups are systematically left out of the data collection and analysis processes (14). This bias can result in AI systems that do not accurately represent or serve the entire population. Some examples are:

• Data missingness: When data are missing or incomplete for groups within a dataset, the AI system may not learn patterns relevant to these groups, leading to poorer performance for them compared with other groups (2,11).
• Underrepresentation: Exclusion of certain demographic groups in clinical trials or datasets can cause AI to be less effective or even harmful to these groups (2,11).
• Access to care: AI tools developed without considering socially or economically marginalized populations might not address the unique barriers these groups face in accessing health care (2,11).

Environment bias

Environment bias arises from the socio-environmental context in which data are collected and used (14). This bias can include the following:

• Social determinants of health: Factors such as income, education, and living conditions can influence health outcomes and need to be adequately represented in datasets (2,11).
• Physical environment: Geographic and environmental factors (eg, urban vs rural settings) can affect health outcomes and must be considered to avoid biased AI predictions (2,11).
• Integration of environmental factors: Ensuring that environmental variables are incorporated into AI models can help in understanding and mitigating health disparities.

Empathy bias

Empathy bias refers to the challenge of incorporating human experiences and subjective elements that are difficult to quantify into AI systems (14). This bias includes:

• Quantitative versus qualitative data: AI systems primarily rely on quantitative data, which can miss nuanced human experiences that affect health outcomes.
• Patient preferences: Empathy bias can occur when AI systems do not consider patient preferences, values, and unique circumstances, leading to recommendations that are misaligned with patient needs (14).
• Human stories: Integrating personal stories and experiences into AI models can enhance their relevance and fairness, although this factor presents a complex challenge.

Evidence bias

Evidence bias involves the processes through which scientific evidence is generated, disseminated, and translated into practice (14). This bias can affect the overall reliability and applicability of AI systems. Examples include:

• Research funding: How research is funded can introduce biases, as funding priorities may not align with the needs of all populations.
• Publication bias: There is often a bias toward publishing positive results, which can skew the evidence base that AI systems rely on.
• Translation to practice: The way evidence is translated into clinical guidelines and policies can introduce biases if it does not consider the diversity of patient populations and contexts.

AI is not a monolithic entity; rather, it comprises various interconnected technologies and data inputs of intricate stacks playing a distinct role, contributing to the overall functionality, outputs, and intelligence of the system. To enhance clarity and understanding about sources of biases, it is beneficial to conceptualize the stack of interconnected technologies and inputs (Table). Biases that occur during the development of AI tools or models were mapped to specific points in the stack, to identify their origins and implement targeted strategies to address them (Table).

To mitigate the risk of bias and promote health equity in AI, several strategic actions are recommended. These actions include collecting data from diverse population groups to ensure AI systems are well-informed and represent the variability in human health; developing AI with explainable outcomes to allow users to understand and trust decisions and ensure accountability in AI-driven processes; continuously monitoring AI systems for biased outcomes; and adjusting algorithms accordingly to ensure they remain equitable over time.

Specific proposed strategies for addressing bias follow.

Addressing experience and expertise bias

• Diverse expert teams: Assemble multidisciplinary teams with diverse expertise, including data scientists, clinicians, ethicists, and social scientists, to inform, develop, and evaluate AI systems.
• Continuous training: Provide ongoing education and training for health care providers on AI technologies to ensure they are proficient in using and interpreting AI tools.
• Standardized protocols: Develop and adhere to standardized protocols for data collection, labeling, and algorithm development to minimize variability due to different levels of expertise.

Addressing exclusion bias

• Inclusive data collection: Ensure datasets include diverse demographic groups by actively recruiting underrepresented populations in data collection efforts.
• Equity audits: Conduct regular equity audits of AI systems to identify and address any exclusion of populations.
• Accessible AI solutions: Design AI tools with accessibility in mind, ensuring that they cater to the needs of socially and economically marginalized populations and do not perpetuate existing barriers to care.

