The results presented below are the outputs specified and validated after step 5 presented upon (Fig. 2).
Description of EI systems and activities
Fifteen epidemiologists from PH and AH agencies in five countries and the ECDC were interviewed during 11 preliminary individual or collective interviews. Among them, 12 were involved for the second round of interviews and 16 other epidemiologists from these networks were selected according to their position for the 28 semi structured interviews (Fig. 1). Five additional practitioners participated in the workshops and HPAI meetings. Table 1 presents the EI activities executed by the 28 interviewees (14 women and 14 men): 61% of them worked in a PH agency, 39% in an AH institution. The average age of the interviewees was 46,6 years and the seniority in the position of 9 years, with higher averages in age and seniority for men than for women (Table 2).
Table 1
Summary of the EI activities of the interviewees
| Country | Sector | EI activities of the interviewees |
| Europe | PH | Modelling of trends and non-pharmaceutical interventions to inform risk assessment (IBS) |
| Detection and assessment of infectious and X threats (EBS) |
| Detection and assessment of infectious and X threats (EBS) |
| Detection and assessment of infectious and X threats (EBS) |
| Italy | PH | Monitoring and risk assessment in entomology (IBS) |
| Surveillance of enteric pathogens (IBS) |
| Detection and risk assessment of infectious threats (national EI, and IBS for VBD) |
| AH | Surveillance in animal health and food safety (IBS mainly) |
| Monitoring program about bacterial diseases, AMR, biosecurity and animal welfare (IBS) |
| Monitoring of West Nile disease virus and Usutu virus (IBS) |
| France | PH | Epidemiologic surveillance of tropical diseases: detection and risk assessment of introduction (IBS) |
| Management of surveillance systems for arboviroses (IBS) |
| Management of surveillance systems for AMR (IBS) |
| Surveillance of hepatitis A, E, tularaemia + coordination of TBD surveillance (IBS) |
| AH | Coordination of the IBS groups and the EI group of the platform (EBS and IBS) |
| Editor of the national epidemiological bulletin (AH and food security); international EI (EBS) |
| Detection and assessment of infectious and X threats EBS (international EI)) |
| Detection of OH threats (Low signals monitoring) (EBS) |
| Finland | PH | Preparedness, Response, risk assessment (IBS) |
| Infectious disease consultant dedicated to the Hotline (EBS) |
| Investigations on bacteria and Food Borne Pathogens (IBS) |
| AH | Modelling risk assessment for animal diseases (IBS) |
| Serbia | PH | Surveillance, Detection and Reporting (IBS) |
| Detection and assessment of VBD (IBS) |
| Spain | AH | Surveillance, detection, management and Elaboration of guidelines (IBS) |
| Management and implementation of European law in disease surveillance and VH (IBS) |
| Implementation of program and reporting on zoonosis and AMR for EC (IBS) |
| PH | Coordination and Management of information system (IBS) |
Table 2
Age, Gender and seniority of the interviewees
| Gender | Number | Average age | Average seniority |
| Female | 14 | 43.8 | 6.9 |
| Male | 14 | 48.7 | 11.1 |
| Total | 28 | 46.4 | 9.0 |
The organization of EI activities (Supplementary file 3) varied in terms of mandate, centralization, human resources and data sources from one institution to another, and between EI teams. EID and biothreats detection and surveillance was the core mandate and the mandate concerning bioterrorism was sometimes delegated to a specific center. AMR was sometimes addressed in different EI networks than the central one (not addressed by the platform ESA (France, AH) and addressed by different agencies (France, PH)). At national level, IBS, based on mandatory notification of diseases, sentinel- and syndromic surveillance had a prominent role in early detection and surveillance. Most of the human resources of national agencies were dedicated to national IBS and very few to national EBS, even less for international EBS. The larger national group dedicated to Epidemic Intelligence (Italy PH) mobilized their 20 officers for national EBS by rotation and their main occupation was IBS. In Spain, international EBS was specifically externalized to a public-private partner. When international EBS was not part of the mandate of the team, it was often implemented in a non-formal manner, according to sanitary situation, on top of the IBS activities (France PH, Serbia, Italy AH).
National Health Information Systems (HIS) were mainly based on heterogeneous non interoperable institution-specific databases, more interoperable dataflows from laboratories and sometimes on shared multi-level repositories from surveillance networks (Spain, Italy). The shared online national platforms (multi-institutions, multi-level and multi-sectoral) were recent or under development and do not concern all the diseases. Some specific challenges were linked to the regionalization of Spain and Italy, and the related autonomy in the organization of disease surveillance.
The use of international platforms for reporting (ECDC (TESSy), EFSA, WHO, ADNS), sharing of events (EWRS/EPIS, WHO, WAHIS) and accessing international validated data (notifications and website of WHO, ADNS, OIE, ECDC) was generalized for the 4 European countries and Serbia also as an ENC.
