The low sensitivity of the mandatory notification system for bovine abortion limits the early detection of any potential introduction of brucellosis or other abortive diseases into France , with potentially important animal and health risks. In that context, our qualitative study investigated the barriers and difficulties that hinder the notification of a suspected case, and possible solutions or incentives to encourage farmers’ and veterinarians’ participation in brucellosis risk prevention. Although our study focused on the sensitive topic of fulfilling mandatory requirements, the confidentiality of the interviews warranted the trustworthiness of participants’ answers , as suggested by farmers admitting that they report none or only some of the detected abortions (Table 2). A review of data on abortion notifications enabled us to validate farmers’ assertions about their participation in the surveillance system.
Specificity of the qualitative approach
In contrast to a quantitative approach, commonly used in epidemiology to estimate the proportion of actors influenced by specific predetermined factors, the goals of our study were to identify which factors (including unknown factors) influence their decision and to understand why those factors are taken into consideration by field actors . The qualitative approach relies on "purposive sampling" to maximise diversity  by covering a broad spectrum of experiences and positions relative to a given phenomenon . In our study, we selected a wide range of participants with different herd characteristics and a variety of attitudes towards their duty to report abortions. We found that "theoretical saturation" was reached after ten interviews for farmers and six interviews for veterinarians, which means that no new information was raised in the last interviews . This finding is in agreement with other studies, for which saturation occurred within the first 12 interviews .
Purposive sampling and theoretical saturation ensured the robustness and trustworthiness of our study and enabled us to generalise the information provided by study participants  into four themes that may be used to interpret the decision-making process of other farmers and veterinarians or decision-making process concerning other clinical surveillance systems such as classical swine fever or foot and mouth disease . In addition, within the four themes, factors influencing the decision-making process may be transferable to other related topics , as suggested by the similarity between our findings and concepts developed in other studies on notification decisions or behaviour regarding vaccination [27, 28].
Farmers’ and veterinarians’ decision-making process for reporting abortions
The typology of farmers according to their risk perception and attitudes towards the abortion notification process underlines little diversity in the driving factors once these two considerations are taken into account [14, 29, 30]. Furthermore, the same themes appear for both farmers and veterinarians. This result suggests that despite profession-related differences (interests, knowledge, position in the surveillance network), these actors share common interests in participating in the surveillance system. This contrasts with their divergent views on other issues such as the implementation of vaccination and culling strategies to control animal epidemics, where farmers have more to lose than veterinarians .
The decision of farmers and veterinarians to report or not to report abortions is driven by three main considerations: the perceived risk of brucellosis (or other abortive diseases); an evaluation of the individual benefits and costs linked to the abortion notification and differential diagnosis; the socio-technical network to which the farmers and veterinarians belong. The first two concepts—risk perception and cost-benefit analysis—are the backbone of the Health Belief Model (HBM), which is a framework widely used to describe health-related actions .
Risk perception is essentially based on lay knowledge and "value and belief-oriented rationality" , especially for farmers who rely on their experience and regular observations of their cattle herd. We also noticed the role of experience for veterinarians, although their judgement depends primarily on their technical and scientific knowledge. In contrast to risk perception, cost-benefit analysis is essentially based on "purposive and instrumental rationality" . Farmers and veterinarians considered calculated reasons (including regulations and health aspects, economic and financial losses, technical and practical factors) and their own self-interest to guide their decision. The importance of some factors and thus the outcome of the cost-benefit analysis varies with the type of production. Indeed, the technical features of beef production compared to dairy cattle  reduce beef cattle farmers’ margin for manoeuvre, as they are bound by both practical and financial constraints. The grazing season increases their difficulties in detecting abortions and carrying out a serological analysis on cows, as they do not have the daily contact that dairy cattle farmers do with their cows. Moreover, in the event of an abortion, the cow is usually immediately culled and sent to the slaughterhouse, and farmers do not feel the need to notify the abortion.
Besides risk perception and cost-benefit analysis, there is a third consideration. The French veterinary services, GDS, GTV, sanitary veterinarians and farmers are all linked through a social and technical network . The mandatory abortion notification surveillance system defines what should be done by each actor to facilitate abortion notification and ensure early detection of a brucellosis outbreak. However, even if farmers are theoretically best placed to report abortions and benefit the most from the surveillance system, it does not make sense for most of them to be required to report abortion as a public duty. This reluctance to report abortions may be compared to the reluctance of humans to vaccinate themselves : in both cases, the decision-making process of most people is not driven by the need to fulfil a public duty but by self-interest. Moreover, some farmers decide not to report an abortion because other farmers do not (peer influence), considering they have performed the necessary cost-benefit analysis to make a wise choice [27, 36].
Perception of brucellosis risk and preventive measures
Like most clinical surveillance systems, the mandatory abortion notification system gives priority to sensitivity rather than specificity, given the sensitive definition of a suspected case. However, farmers and veterinarians do not feel the need to report every abortion to detect a brucellosis outbreak early, as most of them perceive the risk of brucellosis as low. The epidemiological situation regarding brucellosis in France has dramatically changed since the implementation of the surveillance system in 1965, when brucellosis was enzootic. Today, in a brucellosis-free context, farmers are more concerned about abortive enzootic diseases such as bovine viral diarrhoea or Q fever than brucellosis. Furthermore, as it has been cited elsewhere , high levels of confidence reduce perceived risks. In our case, farmers feel confident in their ability to detect a brucellosis case in their herd or in the effectiveness of the annual serological surveillance system for dairy and beef cattle herds. Abortion is a non-specific clinical sign that may or may not be caused by an infectious disease. Thus, reporting the abortion, illness, or death of a single animal does not make sense to them if the objective is to detect an infectious disease .
