AI in Cancer Care: Navigating the Current Landscape and Future Possibilities
The 2023 Clinical Pathways Congress & Cancer Care Business Exchange, which took place in Boston, Massachusetts, October 7-9, featured a session titled “AI in Cancer Care: Separating Fact from Fiction,” spearheaded by moderator Ira Klein, MD, where panelists delved deep into the present and future implications of artificial intelligence (AI) in cancer care. During the discussion, the speakers shed light on key considerations for leveraging AI effectively in cancer care, emphasizing the importance of discerning data accuracy, the role of human oversight, and the need for comprehensive understanding and utilization of AI-generated insights.
The Benefits and Pitfalls of AI in Cancer Care
Will Shapiro, vice president of data insights engineering at Flatiron Health, summarized the fundamental functioning of generative AI language models in layman’s terms. Highlighting the process of predicting text patterns, Shapiro underscored the necessity of establishing a reliable ground truth for ensuring unbiased and accurate AI output. He emphasized the critical role of high-quality data and the significance of collaboration between subject matter experts and AI programmers to achieve reliable results. As Shapiro put it, the old adage of “garbage in, garbage out” holds especially true in the realm of AI.
Building on this, Zachary Taft, MBA, digital ventures lead at Memorial Sloan Kettering, emphasized the indispensable role of human oversight in validating AI-generated data. He stressed the significance of using AI to streamline the delivery of data to a larger patient population, augmenting the capability of specialists, and fostering collaboration between clinical practitioners, providers, and payers. However, he cautioned against overhyping AI, warning that inaccurate data collection could lead to misguided decisions and potential harm to patients.
Echoing a cautious optimism, James Hamrick, MD, vice president of clinical oncology at Flatiron Health, cited the incremental approach to incorporating AI into direct decision-making processes. Despite recognizing the potential of AI to alleviate administrative burdens and refine clinical pathways, Dr Hamrick stressed the importance of performing diligent review and maintaining zero margin for error, especially in clinical decision support tools. Establishing trust in AI output, he emphasized, is pivotal for widespread adoption among clinicians.
Amy Valley, PharmD, vice president of clinical strategy and technology solutions for Cardinal Health, underscored the imperative of understanding the origins and development of AI technology. Stressing the importance of verifying AI-based results, Dr Valley advocated for the integration of social determinants of health (SDOH) into clinical care, thereby enhancing patient tracking, identifying at-risk populations, and improving overall health care outcomes.
Ed Rodgers, director of network development for Elsevier’s ClinicalPath, stressed the significance of maintaining a human-in-the-loop approach and fostering a healthy level of skepticism when utilizing AI in clinical practice. Noting the importance of transparent documentation and training AI models on real-world outcomes, Rodgers emphasized the need to integrate AI seamlessly into existing clinical workflows without undermining the role of human expertise.
The Future of AI in Oncology Care
The participants also shared their views on the evolving role of AI in cancer care. Taft envisioned AI taking on more supportive roles, closing quality-of-care gaps, and aiding in decision-making across various tasks. Shapiro expressed enthusiasm for AI’s potential to expedite drug discovery processes, potentially revolutionizing cancer treatment. Dr Hamrick highlighted AI’s capability to craft evidence-based treatment plans and facilitate patient eligibility for clinical trials. Dr Valley predicted AI’s role in leveling the health care playing field and alleviating administrative burdens, allowing practitioners to focus on delivering optimal care. Rodgers emphasized the need for a symbiotic relationship between humans and machines, with AI primarily handling operational tasks while humans focus on critical decision-making processes. “To quote Fatboy Slim,” Rodgers said, “Machines should work; people should think.”
As the health care industry embraces technological advancements, understanding the practical applications and potential pitfalls of AI in the oncology domain is imperative. In essence, the session highlighted the need for a meticulous approach to AI integration in cancer care, stressing the importance of data accuracy, human oversight, and trust-building measures. While AI holds immense promise in revolutionizing cancer care, the speakers cautioned against overreliance and emphasized the critical need for maintaining the primacy of human expertise and ethical considerations in health care decision-making. As the field continues to evolve, collaboration between health care stakeholders, data scientists, and AI developers will be pivotal in ensuring that AI in cancer care remains firmly grounded in empirical evidence and patient-centered outcomes.