After opening up healthcare data, actuarial models can now support disease-specific recurrence insurance for particular conditions, enabling some individuals previously considered "uninsurable" to be covered. The development of insurance for people with pre-existing conditions has entered its second phase—moving away from a broad one-size-fits-all model to customized development based on disease type, progression stage, and condition control. This shift is driven by efficient medical data sharing, extending the reach of such insurance to more patients.
A year ago, the Shanghai Financial Regulatory Bureau, together with the Municipal Medical Security Bureau and the Municipal Health Commission, released measures aimed at improving health data-sharing capabilities. The goal was to create a collaborative mechanism for sharing data among medical institutions, insurers, and the commercial insurance sector through the municipal big data center and certified third-party organizations. This was designed to support product development for specific groups and specific risk profiles, as long as data security and privacy are maintained. The Shanghai Clinical Innovation and Translation Institute—a market-oriented platform led by the Shanghai Shenkang Hospital Development Center—was established to put this policy into action. Over the past year, the institute has worked with commercial insurers to harness real healthcare data within a secure framework, leading to the development of several new disease-specific recurrence insurance policies.
In a recent conversation, Pan Jiqi, head of the institute's commercial insurance collaboration unit, noted that the key to developing these policies lies in unlocking the value of medical data and translating it into an actuarial language. Yet, with data sharing as the starting point, achieving a complete value chain requires addressing two practical questions: How can medical institutions be encouraged to engage in the full-course management of patients? And how can the challenge of reaching potential buyers for this type of insurance be solved?
With data support, the development timeline for a disease-specific recurrence policy is just three to six months.
Pan Jiqi explained that the institute acts as a bridge, coordinating with government bodies such as the health commission, the data bureau, and the big data center, as well as the ethics review board of the Shanghai Shenkang Hospital Development Center, while also addressing the commercial insurance needs of top-tier municipal hospitals. Under the guidance of the health commission, data bureau, and big data center, the data is reviewed by the ethics committee, then anonymized and registered on a blockchain within the big data center's secure zone. All analysis and development work happens in a dedicated, secure processing space. From the point an insurer commissions a project on recurrence insurance to completing data anonymization, actuarial pricing, and creating a prototype product, we have found that the cycle takes roughly three to six months based on our experience with several product launches. It doesn't take as long as people might think.
Protecting patient information remains a non-negotiable priority.
When asked about data access, Pan clarified that the medical data obtained legally through Shanghai's public data platform is used selectively for each insurer's specific product needs, following the principle of legitimate, necessary, and minimal use. Every request for data must go through the ethics committee for approval. After approval, the big data center retrieves the data, and the Shanghai Data Group performs initial processing—including blockchain recording and anonymization—before handing it off to the institute for analysis. The raw data never leaves the secure domain of the big data center.
Pan described the ethics committee's review process, noting that it draws from a pool of experts in fields like ethics, clinical medicine, and law, as well as representatives from social organizations such as the trade union. Committee members are randomly selected for each meeting to ensure independent judgment. Their review focuses on three main areas: whether the data use is appropriate and secure, whether the proposed development plan is scientifically sound, and whether the purpose and outcome align with ethical standards without harming minority groups. The initial reviews were intense, with applications facing multiple rejections.
On the key disagreements, Pan explained that the biggest hurdles were about whether individual informed consent was needed for each use of patient data and the risk of "precise exclusion" of people with pre-existing conditions. Medical experts initially insisted that every data use requires separate consent, which is ethically correct but nearly impossible for commercial insurance, as you can't go back and get signatures from millions of patients over five years for a pricing model. Ethically, there was also concern that more precise actuarial models could allow insurers to precisely identify and exclude high-risk individuals, turning reasonable risk control into systemic discrimination. After extensive discussions, a consensus emerged. The breakthrough was recognizing that anonymized data research presents minimal risk and does not adversely affect participants, so individual consent is not required. To address the exclusion concern, safeguards were introduced, such as prohibiting analysis targeting specific occupations or regions and requiring that any output group be large enough to prevent identification of individuals. By opening data, we move away from excluding all pre-existing conditions to offering specific recurrence insurance, bringing coverage to people who were previously left out. For example, patients with 10mm lung nodules or stage three breast cancer—who historically couldn't find suitable commercial health insurance—can now be covered. Of course, some conditions, like advanced cancer or old age, remain difficult to insure, and we acknowledge that. But the direction is promising.
