【1】Hospitals “dare not go out”, scientific research “can’t be done”, and patients “can’t get it”
Zhu Tongyu analyzed the problem and pointed out the causes of triple obstruction.
The first is that hospitals “don’t dare to go out”: compliance risks and liability avoidance. Current laws and regulations are extremely strict on accountability for medical data leaks. Hospitals lack the ability to distinguish between "original data" and "data value." It is difficult to judge which cooperation is compliant and can only choose the most conservative solution. In this high-pressure environment, hospital administrators' preferred strategy is to restrict data outflows "one size fits all."
The second is that scientific research is “impossible”: the governance chain is broken. The cross-organizational data retrieval rate is less than 15%, unstructured data accounts for more than 70%, and the data missing rate and incorrect coding rate exceeds 30%. Without professional data management, these "dirty data" cannot support high-quality AI model training. AI companies are difficult to enter, and hospitals themselves lack governance capabilities, creating an endless cycle.
Finally, patients “cannot get it”: obstacles to the realization of rights. Although policies encourage the expansion of the scope of openness of health and medical data to individuals, in practice, patients who want to obtain their complete electronic medical records often face problems such as cumbersome procedures, inconsistent formats, and even rejection.
【2】Comprehensive promotion of the “trusted data space” model
While pointing out the problem, he also proposed a solution, which is to build a "three-way win-win" mechanism.
First, use technology to break the situation and comprehensively promote the "trusted data space" model. The core idea is to change the traditional idea of "data relocation" and implement a new model in which "data does not leave the domain, knowledge can be circulated, and value can be shared".
Specifically, there are three points:
1. Building a trusted data space for the industry: Drawing on the successful experience of Guangzhou’s trusted data space for the health industry, the National Health Commission, in conjunction with the National Data Administration, will pilot the construction of a trusted medical data space in key cities and hospital groups.
2. Deploy privacy computing and intelligent computing power: Deploy technical facilities such as privacy computing and federated learning in the data space, and provide hospitals with intelligent computing power support. The algorithm models of AI companies run in the hospital, and the data are managed and trained locally. Only the parameters and results are taken away, and the original data always remains in the hospital.
3. Build a unified multi-modal common data model (SZCDM): It will greatly improve the efficiency of unstructured data processing and significantly reduce the cost of data cleaning.
The second is compliance protection, establishing hierarchical authorization and full-process traceability mechanisms. The core idea is to transform hospitals from "not daring to go out" to "going out in compliance with regulations." There must be clear rules and traceable technical guarantees.
There are also three specific suggestions:
1. Implement the minimum necessary principle and hierarchical management: follow the minimum available principle and clarify hierarchical access rights for different scenarios such as clinical diagnosis and treatment, scientific research, and teaching.
2. Introduce digital contracts and audit closed loop: Introduce a "digital contract" mechanism into the data space, sign a smart contract for each usage scenario, clarify the purpose, authority and time limit, and achieve full process traceability through blockchain technology.
3. Clarify data ownership and authorization mechanisms: Micro level: Medical and health data involving individuals is personal information, and applications must respect patients’ right to informed consent. For scientific research purposes, a dynamic and specific secondary consent mechanism needs to be established. Macro level: Medical institutions are responsible for application management of desensitized anonymized data, but they must be supervised by government departments.
The last step is to empower patients: get through the “last mile” of data acquisition. The core idea is that patients are the source of data, and they must be made the beneficiaries of data value release, rather than bystanders.
Specific suggestions include:
1. Build a unified patient data query and download platform: Relying on the national national health information platform, promote the construction of a unified "resident health record access portal" in all provinces and cities. After passing the real-name authentication, patients can download their complete electronic medical record data (including structured data and imaging materials) with one click, breaking the deadlock of "the data can be viewed but not taken".
2. Standardize data formats and interface standards: Mandate that patient export data provided by medical institutions comply with national health informatization standards to ensure that data can be identified and reused when patients are transferred between different hospitals.
3. Establish a data circulation mechanism authorized by patients: In scenarios such as patient diagnosis and treatment or commercial health insurance claims, patients are allowed to authorize their data to third-party organizations through authorization codes and other methods to achieve "data follows the patient". Beijing has proposed to explore the establishment of a medical insurance and medical data sharing mechanism to provide support for quick claims for commercial health insurance. This is a typical case of benefiting patients.
The value of medical data cannot be released at the expense of security, nor can innovation be hindered by wasting food. The key to solving the dilemma of "data not being discharged from the hospital" is to build a new data ecosystem with "technically controllable, clear rules, benefit sharing, and patient benefit". It is recommended that relevant departments introduce detailed policies as soon as possible to support the construction of a trusted data space and promote the transformation of medical data from a "sleeping asset" to a "flowing engine."
Reporting from Beijing by Jiupai News special correspondent Li Yang
Editor Wu Diwu Feifei
