From Code to Care: The AI Healthcare Evolution

Authors:

• Khan Siddiqui, MD , Founder and CEO of HOPPR
• Dr Rowland Illing , Chief Medical Officer and Director for Global Healthcare and Nonprofits at Amazon Web Services (AWS)
• Dr. Sandra Elizabeth Ford , Board-Certified Pediatrician and Former Special Assistant to the President for Public Health and Science
• Melek Somai MD MPH Somai, Vice President and Chief Technology and Product Officer at Inception Health

Here is recording of the session:

Abstract

Artificial intelligence (AI) is rapidly transforming healthcare, offering unprecedented opportunities to enhance patient care, improve operational efficiency, and address longstanding challenges such as health disparities and workforce shortages. This article, co-authored by experts in medicine, technology, and health policy, explores the current landscape of AI in healthcare, delves into the challenges of data bias and governance, and envisions an optimistic future fueled by technological advancements and collaborative efforts. By examining clinical applications, operational efficiencies, regulatory considerations, and the critical importance of diversity and ethics, we aim to provide a comprehensive overview that is both educational and actionable for healthcare professionals.

Introduction

The integration of artificial intelligence (AI) into healthcare represents a paradigm shift with the potential to revolutionize patient care. From diagnostic imaging to administrative workflows, AI is making significant inroads, promising to enhance efficiency, reduce clinician burnout, and improve patient outcomes.

At the Health Connect South 2024 conference in Atlanta, GA, we participated in a panel discussion titled "From Code to Care: The AI Healthcare Evolution." Our diverse backgrounds—in radiology, technology, pediatrics, and health policy—provided a multifaceted perspective on the opportunities and challenges that AI presents in the healthcare sector. This article expands upon that discussion, aiming to educate and engage healthcare professionals on the critical aspects of AI integration.

The Current Landscape: AI in Practice

Clinical Applications

AI's impact on clinical practice is perhaps most evident in diagnostic imaging. Advanced algorithms can analyze medical images with remarkable speed and accuracy, assisting radiologists in detecting conditions ranging from fractures to malignancies.

As a radiologist with an academic career in computer vision, machine learning, and AI, Dr. Khan Siddiqui has observed the transformative potential of these technologies. AI models can process vast amounts of imaging data, identifying patterns and anomalies that may be imperceptible to the human eye. For example, deep learning algorithms have demonstrated dermatologist-level classification of skin cancer, offering the potential for early detection and treatment (Esteva et al., 2017).

However, the deployment of AI in clinical settings is not without challenges. The "black box" nature of some AI models raises concerns about transparency and trust (Topol et al., 2019). Explainable AI is becoming increasingly important to ensure that clinicians understand how algorithms arrive at their conclusions, thereby facilitating better clinical decision-making and fostering trust among patients and providers (Yang, et al. 2022).

Operational Efficiency

Beyond direct patient care, AI is streamlining administrative tasks that often burden healthcare professionals (Maleki et al., 2024). Dr. Melek Somai, Vice President and Chief Technology and Product Officer at Inception Health, and assistant professor of clinical and biomedical informatics at the Medical College of Wisconsin, highlighted the use of generative AI in ambient listening, clinical summarization, and inbox management for providers.

For instance, ambient clinical intelligence systems can transcribe and organize patient-provider conversations in real-time, reducing the time clinicians spend on documentation. This not only enhances productivity but also allows clinicians to focus more on patient interaction, improving the overall care experience (Yim et al., 2023 and Owens et al., 2024).

AI is also optimizing scheduling, supply chain management, and predictive analytics for resource allocation. By automating routine tasks, healthcare organizations can reduce operational costs and improve efficiency, which is particularly crucial given the growing patient population and limited healthcare resources.

However, while AI’s implementation promises improvements in efficiency and productivity, early evidence suggests the reality may be more complex (Tai-Seale et al.). Many studies have shown minimal immediate impact on these key metrics. In fact, when considering the initial investment and the significant workflow disruption AI can introduce, the net effect might even seem negative.

Yet, as with any transformative innovation, the true value of AI lies not in short-term returns but in its long-term potential. These early applications should be viewed less as productivity solutions and more as socio-technical experiments that lay the groundwork for more disruptive innovations in the future. The focus for organizations today is on building the foundation—structuring data, aligning workflows, and creating a culture that can fully harness AI’s future capabilities. This foundation is crucial for unlocking the full, transformative power of AI in healthcare.

