From Security to Accuracy


Chief Information Officer Perspective: Charles Gruber shares his thoughts on AI and the medical industry

 

The integration of artificial intelligence (AI) large language models (LLMs) into the healthcare industry symbolizes a significant leap in technological advancement. Having attended the Ai4 conference—a notable industry event showcasing AI innovations across various sectors—I have observed firsthand the remarkable applications of AI in medicine. The medical community appears poised to lead the pursuit of accurate, cost-effective, and responsible use of AI large language models (LLMs), with human expertise serving as an essential component in risk mitigation. This commitment to excellence reminds me of the standards set in the banking sector for information security. Healthcare is setting the stage for the future of LLMs for us all.


Banking's Legacy: A Beacon for Security

The banking sector's vigilance over financial data has led to breakthroughs in information security. Tools such as multifactor authentication, encryption, and biometric verification have fortified the financial industry and inspired others. This history of proactive risk mitigation sets an example for emerging fields, such as AI LLMs.


AI Large Language Models in Healthcare

The utilization of AI LLMs in healthcare has opened doors to remarkable improvements:
  • Improving Diagnosis:
    • Accuracy and Reliability: LLMs can harness clinical data to derive precise information, exceeding traditional capabilities.
    • Efficiency: Streamlining the diagnosis process, LLMs enable medical professionals to focus on critical tasks, enhancing overall productivity.
    • Interpreting Medical Tests: LLMs can interpret complex data, providing valuable insights for diagnosis or treatment, serving as invaluable support for medical practitioners.
  • Reducing Costs: According to a working paper done though the National Bureau of Economic Research (Sahni et al., 2023), the integration of AI could yield substantial financial impact in the U.S. healthcare system:
    • The integration of AI could yield $200-360 billion in annual savings in the U.S. healthcare system, a substantial financial impact.
    • AI can refine hospital operations, optimizing resources and improving patient access.
    • AI can identify patients at risk, facilitating timely interventions to prevent deterioration.
    • AI can bolster clinical decision-making, enhancing quality of care and health outcomes.


The NBER paper also outlines strategies for responsible adoption and mitigation of the risk around AI LLMs:
  • Designing AI integration to retain human decision authority: AI insights should be integrated into clinical workflows while retaining clinician and patient authority over final decisions.
  • Validating algorithms to ensure robustness: AI algorithms need to be validated to ensure that they are clinically robust and safe, such as through FDA approval processes.
  • Establishing governance to manage privacy risks: Governance models and processes should be established to responsibly manage patient data privacy and model risk.
  • Assessing and addressing bias to maintain fairness: Potential biases in AI algorithms and data need to be assessed and addressed to maintain fairness.
  • Refreshing models to adapt to healthcare changes: As healthcare generates new types of data, AI models may need refreshing to avoid cementing biases.
  • Multidisciplinary collaboration for responsible AI development: Collaboration is needed between business, technology, clinical, and other leaders for responsible AI adoption.


Medical Industry: The Vanguard of AI Innovation

Recently, I attended the Ai4 conference, where I witnessed several highly innovative and impressive medical use cases for generative AI LLMs. Two vendors that stood out to me were Vital and ProvARIA.
  • Vital (vital.io) has introduced a "Doctor-to-Patient translator" that simplifies medical visit summaries and test results into explanations accessible at a fifth-grade reading level. This HIPAA-compliant AI tool offers interpretations of various medical documents, including lab results, doctor notes, and discharge summaries. Available for free on any browser, the tool aims to alleviate patient confusion and anxiety, enhance healthcare equity, and potentially reduce physician burnout, which surged to 63% by the end of 2021. The company emphasizes the importance of not merely summarizing the notes but extracting and organizing the most crucial information, reflecting a human understanding of what's significant. Interestingly, the founder of vital.io was also a founder of mint.com, a popular finance platform that was sold to Intuit in 2009.
  • Providence Analytics, a division of Providence Health (providence.org), showcased its ProvARIA patient medical triage system for the medical back office. Dr. Ford Parsons, MD, MS, elucidated how ProvARIA employs AI and LLMs to decipher and automatically categorize patient messages, subsequently generating recommended draft responses. The implementation of this system has led to tangible improvements in efficiency and communication. Specifically, there has been a 39% reduction in turnaround time for patient messages, enhanced prioritization of communications, and a more comprehensive exchange of information with fewer iterations.

Both of these companies, along with others showcased at the Ai4 conference, have meticulously designed their strategies to minimize risk with AI LLMs. A common theme across these strategies is the inclusion of human oversight within the AI-driven processes. This human element, now greatly enhanced in efficiency, is liberated from routine and repetitive tasks, allowing for a concentration on more complex and value-added functions.


The Road Ahead

While the banking sector has set standards for information security through proactiveness, cooperation, and collaboration across industries, the medical sector is now poised to establish norms for the accurate and responsible use of AI LLMs. With careful planning, stringent standards, and ethical commitment, AI LLMs can transform healthcare into a more efficient, cost-effective, and patient-centered domain. As we continue to explore the possibilities of AI, it is crucial to remember that the most likely outcome is not to replace human expertise but to augment it, creating platforms that leverage the best of both human and artificial intelligence. The ongoing collaboration between human expertise and artificial intelligence promises to shape LLMs for the challenges ahead.

The medical community, as pioneers in this field, holds the key to unlocking the full potential of AI LLMs, setting a course that will resonate across all sectors. By leading the responsible integration of AI, medicine can chart a path toward a brighter future.


References 

Sahni, N., Stein, G., Zemmel, R., & Cutler, D. M. (2023). The potential impact of artificial intelligence on healthcare spending. National Bureau of Economic Research. https://www.nber.org/system/files/working_papers/w30857/w30857.pdf