The ECDC is an agency proposing to European countries the outputs of its EI activities or its support to develop their own EI service. The ECDC, ISS from Italy (for PH) and ESA platform from France (for AH) had dedicated teams implementing EBS and standardized procedures for EBS. The professional networks were complex for the daily work of detection and validation and included informal relationships on top of formal collaborations.
The EI team of ECDC was the only centralized team with a rotation of EI officers and a wide network of disease specialists in situ for the daily round table. They also used various EI data sources: official sources at multiple scales on top of their in-house platforms and databases, and many informal sources such as aggregators, blogs and social media.
The Italian EI group (rotation of 20 part-time agents for national EBS and two for international monitoring) was decentralized and used only two EBS tools for national EBS and one additional tool for international monitoring. The focus on only two tools was possible due to the efficiency of the Italian language as a filter to collect national information and the strong technical support of the European Commission Joint Research Centre.
The French AH platform of epidemiological surveillance also had a decentralized group and used a limited number of EI tools like the Italian PH EI group. The three dedicated groups used Standard Operative Procedure (SOPs) for EBS. ECDC has a prominent role to produce and share new SOPs.
In Spain (AH), national and international EBS was implemented without SOP by a network of officers of the ministry and epidemiologists from a private-public company, to complete IBS (ADIS and national databases) for some selected transboundary animal diseases by periodical monitoring of the situation in ENC from various Web sources (aggregators (HealthMap, ProMed), national and international public medias, official websites (WHOAH sites, WHO, etc..).
Specific EBS dataflows based in the notifications of medical practitioners were implemented at national level: a hotline for healthcare professionals to notify suspicions of EID in Finland, the system RYMY to notify FWBD events (Finland), and an online tool e-SIN to notify health-care related infections (France). These data sources were confidential and some tools were not numeric (as the hotline). In Serbia, systematic media monitoring concerned rumors and PH communication.
The multi sectoral collaborations (between PH, AH and wild fauna stakeholders) were often implemented for disease specific activities in an ad hoc way (for food safety, some VBDs and AMR), without formal One Health SOPs. The formalization of the intersectoral collaborations was facilitated in some countries by the institutional integration of AH institutions under the authority of the ministry of Health, like in Italy and in Finland (Table 3) or through national plans (against AMR, for communication about zoonosis, for coordination of alerts in Spain) or agreements between institutions involved in thematic and transversal working groups (France AH).
Main difficulties encountered by EI practitioners
We collected in the interviews 58 difficulties, defined as statements made by the interviewees concerning tasks seen as “difficult to implement” or statements corresponding to unfulfilled expectations. We grouped them into 6 categories (Table 3): the time-consuming tasks (17/58 difficulties, mentioned by 61% of our interviewees), the technical barriers (13/18, mentioned by 46% of our interviewees), the lack of quality (8/58, mentioned by 29% of our interviewees), the lack of collaboration (8/58), the methodological issues (8/58) and the lack of timeliness (4/58, mentioned by 14% of our interviewees). The most frequent difficulties in the interviews were related to different steps of the EI activities: data collection (14/58 difficulties, mentioned by 50% of our interviewees), their processing (9/58, mentioned by 32% of our interviewees), their analysis (15/58, mentioned by 54% of our interviewees), as well as the strategy to organize surveillance (10/58, mentioned by 36% of our interviewees) (Table 3). Many EI practitioners expressed being overwhelmed by the time-consuming data collection, processing, data sharing and reporting activities (14/58, mentioned by 50% of our interviewees), which left them little time for the actual data analysis and interpretation. Their first mandate was to produce consolidated data, but they were also in charge of producing risk assessments. Many difficulties concerned the analysis, because of the poor quality or lack of data, a lack of dedicated personnel with sufficient training, a lack of coordination to review strategic issues, and a lack of knowledge, know-how and tools for implementing the One Health approach.
Table 3
Occurrence of difficulties by category of difficulty, for each EI step (on a total of 58 difficulties identified in 28 interviews)
| Collaboration: difficulties linked to the insufficient or absence of work relations between different agencies, or stakeholders involved in surveillance activities |
| Quality: difficulties linked to the insufficient value of the output/ result of the activity (as expected by the stakeholder) |
| Technical barrier: lack of capacity linked to a tool, data or any technical aspect. |
| Methodology: difficulties linked to a lack of know-how or consensus concerning methods and tools. |
| Time-consuming: constraint linked to the time needed to implement the task. |
| Timeliness: in a broad meaning, the difficulties linked to the capacity to address the task at the accurate moment (according to the user point of view). |
| | Type of difficulties | |
| Tasks | Collaboration | Methodology | Quality | Technical barrier | Time-consuming | Timeliness | Total |
| Strategy | 5 | 5 | 0 | 0 | 0 | 0 | 10 |
| Collection | 0 | 1 | 1 | 3 | 7 | 2 | 14 |
| Processing | 0 | 0 | 1 | 4 | 4 | 0 | 9 |
| Data sharing | 3 | 0 | 0 | 3 | 1 | 0 | 7 |
| Analysis | 0 | 2 | 6 | 3 | 3 | 1 | 15 |
| Reporting | 0 | 0 | 0 | 0 | 2 | 1 | 3 |
| Total | 8 | 8 | 8 | 13 | 17 | 4 | 58 |
Table based on data from Supplementary file 2.
