A New Tool to Assess the Economic Impact of Q Fever on Dairy Cattle Farms

Simple Summary To support decision-making in the field, a tool dedicated to Q fever for farmers and farm advisers was developed. The proposed, modified partial budgeting approach integrates a simplified yearly compartmental model and the main interactions between disorders linked to Q fever. The model concomitantly estimates the yearly burden of Q fever in herd prevaccination as well as the 3-year vaccination benefit. For herds with a moderate or high prevalence of Q fever prevaccination (>30%), a vaccination benefit was observed. The vaccine should then be seen as insurance in herds with low prevalence rates of Q fever prevaccination (≤20%). Abstract To support farmers in their decisions related to Q fever, a dedicated economic assessment tool is developed. The present work describes the calculator, its economic rationale, and the supporting assumptions. The calculator integrates a yearly compartmental model to represent population dynamism and the main interactions between disorders linked to Q fever, especially reproductive disorders (abortion, retained foetal membranes, purulent vaginal discharge and endometritis, extra services, and calving–conception delays). The effects of the nontangible cost of the disease on human health, the welfare of the animals, and the workload of farmers were not integrated into the model. The model shows high-level sensitivity to the prevalence of Q fever in the herd prevaccination and to the costs of abortion and extra days of calving–conception intervals. Breakeven points, i.e., cost values that allow us to achieve positive vaccination benefits, are also reported. For herds with moderate or high prevalence rates of Q fever prevaccination (>30%), a vaccination benefit is observed. The vaccine should be considered a type of insurance in herds with low prevalence rates of Q fever prevaccination (≤20%). The calculator was developed to aid decision-making at the farm level, and no conclusion can be extrapolated as a generic trend based on the present work.


Introduction
Coxiella burnetii is a zoonotic bacterium with a worldwide distribution that is responsible for Q fever.It can be asymptomatic in humans, or it can present with flu-like symptoms [1]; additionally, this bacterium can affect other species [1][2][3].The average overall animal prevalence, interherd prevalence, and intraherd prevalence of Q fever in cattle are reported to be 20%, 40%, and 20%, respectively [4].In France, a 2017 seroprevalence survey [5] reported a herd-level prevalence of 36% for cattle.
The clinical impact of Q fever differs among ruminant species; for instance, abortion related to Q fever is epidemic in goats but more endemic in cattle.In cattle, Coxiella burnetii infection has been found to be responsible for other reproductive disorders, such as endometritis [3,[6][7][8], retained fetal membranes (RFMs) [9][10][11] and subfertility [12][13][14].An inactivated Coxiella burnetii phase-I vaccine (Coxevac ® , Ceva Santé Animale, Libourne, France) has been authorized for cattle, goats, and sheep and is commercially available in many countries.This vaccine contributes to mitigating the zoonotic risk of Q fever through a reduction in shedding via milk, faeces, and vaginal discharge, especially around parturition, as observed by Schulze [15].
No literature is available on the economics of Q fever in cattle and the potential economic benefits for the farmer of vaccination at the herd level.Previous studies have focused on the zoonotic impact of Q fever globally [16][17][18], for instance, by assessing the cost of intervention in the animal sector to prevent human infection [19], the cost-effectiveness of chronic Q fever screening in humans [20], and the economics of Q fever vaccination for agricultural industry workers [21].Farmers face many challenges on their farms and have to make decisions on Q fever management based on either sanitary (animal or public health issues) or economic concerns.To support farmers and their veterinarians in this decision, a Q fever economic assessment calculator is created.Performing the abovementioned studies is challenging since the association between Q fever and health or production disorders is imprecisely quantified, and reproduction is the result of multiple factors, making it difficult to estimate the fraction of reproduction attributable to Q fever.The economic assessment of the total impact of a disease or of the potential benefits of mitigation measures such as vaccination requires precise knowledge of the epidemiology of the disease [22].The lack of precise data on the epidemiologic impact of Q fever prevents any accurate economic assessment and may explain the lack of publications on the economics of Q fever despite the potentially high impact of this zoonotic disease.The present work aims to describe the tool, its rationale, and the supporting assumptions made to create this accurate and easy-to-use economic estimator.

