Cost-utility analysis of deep learning-based opportunistic osteoporosis screening using chest radiographs:
Bo-Xiang Shi1,2, Dung-Jang Tsai3,4, Chin Lin3,4
1School of Pharmacy, College of Pharmacy, National Defense Medical University, Taipei, Taiwan, ROC.
Abstract:
Osteoporosis often goes undetected until a fracture occurs. Using a deep learning model to analyze routine chest radiographs may cost-effectively identify at-risk individuals. Economic benefit depends on how common the disease is, how the model is calibrated, and whether patients follow through with confirmatory bone density testing and pharmacological treatment initiation.
Background:
Osteoporosis is common among postmenopausal women and often unrecognized until fragility fractures occur, despite its clinical and economic burden. Deep learning (DL)-based opportunistic screening using existing chest radiographs is a potential approach, but its economic value may vary with disease prevalence, diagnostic performance, and care pathway completion.
Aim:
To assess the economic value of a DL-based opportunistic osteoporosis screening approach among Taiwanese postmenopausal women.
Methods:
A cost-utility analysis was conducted from the Taiwan healthcare perspective using a hybrid decision tree and Markov model. The model simulated a cohort of 50-year-old postmenopausal women over a 50-year horizon. Model inputs, including diagnostic performance, long-term outcomes, costs, and utilities, were derived from published literature, preferably local sources. Additional analyses examined alternative prevalence settings, sensitivity and specificity combinations, and confirmatory DXA and treatment initiation rates.
Results:
In the base case, assuming full care-pathway completion, DL-based screening was cost-effective versus no screening, with an ICER of 21,827 per QALY, below the willingness-to-pay threshold of one GDP per capita in Taiwan. The approach became cost-effective when osteoporosis prevalence exceeded 14.3%. Across the diagnostic continuum, a sensitivity of 86.6% and specificity of 76.3% generated the greatest economic benefit.
Conclusion:
DL-based opportunistic osteoporosis screening was cost-effective under full care-pathway implementation. Its value depended on disease prevalence, diagnostic performance, and completion of confirmatory DXA and treatment initiation, and was not maintained under real-world examination estimates observed in Taiwan. Strengthening the post-screening care pathway is therefore central to realizing this value.