Data access, processing and validation
The difficulty of accessing good-quality data in a timely manner was a major constraint for national epidemiological data collection, early warning, and long-term international monitoring. It has been highlighted by the COVID-19 crisis, which tested the limits of the EI system.
“Now, after the COVID-19 crisis […] we understood that [most of] the failures that we had in some areas of the emergency response […] were due to the poor data availability.” (Italy P2 PH)
Access to data was divided into several issues, which were all sources of practical difficulties: the knowledge of the existing data sources; the heterogeneity of the data formats; and economic, legal, and political roadblocks. All affected the timeliness of the access and the quality of the data.
Firstly, the quality of the data linked to the deficit of dedicated epidemiological databases beyond punctual studies was also frequently mentioned as a source of difficulty. This was especially salient in EI decentralized systems.
“The problem is that we have very low-quality data from the peripheral labs because they use different systems. They don't really agree on the list of antimicrobials that should be tested because they have diagnostic needs.” (Italy, P2, PH)
Trustful sources of validated data describing local sanitary situations were also missing in some countries (European Neighboring Countries (ENC) in central Europe and north Africa), and worldwide regarding VBDs (e.g., Zika, chikungunya, dengue and malaria). The efforts of data collection were perceived as too time-consuming and were a constraint to improve the travel medicine by more precise risk assessments by location worldwide.
Countries also expressed the challenges in implementing better assessment of the risk of introduction of EID, in relation to additional dataflows that were missing or time-consuming to collect and merge. Concerning PH, the sources for immunization, medication, travel destinations, and risk behaviors came from punctual studies that were not deemed as sufficient. For animal health, the information about legal and illegal movements were collected through time-consuming requests to the customs or researchers and there was no centralized source of composition of food products.
Secondly, EI practitioners had to consult a high number of data sources and the knowledge of the best sources was sometimes person-dependent. The data sources of the EI systems were highly heterogeneous in the responding countries: the structured sources, in particular the mandatory notification reports, still play a major role for early warning. The EI team of ECDC stated that they have to consult a high number of structured and unstructured sources, whereas the national officers made strategic choices to reduce the number of unstructured sources in relation to the workload.
EI practitioners had to manually collect up-to-date diseases data and covariates among many scattered sources. This task was especially important for VBDs, FWBDs and AMR in PH (national and international scales), but also for epizootics in AH at the international scale. The international primary sources of the most up-to-date validated data were scattered. Even in Europe, their identification was based on the practitioner's experience and there was no alert when a new report or notification of new data became available. This fragmentation and heterogeneity of data producers was also salient at the national level between ministries, medical services, and between autonomous regions. This situation led to data sources being left unknown, or known but unused.
“I am sure that there are lots of things that are put officially on certain official ministry websites, with a delay or not. Official data, things that could be useful for us to evaluate epidemiological situations. We do not have access to them because we do not know about them. Then […] we lose time, so we do not persist.” (France, AH, P3)
The COVID-19 pandemic increased the workload by overwhelming some notification tools with a lack of human moderation. For example, the signals for other infectious diseases than Covid-19 were overshadowed by the increased flow of emails from EWRS, the tool dedicated to mutual notification for early warning between EU countries.
“The early warning and response system has a big problem actually, because of COVID, the amount of information has increased […] immensely. […] If every institute of every country in the EU is responding every day, it means like 20 or 30 replies for a single question and there are several discussions going on there. So, this EWRS notifications, they are filling my email box, right now. There might be something interesting, some Listeriosis for instance, but it is difficult to find because most of the notifications are about COVID.” (Finland, PH, P2)
Thirdly, the formats of the accessible data were highly heterogeneous. When databases were available, they were rarely interoperable. When they were not downloadable, or not yet, data could be found in published reports and bulletins, in even less standardized formats. For instance, there was a lack of dedicated databases for epidemiology, in particular for PH, and the numeric entry of epidemiological data by the medical practitioners was partial. In AH, the sources for legal and illegal animal movements and composition of animal by-products were different (customs, researchers, scattered non structured sources) and their extraction was human-based and very time-consuming.
“They do not have a standardized system able to collect information for all human diseases, just for some human diseases. And also, many epidemiological investigations, I mean, investigation data that are quite important, they do not have a complete and informatized system.” (Italy, AH, P3)
Fourthly, economic, legal, and political roadblocks hampered data access. Databases were sometimes expensive – for instance, the International Air Transport Association (IATA) data of flights and passengers. Dealing with the regulations for data protection and licensing agreement required a specific know-how. For instance, in Italy, the databases used by various institutions (laboratories, hospitals, private doctors), sectors, and administrative units could not be merged easily in 2020 in order to respect the new data protection laws. Further, the state authorities sometimes chose not to publish their data or to delay their publication.