Overview
The calculator is based on a modified partial budgeting approach applied at the herd level.The first outcome of interest for the farmer is the total yearly impact of Q fever in his or her herd before any intervention for a given prevalence of Q fever, defined as the yearly production losses due to Q fever (ProdLoss QfPrev ).The second outcome of the calculator is the 3-year benefit of Q fever vaccination for the herd (VaccBenefit QfPrev ).
ProdLoss QfPrev is defined as follows: (i) prevaccination according to the prevalence of Q fever (QfPrev) prior to any intervention (QfPrev BefVacc ) and (ii) postvaccination according to the expected prevalence of Q fever postvaccination (QfPrev AfterVacc ).The 3-year VaccBenefit is then defined as follows: where CostVaccine is the total cost of herd vaccination for the 3-year period.QfPrev BefVacc is considered constant for the 3-year analysis period, considering that there is no intervention by farmers and that circulation within the herd leads to constant prevalence (infected culled cows replaced by infected heifers).QfPrev AfterVacc decreases each year due to herd dynamics (infected culled cows replaced by vaccinated heifers).
The vaccination programme was carried out using Coxevac ® (Ceva, Libourne, France) with the following protocol: all cattle above 3 months of age were vaccinated every year, i.e., 2 injections for all cattle above 3 months of age in the first year plus 2 injections for 3to 12-month-old heifers in years 2 and 3 and an annual booster for other cattle from year 2 onwards.

Production Losses for a Herd with Q Fever
The production losses ProdLoss QfPrev are defined as follows: where CostAbort QfPrev is the cost of extra abortions (Abort) due to Q fever for QfPrev compared to the situation with no Q fever; CostRFM QfPrev is the cost of extra RFMs due to Q fever for QfPrev compared to the situation with no Q fever; CostMet QfPrev is the cost of extra purulent vaginal discharge and endometritis (metritis-Met) due to Q fever for QfPrev compared to the situation with no Q fever; CostAI QfPrev is the cost of extra artificial insemination (AI) due to Q fever for QfPrev compared to the situation with no Q fever; CostCCI QfPrev is the cost of extra days of calving-conception interval (CCI) due to Q fever for QfPrev compared to the situation with no Q fever.

Yearly Prevalence of Q Fever in the Herd
QfPrev is considered the main indicator of the herd infection level.The economic impact of prevaccination and the benefits of vaccination are linked to QfPrev and the decrease in this value due to vaccination, respectively.Due to herd dynamics and the culling of cows with Q fever that are replaced with vaccinated disease-free heifers, the value of QfPrev decreases postvaccination.The model does not consider that cows with Q fever are at greater risk of being culled.
Although the model is static, QfPrev is calculated yearly (QfPrev Yx ) based on the mean between QfPrev at the beginning of the year (QfPrev StartY ) and that at the end of the year (QfPrev EndY ), as indicated in the following: QfPrev StartY1 is an input parameter of the final calculator, representing the prevalence of the herd prior to vaccination.The full vaccination of the herd starts at the beginning of year 1, and boosters are administered yearly.The clinical protection brought about by vaccination is modelled in the present work through a change in QfPrev Yx , even if the prevention of infection through vaccination is not evidenced: this structure of the economic model is in accordance with Equation (2), where the economic consequences of Q fever linked to clinical outcomes are modelled through QfPrev Yx .
The calculation of QfPrev Yx (Equation (3)) is based on a simplified compartmental population approach for a fixed-size herd.The first subpopulation is the Q fever-free heifers joining the in-milk cow herd.This subpopulation is considered to remain Q-fever-free due to the vaccination, and the subpopulation size is calculated based on the culling rate of the herd (CullRate), assuming equal culling and replacement rates (fixed herd size).The second subpopulation is composed of culled cows that leave the herd based on farmer criteria: these cows can have and not have Q fever, with a share equal to QfPrev StartYx , assuming that random culling occurs among the cows (no consideration of Q fever status for culling decisions).The third and final subpopulation is composed of cows remaining in the herd for the whole year, with a share between cows with and without Q fever equal to QfPrev StartYx .Combining the 3 subpopulations leads to the following: Equation ( 4) is applied independently for years 1, 2, and 3 via Equation (5), and the results are combined in Equation (3) to obtain the mean prevalence of Q fever in the herd for each year.