Challenges: Bias and Governance

Data Bias and Health Equity

While AI offers significant benefits, it also poses challenges, particularly concerning data bias and its impact on health equity. Dr. Sandra Elizabeth Ford, a board-certified pediatrician and former Special Assistant to the President for Public Health and Science, emphasized the critical issue of bias in AI algorithms.

Many AI models are trained on datasets that lack diversity, reflecting historical underrepresentation of certain populations in both scholarly publications and clinical trials.? This can lead to algorithms that perform well for some groups but poorly for others, inadvertently perpetuating health disparities.

For example, Dr. Ford discussed a pediatric urinary tract infection (UTI) risk prediction algorithm that included race as a factor. The original model assigned different points based on race, which could lead to underdiagnosis in certain racial groups (Shaikh et al., 2022). This underscores the need to critically assess and update algorithms to avoid racial biases (Obermeyer Z et al., 2019)

Social determinants of health (SDOH)—factors like socioeconomic status, education, and living conditions—also play a significant role in patient outcomes but are frequently omitted from AI models. Ignoring SDOH can result in algorithms that do not account for the full context of a patient's health, reducing the effectiveness of interventions and widening existing gaps in care.

Dr. Siddiqui echoed these concerns by highlighting how AI models can inadvertently learn and reinforce biases present in the data. For instance, AI algorithms have been shown to predict patient self-reported race from medical images with high accuracy, even when race is not explicitly labeled (Gichoya et al., 2022). The mechanisms enabling this prediction are not well understood, raising questions about how these hidden biases might affect diagnostic outcomes.

Regulatory Frameworks and Governance

The responsible implementation of AI in healthcare necessitates robust regulatory frameworks and governance models. Dr. Rowland Illing, Chief Medical Officer and Director for Global Healthcare and Nonprofits at AWS, discussed the importance of risk-based approaches and international standards.

The U.S. Food and Drug Administration (FDA) has begun to address AI in medical devices, but the rapid evolution of AI technologies poses challenges for traditional regulatory models. Dr. Illing highlighted a recent article in Harvard Business Review that suggested #generativeAI might be better regulated in a similar way to medical professionals rather than devices (Blumenthal et al., 2024). Just as physicians undergo continuous education and revalidation, AI models—particularly those that learn and adapt over time—may require ongoing oversight to ensure safety and efficacy.

International standards, such as the ISO/IEC 42001:2023, 23894:2023 and 23053:2022 set international standards for AI Management systems, guidance on risk management and frameworks for Artificial Intelligence (AI) Systems Using Machine Learning (ML). Given that AI is truly global, it is essential that broad consensus can be reached on standards, without unnecessary duplication of existing regulations. These standards aim to address issues of transparency, accountability, and ethical considerations on a global scale, but need to be broadly adopted. Moreover, regulatory bodies are beginning to recognize the need for guidelines that specifically address data diversity and bias mitigation. Ensuring that AI models are trained and validated on diverse populations is becoming a regulatory priority to promote health equity.

The Future: Optimism and Collaboration

Technological Advancements

Advancements in computational power and AI algorithms are accelerating at an unprecedented rate. Dr. Siddiqui shared insights into emerging technologies such as new AI architectures and the next generation of GPUs, which are significantly more efficient than their predecessors.

These technological leaps enable the development of more complex and capable AI models. For instance, increased computational capacity allows for the training of models on larger and more diverse datasets, improving their generalizability and performance across different patient populations.

Furthermore, advancements in natural language processing (NLP) and generative AI are opening new avenues for patient engagement and education. AI-driven chatbots and virtual assistants can provide patients with personalized health information, appointment scheduling, and medication reminders, enhancing patient autonomy and adherence to treatment plans.

Collaboration Across Sectors

The successful integration of AI into healthcare requires a collaborative approach involving healthcare providers, technologists, policymakers, and patients.

Dr. Somai emphasized that AI cannot be developed in isolation. Healthcare organizations must work closely with technology companies to ensure that AI solutions meet clinical needs and are user-friendly. Policymakers play a crucial role in creating supportive regulatory environments that encourage innovation while safeguarding patient interests.

Moreover, patient involvement is essential. Engaging patients in the development and deployment of AI tools helps ensure that these technologies address real-world needs and are accessible to diverse populations.

Dr. Ford highlighted the importance of diversity in AI programming and training programs. Encouraging participation from underrepresented groups in AI development, as well as in clinical trials, can help mitigate biases and promote health equity (Lawrence, 2023)

Education and Training

Investing in education is critical for preparing the healthcare workforce to effectively utilize AI technologies. Clinicians need training not only in operating AI tools but also in understanding their limitations, ethical considerations, and implications for patient care.

