Routinely collected health data (hereafter routinely collected data) is a valuable resource containing large quantities and varieties of information. Routinely collected data are commonly defined as data collected for purposes other than research, such as health service delivery and disease monitoring [1]. Regional and national routine data collections may cover a large proportion of, or entire populations, over extended periods [2]. Common examples include data used to administer health services, disease registries, disease surveillance systems and electronic health records [3]. Such databases are increasingly considered as broad resources with great potential for clinical research, epidemiological studies and health system research [4, 5].

Primary data collection for research has become increasingly resource-intensive and leveraging routinely collected data for research is therefore an attractive and expanding research strategy [6]. Large volumes of data may be accessed in a highly cost-effective way, with many clinical trials, observational studies and health policy and system research around the world using routinely collected data to great effect [7–10]. The use of routinely collected data to assess randomized clinical trial outcomes has been recognized as a disruptive technology for participant recruitment and follow-up [11]. Study participants can be followed at a lower cost and for longer periods to identify long-term effects [12]. In addition, claims data has been used to facilitate pragmatic trials and to do trials embedded within health insurance systems [13, 14].

China has established multiple health databases over the past 2 decades with several examples of these data being used for clinical research—health insurance claims data have been used in a large prospective cohort study [15] and death surveillance data for the identification of fatal outcomes in a large-scale randomized controlled trial [16].

This review was conducted following an established framework—the Preferred Reporting Items for Systematic Reviews and Meta-analyses Extension for Scoping Reviews (PRISMA-ScR) [17, 18]. This review was registered on Open Science Framework (10.17605/OSF.IO/Q5CNB).

We identified 349 potentially eligible databases with most from government agency websites. We excluded 279 mostly because they did not address a specified disease area (n = 225), were not regularly updated (n = 45) or did not have nation-wide coverage (n = 7) (Figure 1). Some databases were ineligible for multiple reasons. There were 70 databases finally included (Supplementary Table S2).

The majority of accessible routinely collected data in China derive from databases established for the surveillance of communicable diseases and are under the custodianship of government agencies. Much fewer data relating to NCDs and injuries are collected and clearly accessible.

The nature of the routinely collected health data available in China reflects the evolution of public health priorities in the country. The earliest systems established in the 1950s [19] focussed on infectious diseases with rapid expansion after 2003 following the severe acute respiratory syndrome (SARS) epidemic. The Chinese government invested significant resources in infectious disease monitoring at this time, to strengthen established public health systems and implement multiple new surveillance programs [20]. These systems were designed primarily to enable better healthcare provision but also allowed for greatly enhanced research activity and reporting on infectious disease epidemiology [21].

The growth in databases related to NCDs has accelerated in the last decade with the launch of the 2009 health-care reform plan, prompting the development of health information systems focused on chronic conditions [22]. The expanding focus on databases recording information about NCDs in China is clearly warranted by the shift in disease burden from communicable to non-communicable diseases that has occurred [23, 24]. However, this review identified only limited (less than 30%) routinely collected databases related to NCDs. The NCD monitoring has to rely on the national health surveys conducted every few years [25]. The monitoring of NCD burdens and healthcare services will hence be limited and delayed. The key challenge of establishing routinely collected NCD data is the multiple source data custodianships making it hard to timely integrate the data. At the same time, the SARS-Coronavirus 2 pandemic has posed new challenges to infectious diseases surveillance systems [26] and there are likely to be multiple new communicable disease databases as a consequence of the pandemic. Novel surveillance methods based on space-time tracing technologies, syndromic surveillance systems and citywide pandemic monitoring platforms have been developed to combat the SARS-Coronavirus 2 pandemic in China, as they have in many other countries around the world [27, 28].

There were no Chinese electronic health record systems identified as eligible for inclusion in the review. This was mainly because all operate at a sub-national level with most patient data, and similarly, health insurance claims data held and managed by individual hospitals of regional administrative bodies. In other Asian countries such as Japan and Malaysia, integrated health information systems have been established by the Ministry of Health to link patient data from individual hospitals, mostly from public hospitals representing more than half of the inpatient admissions in the country [29]. The infrastructures of the existing information systems can serve as the cornerstones to achieving complete population-wide coverage in the future. In a few countries such as the UK, Canada and Australia, data from these systems have been widely used for research purposes, illustrating their enormous potential [30–32]. The National Health Service in the UK, for example, provides access to nation-wide data about primary care consultations and hospital admissions that have been used for studies of disease incidence [33], health service performance [34], medicine prescription patterns [35], as well as to collect outcomes for clinical trials of therapeutic interventions [36]. In regard to the latter, routinely collected data may save considerable resources compared to traditional data collection methods, and has been used in China for this purpose [15, 37], though issues with data quality and completeness have been identified [16]. A key challenge is that the investment in the curation of routinely collected data is typically not as high as might be made for a standalone research project, and the data may be more prone to both systemic and random errors as a consequence [38]. In addition, the infrastructure required to achieve timely data-sharing agreements with data custodians is limited in China, as it is elsewhere around the world [32], and there are significant investments required to implement the technical solutions and operating protocols required to enable data manipulation while ensuring data security.

Provided the large quantities of existing routinely collected databases with nation-wide coverage in China, there has been great potential for such data to be applied in large medical research. The China Kadoorie Biobank follows half a million participants by linking to the routinely collected health insurance claims data to identify disease occurrences [15]. A growing number of large cohort studies in China have used record linkage to routinely collected health data such as health insurance claims, health administrative data and mortality surveillance to follow up study participants over the long run [39–41]. The linkage to health insurance claims data has also been successfully applied in the large randomized controlled trial conducted in China [37]. Likewise, in other low and middle income countries with nation-wide routinely collected health datasets, there may be significant potential to apply these data in high quality medical research.