These pragmatic feelings are supported to some extent by the results of a simulation study which showed that abortion notification (with a reporting rate varying from 20% to 80%) tends to be a less effective method for the early detection of brucellosis than annual bulk milk surveillance . However, some infected animals may not be identified by serological tests as infected cows may be serologically negative for several months until they give birth to a calf or abort [7, 39]. Moreover, brucellosis infection may spread slowly, depending on how the disease is introduced into the herd (from an infected animal or by indirect transmission) and herd management factors, which may influence the amount of contact the animals have with each other, such that only a few animals in the herd may become seropositive or abort, even though brucellosis is a contagious disease [5, 40].
Prospects for improving brucellosis risk prevention
The mandatory notification of each abortion is viewed by most farmers as an externally imposed tool for an externally imposed issue (the need to detect a brucellosis outbreak early) that they are not actually worried about . However, the multiple patterns of brucellosis transmission and dissemination underline the importance of monitoring abortions. In addition, this system is useful for the surveillance of other diseases causing bovine abortion such as Q fever or Rift Valley Fever. Efforts by veterinary services and the GDS to increase farmers’ and veterinarians’ awareness about the need to report abortions has not reduced the probability of under-reporting, which has remained stable since 2009 . As underlined by psychologists, messages to change people’s attitude require three factors: a credible communicator, a high level of "similarity" between the audience and communicator, and finally both the message and communicator must be perceived as trustworthy [42, 43]. The "traditional" communication strategy is based on rational arguments, but requires the actors’ basic motivation and an interest in the topic . Our findings suggest that the difficulties in increasing abortion notification rates may be related to the low level of trust in the communicators (veterinary services and GDS) and the low level of concern about abortions as long as they remain sporadic. Moreover, we found that both farmers and veterinarians do not agree with the decision to report every single abortion, which they consider irrelevant and of no real use . Thus, enhancing risk communication requires not only relying on the "rational choice" model of decision but also taking into account the actors' social context as well as their values, beliefs, and how much they trust the different sources of information .
Our analysis suggests that putting a great deal of effort into increasing the number of reported abortions without updating the surveillance procedure would be inefficient. First, brucellosis infection causes mid-term or late abortions, which occur between five months and the end of pregnancy . Second, the current situation regarding brucellosis in France suggests that the human and financial resources required to meet national and international regulations, i.e. testing about 18.8% of cows (which suffered mid-term or late abortions, or had a calf that died within 48 hours of birth, [20, 47]), seem disproportionate in comparison with the number of brucellosis outbreaks that would be detected promptly (one outbreak in 2003 and two cases in 2012). Third, the performance of such a system will always rely on farmers’ and veterinarians’ willingness to participate in it, and these field actors are not prone to report cows returning to heat or the death of newborn calves. Therefore, there is a real need to revise the surveillance procedure so that the system is more efficient. Considering OIE requirements , these changes need to be considered on an international scale. An initial suggestion would be to revise the definition of abortion by excluding stillborn and newborn calves. Furthermore, a brucellosis analysis could be requested only beyond a certain frequency threshold of abortions within the herd. Farmers could be requested to register each abortion themselves (in their own record or in a national record system) but call their veterinarian only in the event of recurrent abortions (for example, two abortions or more within a month). Besides, cows suspected to have aborted could be sampled after a certain delay should the abortion not be detected early enough, or if practical issues hinder the immediate sampling of the cow.
By helping farmers and veterinarians identify the cause of abortion, the differential diagnosis with respect to other abortive enzootic diseases was supposed to improve their willingness to contribute to the surveillance system and participate in brucellosis risk prevention. However, it has been estimated that the cause was only diagnosed for about one third of the biological submissions . The difficulties in identifying the abortive pathogen arise from the wide range of potential pathogens, the ubiquity of pathogens such as Q fever or salmonellosis, and veterinarians’ lack of knowledge about the type of samples to collect for the analyses. Therefore, a differential diagnosis protocol has been recently drawn up nationally in order to help veterinarians with the sampling process and identification of the cause of the abortion. Improving the coordination of veterinarians by official bodies providing technical support, training and information on the results of the differential diagnosis protocol (in addition to the results from the mandatory surveillance system) is also expected to increase participation in the surveillance system .
These prospects for improving brucellosis risk prevention could be useful for improving other clinical surveillance systems. Indeed, while some factors related to the reporting bias are specific to brucellosis or cattle production, most of them may influence the decision-making process for other diseases. Many exotic diseases with clinical surveillance are seen as low-risk, and the decision-making process is likely to be driven by regulations, health, economic, financial and technical factors. Thus, we expect that farmers and veterinarians would be more likely to report avian influenza or classical swine fever suspicions in the event of high mortality or morbidity than if only a few animals fall ill or die (which is the case if the outbreak is due to low pathogenic AI virus or pestivirus) . Enhancing risk communication, developing differential diagnosis and revising the surveillance procedure to find the best compromise between sensitivity, expenditure and acceptability are ways of improving clinical surveillance systems that should be taken into consideration.
Nonetheless, performance will always rely on farmers’ and veterinarians’ participation, and consequently under-reporting will remain the major limitation of such systems. In this context, it might be useful to develop other surveillance procedures, such as syndromic surveillance. As regards brucellosis, it might be useful to develop an indicator to identify the occurrence of abortions in cattle by using information from additional sources, such as the dates of artificial insemination or calving intervals.