“It is not always possible to find the original source. Especially in some countries, when the ministry of Health is not as transparent as one would want to, or the website is not up to date. In some situations, you need to be more imaginative, also check with other stakeholders, e.g. MSF, Oxfam, WHO regional country offices. This is the next step. So multiple non-direct (i.e. not original) sources would need to be checked.” (ECDC, P2)
The differences in timeliness of data flows were a barrier to their combined analysis needed by stakeholders in order to implement control measures. The scientific literature was considered as an important source of quality but outdated data.
“If you want to be effective in prevention, you have to start right away, when we detect the problem. [You need] tools that [are] very reactive […]. Today, we are already happy to have an annual temporality of about n + 1 […]. [But] the PRIMO mission with the city labs, they are able to give, every month, to the labs the analysis that they make of [their data]. If you want to take data from people, you have to offer them something, otherwise they are not interested!” (France, PH, P3)
The timeliness of reporting was often considered not satisfying. The most prominent European central database repository, TESSy, provided data often seen as outdated for the EI practitioners. Concerning the international early warning at European level, the timeliness of IBS dataflows and notification to international agencies (ECDC, EFSA and WOAH) vary a lot according to the disease. For COVID-19, the health data notification became daily at the national level through dashboards. However, for other infectious threats like tick-borne diseases (TBDs), the delay of notification and the under declaration did not allow an early detection of events.
“This data is not transferred into TESSy straight away. So there is always a delay in reporting. And we have to look for information one by one (per member state). We do not do it often, because it is so time consuming, only when [it is] something big. […] “Most of the states, not all of them, publish on their websites weekly, and monthly bulletins with the numbers of cases reported for all these diseases. And that is the most up to date data that we can access.” (ECDC, P2)
For some countries with low resources, the most up-to-date sources for validated health data was considered to be their ministry websites or social media accounts, but the direct monitoring of social media (not collected by an aggregator) produced a very large amount of data. Facing this difficulty, many practitioners (France and Italy) preferred to monitor only a selection of specific accounts.
The validation of EBS signals was cited as a time-consuming step. When an international event was suspected, it needed to be validated by a committee of experts (in ECDC, Italy PH and France AH). They had the possibility to contact their networks of focal points (from ECDC) in Europe and ENC, reference laboratories, and disease specialists. They also used their personal networks like researchers or other professional stakeholders (NGOs). The validation was subject to discussion between peers in the EI institutions during daily or weekly briefings, and consensus could take some time to be reached when uncertainty was high. Lack of information exchanges between eastern European countries rendered the validation of information difficult between national agencies.
The validation of IBS data at local level was also time-consuming when the number of cases was high and the sources had to be cross-checked like it was the case for COVID-19 in 2020, that mobilized human resources from other priorities.
“In addition to that, when there is a mandatory declaration, there is a whole validation work to be done which is generally done by the regional health agencies. And this validation works, if there are many, many cases, in fact, hundreds of people would be needed on a permanent basis…” (France, PH, P4)
Respondents using IBS data stated they spent time validating them before the risk assessment and field investigation to be sure to have reliable data even if a validation was not their personal duty. The mandatory notification tools and process were not adapted to diseases with many cases (like COVID or Dengue) and lead to additional work to get exhaustive datasets.
“We know that mandatory reporting does not work when there are too many cases to report.” (France, PH, P3)
Data analysis
Many EI officers highlighted a lack of human, methodological and organizational resources to perform sophisticated cross-sectoral analysis about AMR, VBDs and FWBDs. They indicated a lack of consensus about the best intersectoral determinants for the emergence and spread of zoonoses. Missing knowledge in epidemiology (e.g., accurate inputs about immunity interactions) or missing shared resources (e.g. more complete genomics reference library to implement better risk assessments in a One Health approach or access to open-source tools) were cited as important barriers.
Sharing epidemiological datasets between the sectors often required a request of special access that was hampered by data protection issues and by the difficulty to manage the very heterogeneous quality of data. Beyond health data sharing, the lack of standardized covariates sharing between the sectors and countries was a constraint to implement standardized intersectoral risk assessments. Building an intersectoral platform was described as challenging and could not cover all diseases.
“A common database is missing. What happens with Veterinarians is not up to them, it is up to the system. That is established in that way. If they don't have a project and sometimes financial support provided, they cannot give us the data because of Ministry of agriculture” (Serbia, PH, P1)
During the COVID-19 pandemic, the monitoring and comparison of control measures has been challenging, although the countries have timely shared their data through dashboards, and relied on ECDC and research institutes for the analysis. The “human covariates” were deemed not sufficiently standardized (such as movements and implementation of health measures).
The collaboration between the animal, food, environment and human sectors was considered insufficient by most of the respondents, because the coordination from data collection up to analysis was lacking. Multisectoral collaboration was program-specific and related to specific diseases – mainly VBDs or formal programs in AMR, rather than implemented to prepare for bio-terrorism.