Disease Impact before Vaccination
As detailed in Equation ( 6), CostAbort QfPrev is based on the number of cows with Q fever (HerdSize * QfPrev StartY1 ) and the number of heifers with Q fever (HerdSize * QfPrev StartY1 where AbortRate NoQf is the abortion rate of animals with no Q fever; RR AbortIfQf is the relative risk (RR) for abortion in animals with Q fever compared to those without Q fever; Mitigation Heifers is the mitigation coefficient for heifers to account for the expected lower prevalence of Q fever in heifers than in cows in a given herd; UnitCost Abort is the unit cost of abortion.
Similarly, CostRFM QfPrev is based on the number of cows with Q fever and the difference in the risk of RFMs in cows with Q fever compared to that in cows without Q fever (Equation ( 7)): where RFMRate NoQf is the RFM rate for cows with no Q fever; RR RFMIfQf is the RR for RFMs in cows with Q fever compared to those without Q fever; UnitCost RFM is the unit cost for RFM treatment.
Next, CostMet QfPrev is based on the number of cows with Q fever and the difference in the degree of risk for metritis in cows with Q fever compared to cows without Q fever (Equation ( 8)): where MetRate NoQf is the metritis rate in cows with no Q fever; RR MetIfQf is the RR for metritis in cows with Q fever compared to those without Q fever; UnitCost Met is the unit cost for metritis treatment.Finally, CostAI QfPrev is based on the number of animals with Q fever, including heifers, in terms of abortion as well as the number of extra artificial insemination (AI) procedures in the case of Q fever, as indicated as follows: where ExtraAI IfQf is the extra service per conception in patients with Q fever compared to patients without Q fever; UnitCost AI is the unit cost for AI.
In addition to this disorder, the literature also shows an association between Q fever and (i) late AI after previous AI [9], (ii) deteriorated first service conception rate (FSCR; [10]), and (iii) calving to conception interval (CCI; [10]).Based on this literature and expert opinion and as detailed in the calibration section, these 3 contributors are included in the component (CostCCI QfPrev ) of Equation ( 2), as indicated in Equation (10).As a consequence, CostCCI QfPrev is based on the number of animals (cows and heifers) with Q fever and the additional days for CCI for animals with Q fever, considering extra CCIs based on 3 components.First, extra CCIs associated with late AI after previous AI in cases of Q fever (ExtraCCI IfQf_LateAI ) are considered for cows and heifers; second, extra CCIs associated with deteriorated FSCR in cases of Q fever (ExtraCCI IfQf_FSCR ) are applied for cows and heifers; third, extra CCIs directly associated with Q fever (ExtraCCI IfQf_DirectQf ) are included for cows only (Equation ( 10)).
where UnitCost CCI is the unit cost per extra day of CCIs.

Disease Impact of Vaccination and the Cost of Vaccination
The decrease in the value of QfPrev Yx in the case of vaccination is applied directly to Equations ( 7)- (10) to assess CostMet QfPrev , CostRFM QfPrev , CostAI QfPrev , and CostCCI QfPrev for vaccinated herds by replacing QfPrev StartY1 with QfPrev Yx , respectively.This means that the benefit of vaccination is considered to rely only on the decrease in the prevalence of cows with Q fever as permitted by vaccination.
CostAbort QfPrev for vaccinated herds also accounts for the decreased risk of abortion for animals with Q fever and who have been vaccinated compared to cows with Q fever and who have not been vaccinated, as detailed in the following: where RR AbortIfVaccIfQf is the RR for abortion for a cow with Q fever and if vaccinated compared to a non-vaccinated cow.

Calibration
HerdSize, QfPrev StartY1 , and CullRate are considered fixed in the present work (Table 1) but are adjusted for each farm while using the calculator.To keep the calculator as simple as possible for use in the field, QfPrev StartY1 is used as the main indicator for the Q fever level in the herd, and a half-adjustment for heifers (Mitigation Heifers = 0.5) is applied to account for the lower-level prevalence of Q fever in heifers than in cows [15,23].
The base prevalence (i.e., not related to Q fever) of abortion, RFMs, and metritis (AbortRate NoQf , RFMRate NoQf and MetRate NoQf ) were set at 5%, 4%, and 9%, respectively (Table 1), based on a literature review [24,28].The values for relative risks or extra service per AI in cases of Q fever (Table 1) are defined based on the literature, as detailed in Appendix A. The values for ExtraCCI IfQf_LateAI and ExtraCCI IfQf_FSCR are estimated to be 2 and 3 days, respectively, based on the literature, as detailed in Appendix A.
Unit costs (mean and ranges) are set up by the end users of the calculator.The sample of results reported in the present work is based on expert opinion and grey literature.To avoid any overestimation in the Q fever economic assessment, UnitCost Abort , UnitCost RFM , and UnitCost Met consider only direct costs (production losses, treatment, etc.) and exclude extra labour costs and middle-term consequences for fertility (conception per service and long CCI) that are already considered in other components of the total impact of Q fever (CostAI QfPrev and CostCCI QfPrev ).This approach is in accordance with epidemiologic studies linking Q fever and service per conception or the CCI, which are not adjusted for the presence of abortion, RFMs, or metritis.
No discount rate is included due to the limited period (3 years) considered for the model.The vaccination cost includes only the vaccine and excludes the labour cost.