Addressing environment bias

• Integration of social determinants: Include social determinants of health (eg, income, education, housing) in AI models to provide a more holistic understanding of health outcomes.
• Geospatial analysis: Use geospatial analysis to incorporate environmental factors such as air quality, water access, and neighborhood characteristics into health data.
• Contextual adaptation: Adapt AI models to local contexts, ensuring that they account for regional variations in social and environmental factors that affect health.

Addressing empathy bias

• Incorporation of qualitative data: Combine quantitative data with qualitative insights from patient interviews, focus groups, and patient narratives to capture a full picture of health experiences.
• Patient-centered design: Engage patients in the design and development of AI systems to ensure that their preferences, values, and experiences are reflected in the models.
• Ethical review boards: Establish ethical review boards that include patient representatives to oversee the development and deployment of AI tools, ensuring they align with patient needs and ethical standards.

Addressing evidence bias

• Diversification of funding: Advocate for diverse funding sources to support research that addresses the health needs of varied populations, avoiding biases introduced by funding priorities.
• Transparent reporting: Encourage transparent reporting of all research findings, including negative results, to build a comprehensive and unbiased evidence base.
• Inclusive guidelines: Develop clinical guidelines that are inclusive and consider the diverse patient populations and contexts in which they will be applied.

These strategies illustrate that while biases in the development and deployment of AI present challenges to health equity, with careful planning and ethical consideration AI also offers substantial opportunities to enhance health care for all. By prioritizing equity in the design and implementation of AI, public health professionals and medical practitioners can use these powerful tools to not only improve health outcomes but also ensure these improvements are shared across all segments of the population.

Top

Ethical Considerations in the Use of Artificial Intelligence

Ethical frameworks

The deployment of AI in health necessitates adherence to established ethical frameworks designed to guide clinical practice and technological development (18,19). These frameworks typically emphasize principles that must be carefully considered when integrating AI into health care settings (18,19). Principles of beneficence and nonmaleficence ensure that AI technologies benefit patients and do not cause harm, whether through error, bias, or misuse (18,19). Another ethical AI principle is preserving patient autonomy by maintaining transparency and consent in AI interactions (18,19). Fairness and justice principles ensure that AI-driven tools do not create or exacerbate inequalities but rather promote equitable access to health care services (18,19).

Privacy and confidentiality

With AI’s ability to process vast amounts of personal data, safeguarding patient privacy and confidentiality becomes paramount (18,19). These safeguards involve several key concerns about data security, informed consent, and misuse of data. It is critical to implement robust security measures to protect health data against unauthorized access and breaches (18,19). In addition, for populations with limited English proficiency, it is important to make sure informed consent forms are reviewed and explained to patients or translated. In this digital age, we can consider refining consent forms and including concise language for patients on how their data will be used in AI systems to inform their care. Finally, as part of ensuring privacy and confidentiality and limiting potential misuse, we should encourage collecting only data that are necessary for a specific AI application.

Decision-making

AI’s role in clinical decision-making, public health interventions, and population health management introduces complexities in the extent of human oversight and the transparency of AI decisions. To ensure human oversight, we should establish guidelines for human oversight in AI-driven decisions, ensuring that machines augment rather than replace human judgment. To maintain trust and accountability, it is also important to develop AI systems whose actions can be understood and explained to practitioners and patients. Finally, determining how responsibilities and liabilities are shared among AI developers, health professionals, and institutions when AI is used in patient care is a complex and critical component of integrating AI into health care systems. This component involves understanding the roles and obligations of each partner to ensure patient safety, legal compliance, and ethical standards are upheld. Developers are responsible for creating accurate, reliable, and safe AI tools. Health care providers using AI tools must be adequately trained and responsible for interpreting AI outputs correctly, making final clinical decisions based on a combination of AI insights, patient values, and their professional judgment. Public health professionals must be guided by the principles of responsibility and ethics to enhance the ability to analyze data, predict health trends, and implement effective interventions to ensure the well-being of individuals and communities. Institutions need to establish policies and provide oversight to monitor AI performance, ensuring compliance with legal and ethical standards.

