Artificial intelligence (AI) is rapidly transforming health care, with its potential to improve efficiency, reduce provider burnout, and enhance patient outcomes. However, as Rebecca G. Mishuris, MD, MS, MPH, chief medical information officer and VP at Mass General Brigham and faculty member of the Safety, Quality, Informatics, and Leadership program with Harvard Medical School emphasizes, these innovations also bring challenges, particularly in ensuring health equity. “The role of equity in AI, and the role of AI in equity, has to be top of mind for everyone,” Mishuris explains. “From developers to health care systems administrators, we all have a responsibility to ensure these tools serve everyone equitably.”
AI in health care encompasses a range of tools, from long-established analytic AI to the newer, more versatile generative AI. Analytic AI has been a staple for decades, aiding in predictive analytics, interpreting diagnostic tests like electrocardiograms (EKGs), and guiding screening protocols. Generative AI, driven by large language models like ChatGPT, has recently garnered attention for its ability to create human-like text and interpret complex interactions.
“Generative AI is the new kid on the block,” Mishuris says. “It uses tools like large language models to analyze and generate novel content, which is causing renewed focus on AI in health care.” Currently, generative AI is being piloted in administrative and clinical documentation tasks, though its full integration into diagnostic and therapeutic decision-making is still on the horizon.
Equity Challenges in AI Adoption
The adoption of AI in health care raises significant equity concerns across three key dimensions:
- Data Representation
AI models are only as good as the data they are trained on. If the data does not represent diverse populations, the models may generate biased outputs. Mishuris likens this to issues seen in randomized controlled trials: “If you conduct a trial on 60-year-old white men, the results may not apply to 25-year-old Black women. The same is true with AI, but on a much larger scale.”
- Model Bias
Many AI systems rely on historical data, which often reflects existing biases in health care. Without a deliberate effort to adjust for these biases, models can perpetuate or even exacerbate disparities. “If the history is biased, the predictions or new content will continue to be biased if we’re not careful,” Mishuris warns. “But if we actively address these biases, we can create outputs that are more equitable.”
- Access and Availability
The high cost of AI tools creates disparities between well-funded health care systems and those serving under-resourced communities. Mishuris points out, “Many health care systems across the country cannot afford to pilot or implement generative AI, which leaves out providers and patients from potential benefits like reduced provider burnout and improved patient experiences.”
Success Stories in AI Implementation
At Mass General Brigham, efforts to address equity in AI usage have informed every stage of adoption. For instance, the rollout of ambient documentation tools—technology that uses AI to draft clinical notes from patient-provider conversations—has been guided by a focus on inclusivity.
“We evaluated vendors with providers who had different accents and in varied clinical settings, from quiet clinics to noisy emergency rooms,” Mishuris shares. “We also prioritized tools that could handle multilingual interactions, ensuring they worked whether the conversation was in English, another language, or both.”
The impact of this technology has been profound. In the first six weeks of its use with the initial 220 physicians and advanced-practice providers, Mass General Brigham reported a 40% relative reduction in burnout among participating providers, with 60% saying they were likely to extend their clinical careers. Additionally, almost 80% noted that they were able to pay more attention to patients during visits. “It’s a rare case where technology actually improves the patient-provider interaction,” Mishuris observes. “It lets us focus more on the human connection.”
The Path Forward: Responsible Use and Monitoring
While generative AI has already demonstrated its potential, its future in clinical decision-making hinges on overcoming challenges in evaluation and monitoring. AI systems evolve over time, which can lead to changes in their outputs. “We need robust monitoring systems to ensure these tools do not shift in ways we didn’t anticipate,” Mishuris says.
The global health care community must also consider ethical and equitable deployment. Mishuris envisions a future where AI can actively reduce health care disparities—but only if equity is prioritized from the outset. “I see a world where generative AI helps us deliver more equitable care and improve outcomes for all patients,” she states. “But that will only happen if we pay attention to equity as a North Star for these technologies.”
Empowering Clinicians to Advocate for Equity
Beyond organizational efforts, individual health care providers have a role to play in promoting equitable AI use. Mishuris stresses the importance of awareness and advocacy. “Frontline staff need to understand where these systems are appropriate and what their limitations are,” she says. “They should also ask critical questions such as: Have we considered equity in deploying and monitoring these systems?”
This collective responsibility—spanning developers, administrators, and clinicians—can ensure that AI becomes a force for equity in health care.
The intersection of AI and health equity presents both a challenge and an opportunity. While the potential for bias and unequal access exists, deliberate and thoughtful implementation of these technologies can help overcome systemic disparities. As Mishuris puts it, “We must be cognizant of how we use AI to improve equity. If we do this right, AI can help us build a more just and effective health care system.”
This commitment to responsible AI adoption represents a pivotal step in shaping the future of health care quality and safety. As AI continues to evolve, its role in promoting health equity will remain crucial for all stakeholders.