“We have a lot of informal contact with [public health] authorities, but formally, and there is no constant interexchange. So […] it is difficult for us to get all the information from the human side, and sometimes it is difficult for them to get all the information from our animal health side. I mean, One Health is a challenge for every country at this moment, because there is a lot of inertia from the past opposite of this approach and it is difficult to fight against inertia.” (Spain, AH, P1)
The difficulties to implement multi-sectoral collaborations underlined by practitioners were explained by the different concerns in prioritization of pathogens for human and animal health, the structural constraint of the institutions and also the lack of knowledge about intersectoral relations of causality for emergences (AMR, FWBDs). This was a barrier to build or implement a One Health plan for AMR despite the political will.
“Few resources, poor understanding… I think of the problem and at the moment, poor integration between the different parts, the different stakeholders mentioned in that plan.” (Italy, PH, P2)
Many difficulties were related to methodology for implementing intersectoral analysis. Beyond the constraint for collecting and merging the data, some knowledge or skills to perform sophisticated analyses were sometimes missing at the scale of the team.
We noted an overlap between some mandates of surveillance, with a thin line between the descriptive analysis (situation assessment) and risk assessment for decision-making. The data production by multiple stakeholders can lead to duplication of efforts and difficulty of data merging without a strong coordination (e.g., indicators for AMR). There was also a challenge to produce analytical outputs adapted to decision-makers. In the time of pandemics (COVID-19 and HPAI), the production of new modeling outputs was externalized to researchers and specialists. They had to translate their research products into ready-to-use documents for decision-making and communication: taking criteria of decision-makers into account was necessary to warrant their effective use.
“We are asked not to carry out a risk assessment because it is not our role, but we are always obliged to do so when we assess an epidemiological situation, so we have done a minimum and this is a cursor that is difficult to place […] In the first stages of the work (writing notes for epidemiological bulletin), one may be limited in what one has to deal with or not to deal with.” (France, AH, P3).
Notification, reporting, and data sharing
The heterogeneity of reporting systems was one of the main constraints in countries having a high degree of decentralization, organized with autonomous regions and municipalities. In Spain, the lack of homogenization and interoperability was experienced through platforms for national alert network and national database of veterinary antibiotic prescriptions (Table 3) that allowed different interpretations over the regions. In addition, it was pointed out that large rounds of surveys organized though internal forums may be interpreted differently over the regions. Delay of reporting was also stated as a critical problem at national and subregional levels. When the IT system did not allow an automatic uploading of the data, more human resources were needed to fill the forms online.
Some technical difficulties were also related to the constraints about the interface for uploading and frequent changes in format required for official notification. For example, uploading to TESSy required large datasets to be split into smaller ones. Furthermore, the frequent changes in the format of data entries induced coordination costs throughout the surveillance systems. The divergences of reporting outputs according to the international system were also representing a difficulty for analysis.
“According to the ECDC system, you have to split your data set in parts with no more than 500 records. Otherwise, the system is not capable of uploading it. And I have thousands of records. […] [And] It changes quite often. And we have to adapt the system, but […] even a small change in terms of an additional field of data means a big change in the system. Because we have to explain to each lab what the new information means, the way it should be collected, the availability, if it is available or not at their own level. And it cannot be done on a one-year basis. If you want a new information now, you have to start asking that two years in advance at least” (Italy, PH, P2)
"It's really the [member] states that declare after everyone declares in their own way, regularly, in blocks. You have to get to know them. At the beginning, there were also problems between the declarations of the ADNS and the declarations made to the FAO because they were not always consistent. Less and less, there are problems of concordance between the declarations made to the FAO and the ADNS. Now, with the OIE [WOAH], this is not the best way to declare. But there is going to be a new system that will pool declarations between Europe and the OIE [WOAH]. So there will be no more discrepancies.” (France, AH, P3)
At the national level, the problem of reporting could result in a lack of confidence in the reliability of data, for data managers. This problem was also pinpointed in relation to the transparency regarding data coming from extra European countries that needed careful validation through different networks.
Variety of strategic objectives
Strategic stakes appear to be a barrier to analyze data or even to organize surveillance. There is a lack of collective review of EBS objectives and tools between EI practitioners in general and for AMR in particular. End users expressed the need to have a better prospective approach in order to build surveillance systems concerning new threats timelier (this expectation increased since the Zika epidemics). National officers also expressed the need to exchange experiences and procedures between countries about sentinel monitoring and COVID-19 surveillance.
Countries had different priorities with traditional mandatory diseases notification and different resources to complement IBS by EBS. They allocated resources to particular diseases depending on the level of risk and control strategy for their country. VBDs and AMR are important for all the European countries. For Italy, France and Serbia, it primarily concerned TBDs and the mosquitoes borne diseases (in particular West Nile Virus (WNV)). The interest for Aedes-borne diseases was linked to the changing distribution of the mosquito vector species. Italy gave importance to the close monitoring and diagnosis of WNV that have consequences for blood deferral and vector control. France put efforts on the risk of introduction and endemization of Aedes-borne diseases.
