The Sentiworld project: global mapping of sentinel surveillance networks in general practice

Epidemiologic disease surveillance, defined by the World Health Organisation (WHO) as “the systematic collection and use of epidemiologic information for the planning, implementation, and assessment of disease control” [1, 2], consists of data collection over a specific period of time and providing prompt feedback to healthcare administrators in order to adjust health policies accordingly.

In many public health systems, the major pillars of epidemiologic surveillance are mandatory disease reporting and sentinel surveillance networks [3]. Supplemental data are also collected from more modern tools that are used in tandem. These tools include administrative databases [3], drug sales data [4], and participatory syndromic surveillance [5].

Sentinel epidemiologic surveillance relies on dedicated healthcare investigators, often in general practice, who report real-time data on specific health indicators. The objective of this type of surveillance is the early detection of unusual patterns of disease activity, estimation of the burden of disease in the population, and comparison of disease trends with historical data using data that is consistently collected by dedicated reporters using set case definitions [1, 3, 6].

Differences between sentinel networks include the health indicators monitored, representation of populations covered, types of reporters, methodologies for data collection and analysis, case definitions, and frequencies of reporting. Although general guidelines and case definitions are issued by international public health organizations, such as the European Centre for Disease Control and Prevention (ECDC) [7] and the WHO [8, 9], local epidemiological and public health surveillance systems often adapt these guidelines according to their health care system’s specificities and public health objectives. Further, while the International Health Regulations 2005 is a legally binding agreement to combat global epidemics, they pertain mostly to mandatory disease reporting systems, and not to sentinel networks [10, 11]. Sentinel data comparison across geographic areas and data pooling requires adequate knowledge of the available sentinel networks as well as knowledge of the methodology by which data are collected. Understanding this heterogeneity is crucial for improved disease surveillance at higher geographical levels [12].

Several past projects have attempted to identify and describe the sentinel networks that exist at the European level. The first inventories of sentinel networks were conducted in 1987 and 1990 and resulted in a group of 15 European sentinel research networks of general practitioners (GP) called Eurosentinel [13]. Further studies in 2003 and 2006 sought to update this inventory on the European scale and further document the surveillance activities of these networks [14, 15]. Other studies have been carried out by European organizations and institutions, including an inventory carried out by the Euvac project in 2001 [16, 17] and by the ECDC and WHO [18, 19]. However, these surveyed only European networks, focusing mostly on influenza, and are all more than a decade old (with the exception of an up-to-date list in the FluNews project, but this data is restricted to only previously identified European networks [19]). Additionally, epidemiologic surveillance based on primary health care providers acting as sentinel investigators has significantly evolved over the past decades, particularly due to the advent of electronic medical records (EMR) and the advancements in automatic data collection based upon electronic record frameworks.

Inter-network cooperation has long been described as essential at the European level [14], especially for comprehensive flu surveillance [20]. An updated inventory on a global scale would expand upon past works and take a step toward this necessity. For this purpose, the Sentiworld project was launched in 2016 with the objective of creating an inventory of sentinel surveillance networks in general practice at a global scale and providing a public online platform that shares collected network data. Since this time, with the emergence of new global public health crises, such as the COVID-19 pandemic, the need for this cooperation is even more apparent. We envision this project to be a resource that allows for a greater understanding of the structures and operations of different sentinel networks, while facilitating interaction and scientific collaboration between networks, thus taking a step toward improving global disease surveillance. In this study, we present the Sentiworld project and the characteristics of the participating networks.

A descriptive study of sentinel surveillance and health research networks was first conducted in Europe and was then extended progressively to a global scale. This study was conducted using the resources of the French national sentinel surveillance network [21]. The study period was July 2016 to December 2019.

For inclusion in this study, networks had to be: (1) Sentinel-based surveillance networks, (2) composed, at least partially, of general practitioners, (3) monitoring a whole country or specific region of a country, (4) collecting health data on an ongoing or regular basis, and (5) having a goal of health surveillance or research.

The first step consisted of identifying potential surveillance networks in general practice. For this task, we relied initially on previous work on this subject [13‐15, 17, 18]. We then searched for research articles and surveillance reports from different countries on a global scale in order to identify other surveillance networks [22‐25]. We also contacted epidemiological surveillance experts in Europe and worldwide for assistance in identifying sentinel surveillance networks. These contacts included partners of the I-Move (Influenza-Monitoring of Vaccine Effectiveness) [26] and InfluenzaNet projects [27], as well as ECDC members. Finally, internet and literature research was conducted to identify sentinel surveillance networks, especially outside of Europe.

This work presents the characteristics of 48 general practitioner sentinel surveillance networks from around the world. To the best of our knowledge, this study is the first to take inventory of sentinel surveillance networks in general practice on a global scale.

Building on past inventories on the European scale, the Sentiworld inventory of networks has found novel networks in addition to those previously recorded [13‐19]. Many of these previously studied sentinel networks have establishment dates in the mid to late twentieth century, suggesting that these are durable structures that have been able to adapt to changing public health landscapes (Fig. 2). Furthermore, 22 networks were created in 2000 or later, including networks from countries that were not included in previous works. The creation of new sentinel networks both in nations where other such networks did and did not previously exist and the emergence of networks that are monitoring a more varied array of health indicators indicates meaningful, progressive change in the field of epidemiological surveillance.

Previous work by Deckers et al. in 2006 set a threshold of > 1% for the inclusion of general practitioners in sentinel networks [14]. Over two-thirds of included networks had a percentage of general practitioners participating in the network greater than or equal to 1%. Some networks went further and have profiled their sentinel practitioners in terms of regional and demographic representation. This has been noted to help improve network functionality, as gaps in coverage of the population are better and more specifically known [29].