Model Highlighting the Cost of Q Fever for the Farmer and the Benefits of Vaccination
The present model was developed to guide farmers' decisions.This model is not designed to draw general conclusions on Q fever cost components.Through a good understanding of its framework (how it works) and the observation of some results, it remains possible to highlight how the model behaves while avoiding any general extrapolation.
The model shows that abortion and CCI are the two main components of both the economic impact of Q fever and the benefits of the related vaccination (Table 2), regardless of the calibration.Abortion represents 25 to 30% of total costs or vaccination benefits, and changes in the CCI account for 46 to 53%, depending on the model input parameters.Additional AI, RFM, and metritis treatments contributed to 20-25% of the economic impact of Q fever or vaccination benefits (Table 2).This observation (abortion and CCI as the two main components) is in agreement with how the models were built, i.e., with most of the middle-term impact of Q fever on reproduction and with the high-level impact of Q fever on reproduction.These shares of each contributor in the total cost or benefit appear stable even with low or high input parameters.QfPrev Y0 represents the prevalence of Q fever in the herd before vaccination.Vaccine, UnitCost Abort , UnitCost RFM , UnitCost Met , UnitCost AI and UnitCost CCIet are the unit costs for abortion, RFMs, metritis, AI and days of CCI, respectively.% indicates the share of the total costs or benefits.
Table 2 shows that the 3-year total cost of Q fever for 100 cows ranges from EUR 7500 to EUR 23,000 depending on the values used for QfPrev StartY1 and UnitCost.
Case 1 of Table 2 represents an average situation of calibration and shows a positive net benefit of vaccination.The breakeven points represent the values of those input parameters for which the vaccination benefit is just above zero (the beginning of positive returns for vaccination).In herds with a moderate or higher prevalence of Q fever, which was simulated here with an initial prevalence (QfPrev StartY1 ) of 30% or above, a combination of breakeven points for unit costs is EUR 2, EUR 300, EUR 35, EUR 50, and EUR 50 for UnitCost CCI , UnitCost Abort , UnitCost AI , UnitCost RFM , and UnitCost Met , respectively (case 2, Table 2).In these situations of a moderate initial prevalence of Q fever, the calculator can be used to better appreciate the benefits of the vaccination, especially depending on the model input parameters.As abortion and CCI are the main contributors and as the results show that the outcomes are highly sensitive to these two components, users must focus on appropriate values for UnitCost Abort and UnitCost CCI .UnitCost CCI has been reported to be non-uniform: its value dramatically changes, especially based on the mean CCI and price of milk [27].Under the main dairy production system in the EU, the minimum value for UnitCost CCI should be considered equal to EUR 2, and UnitCost CCI increases very quickly for herds with a deteriorated CCI or with high costs for milk not produced.Similarly, UnitCost Abort depends on the genetic level of the herd, the physiologic stage of abortion, the parity of the aborted cow, and the average milk production level of the herd (peak and persistence) [24].Although its influence on the results is limited, UnitCost AI also highly varies across countries and production systems, including criteria such as the genetic level, the actors involved (farmer or veterinarian, for instance), and the costs of semen or drugs [26].
The results are also highly sensitive to QfPrev StartY1 , with, for instance, high-level benefits in the case of QfPrev StartY1 = 40%, regardless of the value of UnitCost (cases 3 and 4, Table 2).In herds with a very limited degree of circulation of Q fever, simulated here with an initial prevalence (QfPrev StartY1 ) of 20% (case 5), a combination of breakeven points for unit costs is EUR 3.5, EUR 400, EUR 40, EUR 60, and EUR 60 for UnitCost CCI , UnitCost Abort , UnitCost AI , UnitCost RFM , and UnitCost Met , respectively.In these herds with low QfPrev StartY1 values, vaccination remains cost-efficient in cases of moderate or high UnitCost.This finding shows that the users of the calculator should pay attention to the two key input parameters, UnitCost CCI and UnitCost Abort .Moreover, in cases of limited benefits, as suggested by the calculator, vaccination can also be considered a zero-cost or limited-cost insurance calculator to prevent any further deterioration of herd performance.