“It is more of a problem of staffing and objectives before SOPs. What do we want to see? What do we need to pay attention to? And when it comes to AMR in animals, it is even worse.” (ECDC, P1)
When vector control strategies were not sufficiently based on the results of the entomological monitoring, it decreased their direct usefulness and thus the quality of their implementation. The lack of cost/benefit analysis of these strategies was a barrier to the revision of these vector control strategies and the lack of perceived data utility a barrier to the quality of the vector data collection.
Professional logics and propositions for possible improvements
Reliance on networks for collective expertise
Mobilizing their professional network was described by EI practitioners as a core feature of their work. To carry out their missions, they thus relied on a large network of epidemiologists in their regional and local administrations, with the addition of private actors or NGOs. These relationships were instrumental in getting access to data from supranational organizations (mainly ECDC, WHO, FAO, WOAH, EFSA), collecting a variety of insights and feedback for risk analysis. In AH, multiple collaborations were made with academia, associations of hunters, key stakeholders for wildlife diseases, farmers, and birds protection associations that were part of national sentinel networks.
They also expressed the need for more peer networking between EI practitioners (such meetings as those organized by ECDC were seen as insufficient), to review their EI strategies, objectives, tools and procedures. A collective thinking about priority objectives and feasibility assessments was considered as useful to redefine the EBS objectives for AMR and thus adapt the EBS tools. In a parallel process, it was pointed out that the intersectoral list of pathogens produced recently by the project EU-JAMRAI could be reviewed and comparisons done between sectors. Moreover, an optimal set of data requirements for the AH sector (beyond food safety) and environmental sector should be identified.
“Could we have networks that can help analysts who may come from different backgrounds? To have a comparable evaluation of the information. That's the human component, that's always very present in the surveillance part of the media environment. But on the other hand, you've got the whole meta analysis of epidemic intelligence data.” (Italy, PH, P1)
Some interviewees pointed out that networking would improve data sharing and consultation among European peers. Epidemiologists performing forecasting for FWBDs would require inter-countries feedback as well as AH specialists regarding experiences of databases for the monitoring of antimicrobial use or for building a comprehensive reference genomics repository. In Spain, exchanges of experiences were also requested regarding a network of public and private veterinarians performing sentinel surveillance through IT application. Such networks should be strengthened at the European level, as well as the regional level – the Balkan region has been cited as a relevant perimeter to exchange official information related to outbreaks.
“We have gathered some groups of private veterinarians, to see which would be their interest in these applications, how we could focus our approach, these sentinel networks of veterinarians. We have started with dairy herds, the outcomes are still to be seen. […] It would be very nice a comparison assessment or report about how are these things done in other countries that are participating” (Spain, AH, P1)
More data integration and interoperability
Improving access to data, in the multiple dimensions (technical, legal, organizational and political) highlighted in 3.1.1., could be achieved by tools and institutional strategies that vary in terms of centralization and standardization.
First, improved digitization of the health data at the level of the private physicians and hospitals was still a major expectation. The laboratory dataflows were more digitized and timelier, but were not sufficient, since the diagnosis of some mandatory notifiable diseases relied on clinical or other complementary examinations, as well as using paper forms in reporting.
The multi-scale integration of data sources in national systems was requested for different epidemiological and administrative databases from the different medical services including pharmacovigilance. It could be solved by text mining tools applied to medico-administrative databases, the use of proxies, agreements to implement epidemiological objectives, or incentives services (by bringing quick analytical results as a feedback). National officers expected more integrated and comprehensive data flows through platforms for mandatory diseases (“notification portal”) or OH platform for enteric pathogens.
“In a hospital, you have a lot of computer systems: you have some in the laboratory, (...), you have bacteriology, but you also have hematology, that's it. There are radiology data, clinical data from the clinical services which are all computerized! And so all this can be put into perspective.” (France, PH, P4)
“The idea is that we need the administrative level of electronic system where any health facility that have a suspicion of any communicable diseases may fill the information about that person, that disease can be in the same time be reported to the district institute and also be visible on national level” (Serbia, PH, P1)
A recurrent hope was to have better data merging and integration in order to save time for analysis and to provide complete datasets for useful analysis. The harmonization and interoperability of infra-national databases could be achieved through a better scientific and institutional concertation of the needed epidemiological datasets (opinion papers) and also a better know-how to manage the new data protection laws. Some issues, AMR in particular, required better coordination between stakeholders to enable better indicators collecting in both sectors.
“The more useful data, the minimum amount of data that would be helpful for forecasting and the description of epidemiological situations at the national or international level and the way in which the constraints related to the regulation about data protection… Concerning the exchange of information, the interoperability of the database can be to some extent faced and some solution can be proposed.” (Italy, PH, P2)
Timeliness and comprehensiveness of dataflows has improved during the COVID-19 pandemic. For example, the real time access to data from sampling laboratories in France was considered to be extended to other diseases. In Italy, the EBS was helpful to bring context to the clusters.