Dr. Illing discussed how generative AI is being used to create training scenarios and assessment tools for healthcare education (Misra, et al., 2024). AI can simulate complex clinical cases, allowing trainees to practice decision-making in a risk-free environment. Additionally, AI-driven analytics can provide personalized feedback, enhancing the learning experience.?

Educational institutions and professional organizations that incorporate AI competencies into curricula and continuing education programs. This will equip clinicians with the skills needed to critically appraise AI tools and integrate them safely and securely into clinical practice.

Conclusion: Embracing the AI Evolution

The evolution from code to care is well underway, marking an exciting era in healthcare. AI has the potential to address some of the most pressing challenges in healthcare delivery, from reducing clinician burnout to improving patient outcomes and promoting health equity.

However, realizing this potential requires a concerted effort to address challenges related to data bias, governance, and education. By fostering collaboration across sectors and prioritizing responsible innovation, we can harness AI to transform healthcare for the better.

We encourage our fellow clinicians, technologists, and policymakers to actively engage with AI technologies. Together, we can shape a future where AI is an invaluable tool in our mission to provide the best possible care to our patients.

References

1. Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115-118. https://doi.org/10.1038/nature21056.
2. Topol, E. J. (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine, 25(1), 44-56. https://doi.org/10.1038/s41591-018-0300-7.
3. Yang, Xi, Aokun Chen, Nima PourNejatian, Hoo Chang Shin, Kaleb E. Smith, Christopher Parisien, Colin Compas, et al. 2022. “A Large Language Model for Electronic Health Records.” Npj Digital Medicine 5 (1): 194. https://doi.org/10.1038/s41746-022-00742-2.?
4. Maleki VS, Forouzanfar M. The Role of AI in Hospitals and Clinics: Transforming Healthcare in the 21st Century. Bioengineering (Basel). 2024 Mar 29;11(4):337.? https://doi.org/10.3390/bioengineering11040337.
5. Owens LM, Wilda JJ, Hahn PY, Koehler T, Fletcher JJ. The association between use of ambient voice technology documentation during primary care patient encounters, documentation burden, and provider burnout. Fam Pract. 2024 Apr 15;41(2):86-91. https://doi.org/10.1093/fampra/cmad092.?
6. Tai-Seale M, Baxter SL, Vaida F, et al. AI-Generated Draft Replies Integrated Into Health Records and Physicians’ Electronic Communication. JAMA Netw Open. 2024;7(4):e246565. https://doi.org/10.1001/jamanetworkopen.2024.6565.?
7. Yim WW, Fu Y, Ben Abacha A, Snider N, Lin T, Yetisgen M. Aci-bench: a Novel Ambient Clinical Intelligence Dataset for Benchmarking Automatic Visit Note Generation. Sci Data. 2023 Sep 6;10(1):586. https://doi.org/10.1038/s41597-023-02487-3.
8. Shaikh N, Lee MC, Stokes LR, Miller E, Kurs-Lasky M, Conway I, Shope TR, Hoberman A. Reassessment of the Role of Race in Calculating the Risk for Urinary Tract Infection: A Systematic Review and Meta-analysis. JAMA Pediatr. 2022 Jun 1;176(6):569-575. https://doi.org/10.1001/jamapediatrics.2022.0700.
9. Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447-453. https://doi.org/10.1126/science.aax2342.
10. Gichoya, J. W., Banerjee, I., Bhimireddy, et al. (2022). AI recognition of patient race in medical imaging: a modeling study. The Lancet Digital Health, 4(6), e406-e414. https://pubmed.ncbi.nlm.nih.gov/35568690/.
11. Blumenthal D., Patel B. (2024). How to Regulate Generative AI in Healthcare. HBR. 2024. https://hbr.org/2024/09/how-to-regulate-generative-ai-in-healthcare.??
12. ISO/IEC TR 24028:2020. (2020). Information technology — Artificial intelligence — Overview of trustworthiness in artificial intelligence. International Organization for Standardization. Retrieved from https://www.iso.org/standard/77608.html
13. Lawrence, C.D. (2023) Hidden in White Sight: How AI Empowers and Deepens Systemic Racism. (Boca Raton, FL: CRC Press, 2023), p.35 https://a.co/d/hB6u7Pe.?
14. Misra SM, Suresh S. Artificial Intelligence and Objective Structured Clinical Examinations: Using ChatGPT to Revolutionize Clinical Skills Assessment in Medical Education. J Med Educ Curric Dev. 2024 Jul 25;11:23821205241263475. doi: https://doi.org/10.1177/23821205241263475.?