The interview partners
Prof. Dr. Patrick Da-Cruz is Professor of Business Administration and Health Management at the Faculty of Health Management at Neu-Ulm University of Applied Sciences (HNU) and Academic Director of the MBA programme in Leadership and Management in Healthcare.
Prior to joining HNU, Mr. Da-Cruz worked for renowned strategy consultancies in the pharmaceutical/healthcare sector and held management positions in healthcare companies in Germany and abroad.
Chenchao Liu is the founder and managing director of SILREAL GmbH. For over a decade, he has been advising ministries, international pharmaceutical companies and healthcare providers on health policy, market access between China and Europe, and digital transformation – with projects for the Federal Ministry of Health, AstraZeneca, Bayer and EFPIA, among others. The PMP-certified chemist (Technical University of Munich) is co-author of several specialist publications and regularly publishes articles in leading media outlets such as the Neue Zürcher Zeitung, the World Economic Forum and the New York Times.
China is considered a pioneer in the use of AI in medicine – which specific applications in Chinese hospitals impress you the most?
Chenchao Liu: I am particularly impressed by radiology: AI algorithms triage CT/MRI images, mark abnormalities and prioritise findings, which reduces ‘time-to-treat’ in cases such as stroke. Products such as InferRead CT Lung from Infervision have been shown in studies and approvals that AI can increase sensitivity in lung cancer detection – some with FDA/NMPA approval and proven performance advantages in the detection of fine nodules.
In addition, I see AI-supported triage and chatbots in outpatient clinics/polyclinics, structuring medical histories and directing patient flows. Evaluation studies show that such ‘intelligent guidance chatbots’ are usable in practice in China – a building block for relieving scarce resources.
Third: platform approaches by large technology companies. Tencent AIMIS, for example, links image data, research and workflow – an example of how infrastructure layers facilitate the scaling of clinical AI.
Fourthly: AI assistance at the point of care. According to the company, Ping An AskBob already supports over a million doctors with decision-making knowledge and pathways – especially in underserved regions.
What specific advantages does China's use of AI in hospitals offer?
Chenchao Liu: Three points:
- Scaling: China is rapidly rolling out successful digital/AI models. Official figures: Over 3,000 internet medical institutions have been offering online services since the end of 2024, illustrating the ability to spread digital care models nationwide.
- Database & standards: National programmes such as Smart Hospital Grading and thematic guidelines create compatibility in processes/IT – a foundation on which AI can be reliably embedded in clinical pathways.
- Care logic: AI is primarily understood as an aid – for prioritisation, quality assurance and workload reduction, not as a replacement. This facilitates broad acceptance (see chatbot usability, AskBob usage).
How are medical and nursing staff in China being inspired and trained to use AI?
Chenchao Liu: Success comes from what happens in the workplace: ‘super-user’ models that provide hands-on support to ward teams; gradual introductions per use case (e.g. radiology triage), measurable via turnaround time and diagnostic quality. There are also frameworks (e.g. smart hospital standards) and thematic guidelines for AI application scenarios that specify processes, responsibilities and quality assurance – this reduces implementation risks and makes training easier to plan.
What are the biggest hurdles to integrating AI in German hospitals – compared to China?
Chenchao Liu:
- Fragmented IT & data: Many hospitals struggle with heterogeneous HIS/RIS/PACS landscapes and low interoperability – but AI needs consistent, structured data.
- Regulatory & compliance path: Under the EU AI Act, AI systems in medicine are considered high-risk with additional requirements (including risk management, data quality, monitoring). Deadlines are staggered – bans apply from February 2025, GPAI rules from August 2025 and transitions for high-risk systems embedded in products until August 2027.
- Change & skill gap: Implementation rarely fails because of algorithms, but more often because of process redesign, training and governance.
- Data availability: Even though Germany is taking a big step forward with the ePA-for-all (opt-out) from 2025, clear, clinic-oriented data strategies and usage paths for care and quality are needed.
If you had to develop an AI strategy for a German hospital tomorrow, which approach from China would you adopt immediately?
Chenchao Liu:
- ‘Small bets, fast scaling’ – a systematic pilot and rollout approach. Start with two clinical beacons (e.g. radiology triage and sepsis/bed forecasts). Define clear outcome KPIs (findings turnaround, door-to-needle, length of stay, transfers). Successful projects are scaled to other hospitals/departments within 3–6 months – playbooks and parameters are carried over rather than reinvented. (Role model: rapid scaling of digital care models in China.)
- Data strategy as a top priority. Build an interoperable data layer (FHIR profiles, terminologies), ensure data quality, and establish use case governance (clinical responsibility, metrics, drift monitoring). This creates the basis for robust AI operation as a process component – and at the same time fits in with the EU AI Act path (risk management, logging, post-market monitoring).
- Workforce enablement ‘in the line’. Train at the point of care with super users, simulate real cases, introduce safety/ethics checks in SOPs, and link the whole thing to the department's quality targets (e.g. stroke unit times). China's smart hospital thinking shows that standards + on-site enablement increase acceptance.
- Platform over point solutions. Focus on platforms (PACS/VNA/workflow orchestration) that integrate multiple AI models and enable A/B comparisons – this allows you to remain vendor-agnostic and prioritise use cases (radiology, pathology, surgical planning). Examples such as Tencent AIMIS illustrate the benefits of platforms – it is the architectural principle that is transferable, not the specific solution.
- Public-private learning loops. Work with industry/start-ups in outcome-based contracts (e.g. linking remuneration to time/quality metrics). In China, for example, reliable evidence paths for radiology AI were quickly established (e.g. lung screening).
- Policy/payer alignment. Align in-house AI strategy with national programmes (ePA for all, cross-sector data rooms) – this creates reimbursement and scaling pathways instead of isolated solutions.
Practical relevance: Policy as a catalyst
A good example of how policy drives and scales innovation: COPD was included in China's National Basic Public Health Service in 2024 – such decisions increase data availability and clarity of care pathways, which AI can then build on. For Germany, this means that internal hospital AI strategies should be integrated with public health pathways and registries at an early stage.
Thank you very much for talking to us!
The content and statements presented in the interviews reflect the perspective of the interviewees and do not necessarily correspond to the position of the editorial team.