Partial Budgeting Approach: Adaptations Performed to Scope with Its Limitations
Partial budgeting is no longer considered an appropriate approach for assessing the economic impact of diseases or the benefit of any mitigation measures, especially in dairy production systems.The bias arising from an oversimplified partial budget was discussed in detail previously [25].This bias arises from (i) the static approach, although dairy production is dynamic and follows a long-term pattern; (ii) the deterministic framework, although both epidemiology and economic stochasticity are observed in the field; (iii) the slicing between the different disorders linked to the studied disease despite multiple interactions between them.Despite these limitations, partial budgeting is preferred to any dynamic stochastic and interactive model here because the first objective is to use a herd-side calculator to support farmer decisions and address the enormous limitations in terms of the epidemiological data available in the studied domain.A cutting-edge methodological model with imprecise calibration is incoherent and can even bias users' feelings with an overestimation of the robustness of the results they obtain.
One major concern in the economic evaluation of Q fever economic burden or vaccine benefit at the farm level is the precision of epidemiological studies examining the impact of Q fever on animal performance.For instance, in the case of no vaccination, the present model considers the same abortion rate each year after the introduction of Q fever in the herd for the 3-year period of the analysis.The literature shows that after the Q fever-related abortion outbreak in goats, the number of abortions decreased the following year.For cattle, no data are available on the relative epidemiologic pattern of abortion annually.Because Q-fever abortion is more common in goats than in cattle, we assumed that the prevalence of abortion in years 2 and 3 after the outbreak was the same as that in year 1.Similarly, the association between Q fever and endometritis is not consistent [3,[6][7][8], and the model considered this association, as indicated in Table 1.The authors believe that this parameter was the most appropriate given the data available in the literature.The results show that endometritis only slightly contributes to the total cost or vaccine benefits, suggesting that this assumption has a very limited impact on the results.
This example also clearly demonstrates that more studies are required to better understand the epidemiology of Q fever in cattle.
The modifications we provide to the partial budget approach allow for a precise simulation of herd dynamics: compartmental modelling year by year for 3 consecutive years succeeds in precisely simulating QfPrev Yx for the 3 years following vaccination (Table 3).The high-level sensitivity of the results to QfPrev StartY1 (Table 2) and the large change in QfPrev Yx over time (Table 3) demonstrate the usefulness of compartmental modelling for improving economic model accuracy despite its partial budget basis.The initial prevalence of Q fever in cows (QfPrev StartY1 ) is not an indicator that is easy to estimate in the field.The prevalence of Q fever is likely a good proxy for QfPrev StartY1 despite its inability to determine the infectious status of cows.Moreover, the prevalence of this disease is not known among farmers in the field.The prevalence of seropositive cows within milk herds is reported to be 40% under French conditions, ranging between 20 and 60% [29].The mean prevalence in France was estimated to be 36% in a recent study [5].For practical use of the calculator in the field, we suggest considering 20% of the minimum value for QfPrev StartY1 .Identifying herds whose mean prevalence is less than or greater than 20% is possible because of the use of milk ELISA tests [30].In the case of a substantial number of abortions in a herd, assuming that the Q fever-related abortion rate is a good proxy of the herd's Q fever infection status, we suggest using the value of QfPrev StartY1 = 40% (i.e., the approximate mean of French prevalence).Alternatively, the average value for QfPrev StartY1 can reach 30%.
The model framework considers the impact of Q fever on heifers applied to the same percentage (i.e., 20% or 40%) of animals despite evidence of a lower prevalence in heifers than in cows [15,23].To adjust for this factor, the impact for heifers was considered to be 0.5-fold that of cows (Mitigation Heifer = 0.5).This assumption enables the use of one value of QfPrev StartY1 for the calibration of the entire model in accordance with the difficulties in assessing QfPrev StartY1 in the field.

