In February 2019, a Chinese study found that an artificial intelligence (AI) system trained on data from 1.3 million patient visits to a Guangzhou medical center could outperform some doctors in diagnosing common childhood diseases. Eight months earlier, another AI system had scored 2:0 against top Beijing physicians in diagnosing brain tumors. Meanwhile, Chinese tech giant Tencent is in the midst of clinical trials for an AI program that will be able to diagnose patients with Parkinson’s disease.
AI is transforming China’s healthcare industry. Hospitals across the country have started adopting AI-assisted medical products and services that include not just image-assisted diagnosis, but also range from surgical robots to smart patient health monitoring and virtual medical assistants. New medical AI start-ups are emerging by the minute while traditional healthcare companies and leading technology firms are betting big on AI solutions.
China’s medical AI boom is strongly supported by the central government, which aims for China to become the world’s leading AI innovation center by 2030. The application of AI in healthcare is a priority as Chinese leaders look for ways to relieve the immense pressure that a rapidly urbanizing and aging population is putting on the country’s healthcare system. The hope is that new AI-driven treatment methods can address doctor shortages, high misdiagnosis rates, inaccuracies in early disease prediction, and much more.
As a result, China’s AI health market is booming, with estimations putting its size at RMB 20 billion (roughly $3 billion) in 2018, nearly double its size in 2017. Yet, mainstream media coverage often fails to fully acknowledge the vast range of regulatory challenges that may prevent US and foreign companies from tapping into these opportunities. If multinationals want to be successful in providing medical AI solutions to Chinese patients, they must be acutely aware of and prepared to mitigate a range of operational risks.
The biggest wildcard for US companies is how open the Chinese government will be to foreign participation in its domestic AI market.
Beijing is putting large amounts of resources and support behind the development of domestic companies. China has selected a “national team” to spearhead AI development. Tencent is one of these companies designated to become a leader in AI-assisted medical diagnostics.
Furthermore, medical robotics manufacturers such as Tinavi have secured millions of dollars in state funding and are receiving preferential treatment in the tendering, registration, and procurement processes.
China also still lacks a robust regulatory framework for AI-powered software and devices. Much like in the United States and other parts of the world, many products in China are still a long way from being commercialized.
Led by the National Health Commission and the National Medical Product Administration, healthcare regulators have been actively working on creating classification catalogs, clinical trial guidelines, and technical review documents specifically for medical AI. In February 2019, the Center for Medical Device Evaluation issued evaluation guidelines for medical devices using AI for clinical decision making. Several working groups and expert committees have been created with the specific task of creating further standards and guidelines.
Yet, the lack of unified industry standards for risk classification and evaluation of medical AI is still causing significant delays in product registrations. Regulators are taking a distinctly conservative approach, with current medical device catalogs favoring the highest risk classification (Class III) for products or solutions as soon as they incorporate deep learning with explicit diagnosis functions. Many of the 50+ start-ups in China that have invested heavily in algorithm-driven diagnostic products may not survive due to the complexities and high costs of medical device registration. Increasingly, analysts suspect that as much as 90 percent of all Chinese AI start-ups could fail amid fierce competition, pressure to commercialize, and a wider economic slowdown.
The data conundrum
AI’s inherent dependence on data is at odds with the government’s goal of protecting sensitive medical data.
Data sharing is essential to ensuring the wide and safe application of AI-driven medical products. Datasets used to train AI systems need to be drawn from a wide variety of hospitals and patients. For this reason, regulators have recently taken steps to improve China’s currently underdeveloped and unstandardized electronic resource infrastructure. The National Health Commission has issued policy documents to promote the use of electronic medical records across hospitals and assigned local regulators to assess how well they are being adopted.
However, under China’s Cybersecurity Law and related implementation measures, much of the data generated in a medical context are classified as sensitive, potentially subjecting them to strict data localization requirements or security reviews. Initial healthcare-specific guidelines are requiring medical institutions to store patient information on secure servers within China and to apply for security evaluations if they wish to transfer it overseas.
Foreign companies are therefore at risk of cross-border data flow regulations hindering their ability to collect the necessary datasets to localize their products in the Chinese market.
Ensuring ethical and safe medical AI
The quality of the patient data that is being collected and processed will also come under growing scrutiny. Experts in the US and other parts of the world are highlighting that the very data that fuel the algorithms producing medical diagnoses are often of poor quality and have clinical relevance of only four months.
In recent months, ethical and safety considerations have been moving to the forefront of Chinese AI conversations among policymakers and industry experts. China’s Ministry of Science and Technology recently issued a widely anticipated document providing eight principles for the governance of AI in order to promote its “safe, controllable and responsible use.” In doing so, China may be taking its own path to regulate ethical and safety issues—it was not one of the 42 countries that endorsed the OECD’s first international set of AI guidelines issued just a few weeks earlier.
With the full safety risks of medical AI still unclear and lives potentially at stake, it remains to be seen to what extent Chinese regulators and patients will be comfortable with letting algorithms make complex disease diagnoses and devise treatment plans based on still fragmented and unstandardized medical records. Other question marks remain over how legal liability will be determined for patient injuries resulting from AI devices and solutions.
Until guidelines are clarified at the national and sectoral levels in China, companies will face uncertainty regarding the commercialization of their AI products.
Capitalizing on opportunities
US companies eyeing the significant opportunities that China’s healthcare system offers for medical AI applications should not underestimate the challenges of market entry and expansion. Businesses that do not have the foresight and resources to deal with these obstacles may struggle to thrive in a market that will face increasing regulatory scrutiny and growing numbers of ambitious Chinese competitors propped up by state funding.
Nevertheless, American companies with top-notch AI solutions that help China address critical healthcare issues can still be highly successful if they position themselves strategically and prepare for regulatory risks.
Even with substantial government funding, many Chinese companies currently struggle to close the technology gap between them and their foreign competitors, and fierce competition from smaller start-ups makes the leadership of China’s designated “national champions” far from certain. The healthcare sector is also unlikely to see overly excessive favoritism of domestic companies since the government cannot risk jeopardizing the quality of medical services.
Foreign companies with strong technical capabilities are attractive partners for Chinese companies to co-develop medical AI applications. Siemens Healthineers, for example, has positioned itself as a key partner in the development of China’s digital healthcare ecosystem. The company is working with Chinese players including Tencent Medical, Deepwise, and Infervision to launch a platform for AI-assisted diagnosis and cancer screening.
At this crucial stage of development, businesses with strong policy expertise also have the unique chance to help Chinese regulators keep a balanced view of the practical risks and benefits of medical AI by sharing best practices. China’s various AI technical committees and working groups may provide useful channels for providing technical advice on how to approach critical issues surrounding data, safety, and ethics without hindering AI development or disadvantaging US companies. Advocacy efforts should include ensuring unhindered data flow across borders and definitions of “secure and controllable” medical AI that are not based on the nationality of the product or service.
* * *
The rewards of engaging the various players in China’s AI ecosystem on the ground are noteworthy as China continues its rapid push for AI commercialization. Amid emerging Chinese ambitions to set AI standards at a global level, being a part of early AI discussions in China will be crucial for companies’ global success.
Caroline Meinhardt is a project consultant in APCO Worldwide’s Beijing office, where she consults multinational clients on how to navigate China’s regulatory landscape. Her work focuses on the technology and healthcare sectors, especially China’s industrial upgrading initiatives, the government’s push for artificial intelligence, and its cybersecurity regime. Read Caroline’s full bio.
Photos provided by Infervision.