“The example of the laboratory data that comes back on a daily basis, which makes it possible to know how many people have been tested for COVID and how many have tested positive from the sampling laboratories. This raises the question of whether or not to make it sustainable. […] in what form? Will it be anonymous or not? Will it be for all diseases, will it be only for COVID? […] And now we're going to try to set up the same thing for arboviroses as well, using the same pipelines we set up for the COVID, where for the COVID we have direct information from the sampling laboratories” (France, PH, P1)
At the international level also, a shared system would be welcomed by several respondents asking for One Health IT system to formalize the OH network and integrate climate and other environmental data and animal data. Indeed, EI practitioners requested a better identification of health determinants, in particular for VBDs, and risk thresholds validated by specialists. This identification could be partly done by literature review: it would help to better focus on the most important dataflows to merge. Moreover, it could help the agencies to get more standardized analysis and allow them to compare their risk assessments. These data would be gathered with a OH approach that was considered as more efficient in terms of decision making. A preference for an interoperable system with the existing platforms was elicited. The general trend was to reduce the number of platforms. A user-friendly access to already processed and standardized covariates would allow EI practitioners to save time and to perform more sophisticated intersectoral analysis.
“All the information in one single place. So, for example, if I see […] that they found anthrax in cattle somewhere in the EU, I want to understand, you know, if that is relevant or not. So, is it the first time? [If not], would it be transmitted to humans at some point, or some other zoonosis? […] There is a platform that deals with the zoonotic diseases in animals, another [one] that deals with […] the presence of the disease in humans, [and] the movement of people is on another platform. […] So, to assess which are the relevant [health determinants], we need to jump from one platform to another.” (ECDC, P2)
Building preselected datasets related to the sanitary context of worldwide locations would help EI officers to perform the risk assessment of the introduction of EIDs by travelers (PH). The information about travel locations, mobility and behaviors of travelers could be collected through app for travelers, social media mining or proxies (TripAdvisor, WTO) and additional dataflows (Eurogate project and covariates repositories).
“If you can help to centralize or to have processed or to store. I found an example: hantavirus in China, which was everywhere in the media last week. I want, for example, to know what is the latest outbreak of Hantavirus in China. Where is this outbreak? How many people live in the cities that are infected? How many people from Europe are traveling to and from this City; what animal reservoir is susceptible to be in this city in China? Will I have a mass gathering in this city? All this kind of information.” (ECDC, P1)
Concerning the risk assessment of introduction by international movements of animals and importation of animal by-products, a European repository could centralize the composition of food products. Another European repository could allow the queries of animal and by-products importation and (legal and illegal) movements between countries based on existing dataflows (Eurostat, TRACE database, Movebank and covariates for WNV wildlife etc).
It was pointed out that institutional collaboration took the form of externalization of some complex and time-consuming tasks, in particular to ECDC or research institutes. A methodological support to choose R libraries and predefined analytical tools would strengthen the capacities, as well as the continuous training.
This strategy was also often applied regarding EBS or modeling in crisis time. Many countries externalized this service to ECDC, OIE and EFSA and preferred to use their reports or get their notifications (Table 3). Specialized EBS tools for early warning, like social media analysis, were used mainly in dedicated EI teams (EI team of ECDC, Italian network of EI, VSI team of ESA platform), as they were seen as demanding in terms of know-how and continuity of service. In Italy, EBS was used to bring context to the outbreaks of COVID-19 (at the step of community spread). Thus, the EBS was seen as very flexible for the national EI dedicated teams. The standardization of the EBS tools and procedures and communication about their capacities and roles would help the institutions to adapt their strategies. ECDC had the mandate to provide training and support to the Member states.
“In Italy, the objective is to support IBS on potential epidemics, or ongoing epidemics in the country. So, we are also interested in information that normally is not very interesting because it is considered normal to have a certain number of infections in the country. In other countries, in other systems that look for what is unusual, what is different from what you expect. So, EBS is very flexible, you can choose your objectives, you can manage it according to what you need in the countries.” (Italy, PH, P1)
The COVID-19 crisis increased the need for EI practitioners to have more knowledge of social and political dimensions, for risk analysis as well as communication. Identified other data needs included rumors, context, beliefs and perceptions to monitor trust related to health measures by using sentiment analysis. Possible uses of existing and new tools (EpitweetR, for example) should be shared among Member States. Complementary tools should also be user-friendly with a manageable quantity of results and settings to monitor trends and visualization, and be standardized to avoid differences of interpretation between practitioners. In terms of relationships with decision-makers, preparedness during “peacetime” has been identified as a need. For instance, modeling was used in an extensive way to support decision-making, and a strengthening of the capacities of the modelers for a better use of their results by decision-makers has been elicited.
