Three‐Stage Evolution of China's Breakthrough Therapy Drug Policy: AI‐Driven Quantitative Evidence and Global Implications

Background Many countries have accelerated the approval of innovative drugs through breakthrough therapy policies to meet public health needs. China needs to optimise its policy system to narrow the gap with international practices. Objective To identify shortcomings in China's breakthrough therapy policies and formulate optimization strategies via quantitative analysis of their dynamic evolution and cross‐jurisdictional comparisons. Methods Adopting a quantitative design, this study employs AI technologies (LDA topic modelling + time‐series deduction), combined with cross‐jurisdictional policy element matrix comparison and typical drug case analysis, to systematically analyse the evolutionary path and implementation effects of China's breakthrough therapy drug policies. Results Quantitative analysis of 59 policy texts (2020–2024) reveals that China's policies have undergone a three‐stage evolution: ‘approval‐driven (2020–2022)—collaborative transition (2023–2024)—ecological construction (exploratory scenario deduction after 2024)’. The core topics focus on ‘drug review’, ‘industrial innovation’, and ‘medical services’, among which ‘drug review’ and ‘industrial innovation’ form a dual‐dominant pattern. However, the persistently low intensity of the ‘medical services’ topic reflects an ecological imbalance of ‘accelerated approval—industrial innovation—lagging accessibility’. Cross‐jurisdictional comparisons show gaps in early innovation incentives, international collaboration, and medical support. Conclusions Based on the quantitative analysis of policy texts, this study proposes six optimization suggestions. These targeted strategies aim to address the imbalance of ‘fast approval but slow accessibility,’ support the ‘Healthy China 2030’ strategic goal, and provide references for developing countries in balancing resource allocation, accelerating research and development through the integration of artificial intelligence, and avoiding accessibility disparities via risk‐sharing mechanisms.