Issue of the Reproduction Complex in the Economic Assessment of Dairy Production Outcomes
The economic assessment of Q fever is a typical example of the difficulties that arise when a static model tries to simulate a dynamic and complex system with multiple interactions between the disorders associated with Q fever.First, Q fever simultaneously impacts RFMs, metritis, service performance, and abortion, four items in dynamic interaction, and the quantifications available around Q fever and these four items are very limited.
For RR AbortIfQf , RR RFMIfQf , and RR MetIfQf , the accuracy of the literature seems good (Table 1 and Appendix A).Although only one publication has reported a trend towards decreased degrees of abortion risk in the case of the vaccination of infected cows and heifers, this association is considered in the present economic model [9].No benefit of the vaccination of infected cows in the reduction of RFMs or metritis is considered in the model due to a lack of evidence in the literature.The risk of overestimation in the Q fever economic assessment is limited in the present work, as only direct costs or treatments for UnitCost Abort , UnitCost Met , UnitCostRFM, and UnitCost AI and middle-term consequences on reproduction performance have been accounted for through the CCI (Appendix A).Similarly, late AI after previous AI [9] and FSCR [10] are considered only through extra CCIs and are associated with two and three CCIs, respectively, as described in Appendix A. Although these calculations are very approximate, the small impact of these two contributors in terms of the CCI when compared to the direct impact of Q fever (+14 days of CCI) shows that the risk of overestimation through these raw calculations is very limited.
As detailed in Appendix A, no data are available on the impact of late AI after previous AI, extra service per conception, or extra CCI when comparing infected herds to noninfected herds, but the opposite association is described when comparing vaccinated populations to nonvaccinated populations [9,10].Despite this lack of data, the authors consider that any benefit of vaccination can be observed only in the case of the impact of the disease, and the authors extrapolate these three impacts in infected cows and heifers compared to in non-infected cows and heifers.This extrapolation is required to avoid any appropriate estimation of the impact of Q fever when no vaccination is performed, which is an objective of the present work.The lack of consideration of these contributors may lead to severe underestimation.
Building a bioeconomic model requires many trade-offs within its design and calibration.These trade-offs should not be seen as a decrease in the quality of the work or in the precision or robustness of the results; they only represent the best use of the data and knowledge available at this time, considering not only the epidemiological part but also the economic rationale.The concerns previously highlighted about the precision of the model (abortions in case of no vaccination; Q fever and endometritis) or the fact that the slight decrease in milk after vaccination [31] is not considered are in accordance with the assumptions made for the economic part of the model and related to the limitation in the accuracy of the unit costs (Table 1).The vaccination strategy that farmers may develop in the field aims at (i) maximizing the decrease in the bacterial load, (ii) reducing the number of shedders, and (iii) improving reproductive performances.The present model considers that vaccination is occurring at the beginning of the 3-year period and within a herd with Q fever.The available literature does not provide any possibility to consider the bacterial load and number of shedders in the economic assessment in spite of this may influence the economic benefit of vaccination.The present model would benefit from updates as additional information becomes available.

Focus on Avoiding Overestimation
As detailed above for the population dynamics and for the reproduction complex simulation, the overestimation of the total cost and vaccination benefits is a key driver of the conception and calibration of the estimator.This state of mind also leads to other decisions, such as (i) no cost of labour being included in the model, despite labour costs for the management of the consequences of Q fever possibly being higher than those for vaccination, and (ii) no overculling of cows with Q fever, despite these cows having a higher degree of abortion risk and deteriorated reproductive performance, thus being more likely to be culled.Similarly, the positive effects of vaccination on the public health and health status of neighbouring farms are not accounted for since the model is focused only on the focal farm.Finally, the model considers only 3 years after the start of the vaccination in accordance with the limitations of the model calibration, as previously highlighted, and the difficulties in quantifying the dynamics of Q fever in the herd afterwards.This choice by the authors to ensure the accuracy of the calculator also contributes to vaccination benefit underestimation since we expect greater benefits a few years after the start of the vaccination.
As a tool to be used in the field, the present model is built considering the vaccination of a herd with Q fever already present on the farm, despite vaccination occurring before any infection, as a form of prevention support.This choice to use the vaccine in an infected herd reduces the expected benefits of the vaccine and may underestimate the usefulness of vaccination in Q fever-free herds.

Conclusions
The economic assessment of Q fever and the potential benefits of vaccination is very challenging due to the scarcity of available data and the dynamic and long-term pattern of dairy production.The improved partial budget model applied to Q fever integrates a simplified yearly compartmental model for a better representation of population dynamism and accounts for the main interactions among the reproduction consequences of Q fever.The model shows a high degree of sensitivity to QfPrev StartY1 , UnitCost Abort , and UnitCost CCI .The calculator aims to help farmers make decisions at the farm level, but the specific results provided here are not generalizable.
1: Details in Appendix A.

Table 2 .
Example of results (total cost of Q fever and vaccination benefits; right) for 5 cases of combinations of input parameters (left).The results are expressed in euros for each year and the 3-year period (bold).

Table 3 .
Values of QfPrev Yx obtained with the model depending on QfPrev StartY1 .QfPrev Yx represents the average prevalence of Q fever in the herd for year x.QfPrev StartY1 represents the prevalence of Q fever in the herd before vaccination.