The main and often exclusively monitored health indicator is influenza. Previous studies support this finding [14, 15, 18, 19]. However, our study also shows that networks monitor many other health indicators, including infectious diseases that cause acute health problems (outside of influenza), non-infectious (chronic) disease incidences, and other, non-disease indicators such as suicides.

Data collection for sentinel surveillance purposes has evolved over time. Networks have trended from paper forms to electronic resources like dedicated websites and software. In recent years, with the wider use of EMR, sentinel epidemiology has trended towards the use of direct, automatic data extraction from these records to be used in tandem or in place of traditional sentinel collection mediums [30, 31]. In theory, EMR systems make many improvements over traditional reporting methods, as they provide data on more patients in a longitudinal sense and on more indicators [30]. Access to EMR data is also more rapid and has been shown to reduce the risk errors in disease surveillance [32]. We have found from our own data that EMR data collection does in fact have a strong positive correlation with the number of indicators surveyed, the surveillance of non-infectious indicators, and non-infectious indicator surveillance. This is likely because of the larger scope of data contained within EMR. EMR surveillance data is also better standardized than traditional sentinel data and thus could be a solution to homologizing sentinel surveillance, given the diversity of networks included in the Sentiworld study. However, EMR data collection cannot be a fix-all solution to all gaps in sentinel surveillance nor in international collaboration, as EMR data collection is also complicated in places where many private companies control medical data with data systems, coding or diagnoses are unstandardized, or data that is unsuited for epidemiological modelling.

Our results show that sentinel networks, despite all existing in general practice, appear to be heterogeneous. It is thus crucial to understand the differences between nations in healthcare management, in the laws and practice of general medicine, and in population health habits [33] when engaging in global health collaborations. It is also necessary to work towards a better understanding of the differences between networks, including case definitions, when comparing and sharing data. As of this moment, limited collaboration between sentinel networks is taking place, even in Europe where sentinel networks are most plentiful and established, and even in the face of the COVID-19 pandemic. This can be attributed to a lack of data homogeneity (compared to mandatory reporting systems) as well as an even broader lack of comprehension of foreign sentinel surveillance and health systems. Despite the challenges posed by international collaboration, it is important that this cooperation exists in our globalising society in where it is possible to fly from one part of the world to another in less time than the incubation period for most infectious diseases [10]. This idea held true certainly at the start of the Sentiworld project in 2016 and does so even more in 2021, amidst the worst public health crisis in a century, COVID-19. In this time, we have seen unprecedented collaboration among healthcare professionals, researchers, and surveillance systems across borders and continents [34]. With sentinel systems necessary to supplement research and mandatory surveillance data in the context of infection management, policy decisions, and vaccine monitoring, we are again confronted with the barrier of heterogeneity between systems. Our cartography allows networks to learn more about each other, which bridges many gaps in information that previously existed and allows for potentially greater inter-network cooperation.

One of the major strengths of this study is that it is the first of its kind in many respects. First, it extends its area of study to a global scale by contacting 148 networks in 116 countries over 6 continents. Despite a low overall response rate of 43.2%, compared to other studies of the like, we contacted many more networks, and were able to include many outside of Europe. Additionally, because data was supplied by network administrators, they were more likely to be reliable and exact. Data sharing on our website also improves network visibility and improves cooperation among networks.

However, there are also limitations present in our study. Despite contacting many networks outside of Europe, the surveillance networks included in this study were largely European. Because we are based in France and limited linguistically to French and English, it was difficult to gather data from countries that were geographically distant, had vastly different health systems, or were less involved in research projects and, thus, more difficult to reach. Difficulty in finding accurate, up-to-date network contact information online often hampered efforts to expand our study as well. Additionally, even when contact information was found, data on the networks may have been scant, leading to doubt that even with a higher response rate, all of our 148 potential sentinel networks contacted would have fit our inclusion criteria. Given the anticipated health system differences, we attempted to use a standardized electronic survey, but realize that it was likely not adapted for all types of sentinel networks, especially those that differed greatly from our own or had more limited knowledge of English. Overcoming these linguistic, cultural, and technological challenges could have increased the number of included networks and broadened our geographic inclusion. Further, we were not able to engage in additional research to profile all physicians in our studied networks, which would have been interesting to add to this study. While not the subject of this study, further investigation could take place regarding the sensitivity of inter-network collaboration to detect early epidemic warning signals or a quantification of the value of cross-border collaboration or even international standardization of sentinel surveillance.

The findings of Sentiworld show that there is great potential for future work in the area of international collaboration. The heterogeneity of sentinel surveillance networks leads to questions about which systems are most effective and future studies could be focused on statistical associations among characteristics of sentinel networks in order to determine which methods of surveillance are most effective, at least in theory. Further, Sentiworld data and its website can be used as a tool for communication among sentinel networks. Understanding differences between sentinel networks could allow for easier sharing of data in future studies and when dealing with public health crises. In fact, correspondence that took place for the Sentiworld study later led to the formation of a small working group of eight sentinel surveillance networks in seven European countries in 2020. The group was later able to collaborate on early COVID-19 research on initial risk perception and sentiments of preparedness among primary care physicians regarding the COVID-19 pandemic [35].

The Sentiworld website [36] allows the public to consult the information pertaining to each network detailed in this article. Each network has a dedicated webpage that can be found on a world map on the website’s homepage, on a list of the networks, or by searching keywords using a search engine. The site is expected to grow and develop with the expansion of this project and with the refinement of its features. It will be updated on a yearly basis.