AI and Human Radiologists Team Up to Reduce Mammography Screening Costs by Up to 30%

New research co-authored by Mehmet Eren Ahsen, a professor of business administration and Deloitte Scholar at the University of Illinois Urbana-Champaign, suggests that integrating artificial intelligence (AI) with human radiologists in breast cancer screening can significantly reduce costs without compromising patient safety. The study, published in Nature Communications, examined the effectiveness of three screening strategies: expert-alone (current norm), full automation, and a delegation model. In the expert-alone approach, radiologists individually assess every mammogram. While this method ensures high accuracy, it is labor-intensive and expensive. Full automation, on the other hand, relies entirely on AI algorithms to screen mammograms. However, this approach is fraught with risks, as current AI systems are inadequate for handling complex or borderline cases, leading to potential misdiagnoses. The delegation strategy stands out, where AI initially screens mammograms, identifies low-risk cases, and flags high-risk or ambiguous ones for radiologists to review. This approach can cut screening costs by up to 30% without sacrificing diagnostic accuracy, according to the study. The researchers developed a decision model to evaluate the economic and clinical outcomes of each strategy using real-world data from a global AI crowdsourcing challenge. The challenge, part of the White House Office of Science and Technology Policy's Cancer Moonshot initiative (2016-17), provided a robust dataset to test the models. The delegation strategy emerged as the most cost-effective, saving up to 30.1% compared to other methods. This model can streamline the screening process, potentially reducing the number of false positives and the associated emotional and financial burdens on patients. Annual mammography screenings in the U.S. number nearly 40 million, generating significant costs due to follow-up procedures for false positives. A 10% false positive rate translates to about four million women being unnecessarily recalled, leading to additional appointments, screenings, and tests. This not only exacerbates healthcare spending but also causes considerable stress and anxiety for patients. With AI's ability to handle straightforward, low-risk cases, the delegation model aims to alleviate these issues by freeing radiologists to focus on more challenging and critical cases. Ahsen highlighted the importance of AI's role in enhancing workflow efficiency. "AI can be used around the clock, without breaks, and it excels at identifying low-risk mammograms quickly and accurately," he said. "By flagging high-risk cases immediately, AI can help ensure that patients receive follow-up care more promptly, reducing wait times and alleviating some of the anxiety associated with uncertainty." The study's implications extend beyond breast cancer screening. It raises questions about the broader application of AI in medicine, particularly in fields like pathology and dermatology, where diagnostic accuracy is paramount. The researchers found that the delegation model works optimally in populations with low to moderate breast cancer prevalence, suggesting that human expertise remains crucial in high-prevalence settings. Additionally, the legal and regulatory landscape must adapt to accommodate AI in healthcare, addressing concerns about liability if AI systems are held to stricter standards than human clinicians. Ahsen emphasizes the potential benefits of the delegation model in resource-scarce settings, such as developing countries with limited radiologist availability. "An AI-heavy strategy could be a game-changer in these regions, improving access to critical diagnostics," he noted. The research provides a comprehensive framework for hospitals, insurers, policymakers, and healthcare practitioners to make informed decisions about AI integration. Ahsen underscores the importance of evidence-based decision-making, considering not just what AI can do, but also whether and when it should be deployed. "We need to explore the conditions under which AI can be a valuable tool to help human experts, rather than a replacement," he said. Industry insiders agree that the delegation model holds promise. They believe it could set a precedent for AI integration in other medical diagnostics, leading to more efficient and equitable healthcare systems. The University of Illinois Urbana-Champaign, known for its interdisciplinary approach to research, continues to spearhead innovations at the intersection of technology and healthcare, positioning itself as a leader in the field. Overall, the study demonstrates the potential of AI-human collaboration to enhance the efficiency and accuracy of mammography screening, offering a scalable solution to address the growing demand for early breast cancer detection and the radiologist shortage. As AI technology evolves, the delegation model presents a pragmatic and ethical approach to integrating these advancements into healthcare practices.