Community engagement

Involving diverse communities in the AI development lifecycle is essential for its ethical application in public health and medicine. This approach ensures that AI systems are developed with a comprehensive understanding of the unique needs and challenges faced by various populations. Benefits of community engagement include enhanced relevance of the AI system to address the actual needs and preferences of the population, leading to better outcomes, and an increased trust and acceptance, with likelihood of successful implementation of the AI system (20).

Top

Challenges and Opportunities

The ethical integration of AI in health care and public health presents both challenges and opportunities. For example, AI can potentially streamline workflows and enhance diagnostic accuracy, but it also raises issues such as the potential for dehumanization in care and reduced patient–provider interactions (17). By addressing these ethical considerations proactively, working partners in public health and medicine can leverage AI to improve population health and health care outcomes while maintaining a commitment to ethical practice. As we delve into the transformative potential of AI in public health and medicine, it becomes increasingly apparent that while AI offers substantial benefits for health care efficiency and effectiveness, it also introduces substantial ethical and equity challenges.

To promote health equity and ethical AI use in public health and medicine, it is recommended to develop inclusive AI policies, enhance ethical frameworks, and ensure transparency and accountability (Figure). Investing in public and professional education about AI, fostering community engagement, and integrating social determinants of health into AI models are essential. Additionally, diverse funding for research and evidence, continuous monitoring and evaluation of AI systems, and interdisciplinary collaboration are crucial strategies to ensure AI technologies are fair, equitable, and beneficial for all populations (Figure).

Figure.
Multifaceted approach for ethical and equitable implementation of artificial intelligence (AI) in public health and medicine. [A text version of this figure is available.]

To advance public health and medicine responsibly, it is also imperative that partners work collaboratively to ensure that AI technologies not only meet the highest standards of innovation but also adhere to ethical and equitable practices. By implementing these recommendations, health care and public health professionals can leverage AI to enhance health care outcomes while safeguarding against potential inequalities and ethical transgressions.

This comprehensive approach ensures that AI serves as a tool for positive change, propelling public health and medicine into a future where technology and human values are aligned to promote the well-being of all individuals.

Top

Acknowledgments

The author received no external financial support for the research, authorship or publication of this article. The author declares no potential conflicts of interest with respect to the research, authorship or publication of this article. No copyrighted material, surveys, instruments, or tools were used in the research described in this article.

Top

Author Information

Corresponding Author: Irene Dankwa-Mullan, MD, MPH, Department of Health Policy and Management, Milken Institute School of Public Health, The George Washington University, 2175 K Street NW, Washington, DC 20037 (Irene.DankwaMullan@gwu.edu).

Author Affiliations: ¹Department of Health Policy and Management, Milken Institute School of Public Health, The George Washington University, Washington, District of Columbia.

