Economic Viability of Artificial Intelligence Assistance in Lung Cancer Screening Using Computed Tomography

ABSTRACT In Australia, health authorities have not implemented lung cancer screening but have explored its potential implementation through various inquiries. Evidence suggests that combining AI and CT scans for screening holds promise in terms of cost–benefit ratio, cost savings, improved healthcare outcomes, optimized resource allocation, and enhanced efficiency in lung cancer diagnosis by radiologists. Therefore, this study aimed to evaluate the cost‐effectiveness of an AI system for the initial annual lung cancer screening. In this retrospective analysis, a Markov model—built using data from published sources—was applied to estimate quality‐adjusted life years (QALYs) and lifetime costs for various diagnostic strategies. To ensure a thorough evaluation of model uncertainty and AI system expenses, the study applied both deterministic and probabilistic sensitivity analyses, capturing variability across diverse input scenarios. Using AI significantly reduces the cost of detecting lung cancer, from $150,000 to $300, and from €138,000 to €276. Simultaneously, there is a marginal increase in QALYs from 14.0 to 14.2. These findings claim that AI can reduce the cost of screening by 99.8% and increase the QALYs by 0.2. AI (reduced price) for lung cancer screening is more cost‐effective than the original CT cost scenario. Both methods reach full cost‐effectiveness, with AI (reduced cost) lowering mortality by 20% and standard CT achieving a 40% drop. Beyond 50% reduction, cost‐effectiveness wanes; the optimal range remains between 20% and 50%, where CT plus AI proves economically strong. Integrating AI with CT scans in lung cancer screening is a cost‐effective approach that provides superior value for money. Defining cost thresholds for AI can speed up its adoption in clinical practice, and further research is required to assess the combination of AI models with specialized radiologists and address uncertainties regarding AI system accuracy.