When discussing how to reach patients, Pan acknowledged that selling cancer recurrence insurance is even harder than pricing it. The target audience is people who have had cancer and have since left the medical system, making them difficult to contact. For the Shen Ai Bao policy for specific gynecological cancers, the institute has taken two steps. First, they've made promotional materials available in hospitals, placing them in designated, legal areas, respecting each hospital's right to participate. After a year of piloting, the results have been positive. Second, they established a standardized case management system for tumor patients—an exclusive service offered by the institute. This involves training case managers to follow up with cancer patients in hospitals. However, there's a strict boundary: sales staff cannot sell in hospitals, and medical staff or case managers are not allowed to promote insurance products. The practical approach has been to have insurance staff work in a dedicated area within the hospital. When patients see the information and have questions, they can consult with a case manager who can then direct them to the insurance staff. This allows the insurer to reach patients in a compliant way without any in-hospital selling. Still, moving from access to actual sales and successful claims takes time. There's a clear gap in the number of in-person sales personnel, particularly those certified to work in hospitals, and patient awareness and acceptance of these products are still limited. Reaching a wider audience won't happen overnight; it demands progress on multiple fronts including access, personnel, and patient education.
On the question of how to ensure patients at various disease stages find suitable products, Pan emphasized that the goal is to build a multi-tiered system, not rely on one product for everything. Taking lung disease as an example, the approach is a staged and progressive coverage chain. For patients with nodules, insurers in Shanghai have launched tailored products that cover those with 8-10mm nodules and even pre-cancerous conditions or early-stage adenocarcinoma. If the disease advances to certain stages of lung cancer, coverage shifts to appropriate recurrence insurance. Unfortunately, stage four patients still lack access to most recurrence policies. In summary, we can't expect a single policy to cover all scenarios. Instead, we need to offer the right options to patients at different stages. This system works best when potential policyholders are numerous, medical data is comprehensive, and the insurance leverage is meaningful. With a wider array of products on the market, patients at different stages can find coverage that fits their needs.
Regarding why chronic diseases have few specialized insurance options, Pan noted that developing policies for these conditions is harder than for cancer recurrence. This is due to several factors: the long duration of chronic disease, which makes complete follow-up cohorts rare; the complex progression paths with multiple complications that complicate risk modeling; the difficulty of intervening in lifestyle choices that heavily influence outcomes; and the need for continuous, cross-institutional, long-term data. Currently, the institute is planning data cohorts for vascular and brain diseases to lay the groundwork for future chronic disease insurance, with cardiovascular and cerebrovascular conditions as a priority.
When asked what key lessons from Shanghai's data-sharing experience could be shared, Pan highlighted two core practices. First, establish a unified ethical review framework for data use at the top level, with a single translation platform for data processing that handles anonymization, blockchain registration, and security to speed up commercial application without compromising compliance. Second, achieve a closed loop of business value. Many regions have tried opening data but lack the necessary full-course management services. Shanghai's experience shows that insurance for pre-existing conditions needs standardized, customized follow-up services to balance risk and achieve a sustainable business model.
On whether policies developed using Shanghai's data could benefit patients elsewhere, Pan said that currently, products are not restricted by region, and patients from other provinces can purchase them. The product design often encourages policyholders to seek care in Shanghai, with benefits like travel allowances for medical visits and access to local treatment and management. However, since Shanghai's health data is specific to its population, healthcare standards, cost structure, and drug accessibility, directly applying its models to other regions would result in pricing inaccuracies. A more practical approach is for other provinces to use Shanghai as a reference and adjust for local data. In the long run, Shanghai could explore joint modeling across regions and platforms, building a network for data collaboration on insurance for pre-existing conditions nationwide.