Top

References

1. Alowais SA, Alghamdi SS, Alsuhebany N, Alqahtani T, Alshaya AI, Almohareb SN, et al. . Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ. 2023;23(1):689. PubMed doi:10.1186/s12909-023-04698-z
2. Roski J, Chapman W, Heffner J, Trivedi R, Del Fiol G, Kukafka R, et al. . Chapter 3: How artificial intelligence is changing health and health care. In: Matheny M, Israni ST, Ahmed M, Whicher D, eds. Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril. The National Academies Press; 2019:65–98.
3. Rajpurkar P, Chen E, Banerjee O, Topol EJ. AI in health and medicine. Nat Med. 2022;28(1):31–38. PubMed doi:10.1038/s41591-021-01614-0
4. Zeng D, Cao Z, Neill DB. Artificial intelligence–enabled public health surveillance — from local detection to global epidemic monitoring and control. Artif Intell Med. 2021:437–453.
5. Romero-Brufau S, Wyatt KD, Boyum P, Mickelson M, Moore M, Cognetta-Rieke C. Implementation of artificial intelligence-based clinical decision support to reduce hospital readmissions at a regional hospital. Appl Clin Inform. 2020;11(4):570–577. PubMed doi:10.1055/s-0040-1715827
6. Rashid J, Batool S, Kim J, Wasif Nisar M, Hussain A, Juneja S, et al. . An augmented artificial intelligence approach for chronic diseases prediction. Front Public Health. 2022;10:860396. PubMed doi:10.3389/fpubh.2022.860396
7. Johnson KB, Wei WQ, Weeraratne D, Frisse ME, Misulis K, Rhee K, et al. . Precision medicine, AI, and the future of personalized health care. Clin Transl Sci. 2021;14(1):86–93. PubMed doi:10.1111/cts.12884
8. Gurevich E, El Hassan B, El Morr C. Equity within AI systems: what can health leaders expect? Healthc Manage Forum. 2023;36(2):119–124. PubMed doi:10.1177/08404704221125368
9. Braveman P, Gruskin S. Defining equity in health. J Epidemiol Community Health. 2003;57(4):254–258. PubMed doi:10.1136/jech.57.4.254
10. Schwartz R, Vassilev A, Greene K, Perine L, Burt A, Hall P. Special Publication 1270: Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. National Institutes of Standards and Technology. March 2022. Accessed May 29, 2024. https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.1270.pdf
11. Nazer LH, Zatarah R, Waldrip S, Ke JXC, Moukheiber M, Khanna AK, et al. . Bias in artificial intelligence algorithms and recommendations for mitigation. PLOS Digit Health. 2023;2(6):e0000278. PubMed doi:10.1371/journal.pdig.0000278
12. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447–453. PubMed doi:10.1126/science.aax2342
13. Mann H. Do all AI systems need to be explainable?. Stanford Social Innovation Review. November 15, 2023. Accessed July 10, 2024. https://ssir.org/articles/entry/do_ai_systems_need_to_be_explainable
14. Dankwa-Mullan I, Weeraratne D. Artificial intelligence and machine learning technologies in cancer care: addressing disparities, bias, and data diversity. Cancer Discov. 2022;12(6):1423–1427. PubMed doi:10.1158/2159-8290.CD-22-0373
15. Yang J, Soltan AAS, Eyre DW, Yang Y, Clifton DA. An adversarial training framework for mitigating algorithmic biases in clinical machine learning. NPJ Digit Med. 2023;6(1):55. PubMed doi:10.1038/s41746-023-00805-y
16. Chen Y, Clayton EW, Novak LL, Anders S, Malin B. Human-centered design to address biases in artificial intelligence. J Med Internet Res. 2023;25:e43251. PubMed doi:10.2196/43251
17. Ferrara E. Fairness and bias in artificial intelligence: a brief survey of sources, impacts, and mitigation strategies. Sci. 2024;6(1):3.
18. Mennella C, Maniscalco U, De Pietro G, Esposito M. Ethical and regulatory challenges of AI technologies in healthcare: a narrative review. Heliyon. 2024;10(4):e26297. PubMed doi:10.1016/j.heliyon.2024.e26297
19. Dankwa-Mullan I, Scheufele EL, Matheny M, Quintana Y, Chapman W, Jackson G, et al. . A proposed framework on integrating health equity and racial justice into the artificial intelligence development lifecycle. J Health Care Poor Underserved. 2021;32(2):300–317.
20. Rajamani G, Rodriguez Espinosa P, Rosas LG. Intersection of health informatics tools and community engagement in health-related research to reduce health inequities: scoping review. J Particip Med. 2021;13(3):e30062. PubMed doi:10.2196/30062

Top

Table

Top