“You have your indicator-based, you have your EBS, you have a lot of possibilities that you can maintain or reuse, refit as we did for the EBS, so you can put them back when you need them, so it's good to have the instruments in place. And then you can modulate the monitoring response to better suit the situation!” (Italy, PH, P1)
Tools for automation à la carte
The need for a One Health methodological support was recurrent and covered various aspects: from the identification of the best drivers and thresholds of risk for zoonosis (in particular VBD) to the guidance to preselected open-source tools or R libraries, and the use of machine learning to better analyze environmental covariates, genomic data (to detect new strains), and trends for endemic diseases (causal inference and prediction).
A recurring message is that EI practitioners wanted to keep control of their workflow and their data, rather than relying on black boxes. They want to use their own data or/and to choose the data sources.
Practitioners were interested in methodological support to be able to choose predefined analytical tools through a decision tree for example. They expressed the need for tools that can help them accelerate specific steps on their workflows, depending on the specific characteristics (manual validation or not) of the situations at hand. Although the standardization of the risk assessment was an important objective, the practitioners preferred semi-automation that allows flexibility of the analyses and a research mindset: for example, a tool that could import and visualize processed covariates and merge to the series of points corresponding to the health/disease data owned by the users would be useful.
“There is always a physical analyst looking at things, so we have dropped the idea of an artificial intelligence doing everything. But at the same time, I think automation is enhancing the analyst's job in ways that can better organize the things that you can find, to allow the analyst to have a systematic visualization of things that are together very similar. So you can handle volumes of information more easily.” (Italy, PH, P1)
One core need of automation was to get more timely access to validated and up-to-date data. An alert system indicating the availability of new publications from the European laboratories and centers of reference or local bulletins is cited as a useful solution or a tool allowing queries of multiple sources (gathering of international and national validated health data, migration flows and customs for AH, or aggregation of cases published in local bulletins for PH) to assess sanitary situations. The centralization of the alerts and signals in the same formats is also important.
“Digested information, like for example, to take the same example, it's Hungary with a case of AI in a farm. So, I type in my software Hungary France, turkey AI. And there, it will take me all the information concerning Hungary from the OIE, from the ADNS, which will take me information on migration. So I won't have to go and look for articles on migration. It's going to look for articles on the website of the Hungarian Ministry of Agriculture and then translate them for me so that I can say here is the information that is available on the website of the Hungarian Ministry of Agriculture. So, I won't have to look for it myself. So here are the channels, the Hungarian professional channels, their website on poultry farming, here is what they have put as information and so everything comes to me already digested.” (France, AH, P3)
Machine learning was seen as useful when recurrent complex analyses are needed (to analyze clusters of bacterial strains, trends and outliers in endemic diseases). The production of risk assessments could possibly benefit from analytical tools merging covariates and health data linked to machine learning. For instance, in Italy, respondents would have liked to integrate environmental covariates and molecular typing of enteric bacterial pathogens to better understand the correlation between outbreaks of FWBDs and environment and get automatic alerts of new dynamics or abnormalities (causal inferences and prediction).
“The way the analysis that we are producing now is descriptive only. Some basic reporting concerning the number of isolates, the trends, the difference of isolation according to the different regions and labs and pathogens. We had some previous studies in which we apply some special analysis on the specific outbreaks and we try to include also some environmental correlates. And that [gave us] additional information concerning the dynamic, the ecology of some pathogens that were more linked to the environment than others. We would like to implement this kind of analysis.” (Italy, PH, P2)
User-friendly visualization tools were also mentioned as useful. The building of risk maps to support decision making implies to take into account the criteria of decision-makers (for HPAI, the accurate administrative resolution of the results should avoid stigmatization of farms and allow control measures). The other main use cases were analyzing social media and identifying trends, visualizing health data and their covariates, and monitoring outbreaks or endemic diseases in real time. Practitioners would like to have access to settings to choose the analysis period and thresholds of alerts and detection of outliers.
“It can be much more reworked to maybe even generate buzz level graphs and alerts like that, and that's where it could be improved! Now we have the raw information which can be tedious to rework manually, but if we set up with macros or with a way of reworking the data that is here in relation to graphs. I think we can go a little bit... We can lighten the information gathering.” (France, P4, AH)
“It would be fantastic if we could then apply some formulas that would detect / trigger an alert when there is an increase somewhere and in neighboring countries. E.g. a severe increase in scarlet fever, the month before they had 200 cases and the month before that only one case, it is very difficult for a human to [detect] this stuff, unless someone is checking proactively, this could go unnoted, [until] one of the neighboring countries notifies it.” (ECDC, P2)
Finally, practitioners expressed the need for a sustainable improvement of practices: they wanted to increase their skills at the scale of the team or institution and replace time-consuming practices in a sustainable way. The maintenance of tools and regular update of data flows were important issues identified to ensure efficient EI systems.
Additional quotes from interviews can be consulted in Supplementary File 4.