4 Health impact assessment and burden of zoonotic diseases

Numerous zoonotic diseases cause morbidity, mortality and productivity losses in both humans and animal populations. Recent studies suggest that these diseases can produce large societal impacts in endemic areas. Estimates of monetary impact and disease burden provide essential, evidence-based data for conducting cost-benefit and cost-utility analyses that can contribute to securing political will and financial and technical resources. To evaluate burden, monetary and non-monetary impacts of zoonoses on human health, agriculture and society should be comprehensively considered. This chapter reviews the framework used to assess the health impact and burden of zoonoses and the data needed to estimate the extent of the problem for societies. Case studies are presented to illustrate the use of burden of disease assessment for the zoonotic diseases cystic echinococcosis, Taenia solium cysticercosis, brucellosis and rabies.

Numerous zoonotic infections result in ill health and economic losses in humans and animals. The challenge in assessing the burden of a zoonotic infection is that most approaches will estimate the impact of the infection in one species at a time. In addition, some approaches are only applicable to measuring human health impact, and are not applicable to animal health impact. To overcome this challenge, a number of large-scale initiatives and individual researchers have endeavoured to assess the burden of these infections in both non-monetary and monetary terms. Non-monetary assessment can include measures of mortality, morbidity (which often includes reduction in productivity in animals), and health adjusted life year (HALY) measures which include the quality adjusted life year (QALY) and the disability adjusted life year (DALY). Monetary assessment of these conditions should include both direct and indirect costs associated with disease in both human and animal hosts.

Mortality has traditionally been used to measure and compare the health status of populations (Hyder and Morrow 2001). For humans, vital statistics data have been shown to be available from only 115 out of 192 member states of the World Health Organization (WHO) (Mathers et al. 2005). Of those 115 member states, death registration was considered complete in only 64 (33%). In addition, among 106 countries with recent data at least 50% complete, the quality of the data on classification of the causes of death was considered as high, medium and poor in 23, 55 and 28 countries, respectively. These data underline the difficulty of measuring burden across states even using what is usually considered an objective and reliable measure, death.

There is no systematic data collection on causes of animal deaths, except in the case of the occurrence of an outbreak. In livestock, slaughterhouse data, when available, may be used to estimate the prevalence of infections (see later), but do not reflect causes of death nor natural (non-slaughter) death rates. There is no vital statistics system for pet animals. Therefore, while there is a way to compare human death rates across countries, even if they are not completely accurate, such measures are not available for animal populations.

Causes of death data may accurately measure the impact of deadly diseases, but ignore those diseases that may be chronic and disabling, but not or rarely fatal. In humans, such data can be estimated from clinical records systems, health insurance claims databases or notifiable disease or cancer registries. Medical records systems will only reflect the incidence rates of diseases among people seeking medical care, even in countries with universal health coverage. Notifiable disease surveillance data are known to underestimate the true frequency of infectious diseases, but are helpful to detect outbreaks and analyse temporal and spatial trends (Trottier et al. 2006; Giesecke 2002; Chorba 2001; Nelson and Sifakis 2007). In addition, models have been developed for some diseases, such as measles, to adjust for under-reporting (Fine and Clarkson 1982). The reliability of the classification of diseases from medical records is often poor (De Coster et al. 2006). Finally, availability of medical record data is limited to only a few developed countries.

In animals, data from slaughterhouses or official carcass inspections in markets can be used, where available, to estimate the frequency of diseases in animals. Such data are only useful in countries where home slaughter is uncommon and where the majority of carcasses will be inspected. The advantage of these data is that most animals will be eventually slaughtered thus providing reliable estimates of the frequency of lesions associated with infections under surveillance. For example, slaughterhouse data were used in a recent publication aimed at estimating the monetary burden of cystic echinococcosis in Spain (Benner et al. 2010). Unfortunately, reliable abattoir data are only available from developed countries and are very limited elsewhere.

Companion animals can be the source of a number of zoonoses. Unfortunately, accurate data on diseases in such animals are rarely available except with certain notifiable diseases (e.g. rabies). However, an initiative to address this problem, the National Companion Animal Surveillance Program, is being undertaken in the USA using the electronic records of over 500 small animal hospitals and the electronic reports from diagnostic laboratories serving over 18, 000 private veterinary practices (Glickman et al. 2006). This initiative aims to provide better information about zoonotic and emerging diseases in companion animals. These data will, however, still present the same limitations in that they only represent those animals under care that were accurately diagnosed, but this is a considerable improvement on the current lack of information.

Surveillance data may also be used in animals. The World Organization for Animal Health (OIE) recently launched the World Animal Health Information Database (WAHID), which aims at collecting data on notifiable animal diseases, including zoonoses. Even though an improvement from what has been traditionally available, this source of data has the same limitation as human notifiable disease data, in that it is highly dependent on the willingness of people to report cases of disease. Also, for several of these notifiable infections, only outbreaks are being reported and information on endemic infections is not available. Another common limitation in estimating the frequency of infection in animals is the lack of accurate estimates of animal populations, leading to the absence of valid denominators to estimate frequency measures such as incidence rates or prevalence proportions.

In the absence of reliable data from surveillance or medical records in humans and slaughterhouse or surveillance data in animals, estimates of prevalence can be obtained through surveys conducted by independent researchers. However, these types of surveys are problematic because they are often conducted in areas known to be endemic for the infection of interest, which leads to an overestimation of the true frequency of the infection. In addition, estimates of disease prevalence are often obtained via serological testing (e.g. to diagnose cystic echinococcosis or T. solium infection in humans and livestock). These tests are developed to detect exposure to the infectious agent (tests designed to measure antibodies) or to detect the presence of the infectious agent (tests designed to measure the agent itself or its antigens). Exposure to the agent may have occurred in the past and its association with the development of symptoms, which cause disability, may be very weak. Similarly, the presence of the agent in an infected individual will not always result in the development of symptoms. For example, not all larvae of T. solium will migrate and cause lesions in the brain and not all brain lesions will be associated with symptoms. In a cross-sectional survey of people with epilepsy in South Africa, an ELISA test for the detection of the antigens to the larval stages of T. solium was estimated to have a sensitivity of only 17.4% to detect lesions of neurocysticercosis identified by CT-scan (Foyaca-Sibat et al. 2009). In a cross-sectional survey conducted among healthy volunteers in Mexico, 9.1% demonstrated asymptomatic lesions of neurocysticercosis on a brain CT-scan (Fleury et al. 2003).

The poor performance of serological tests to detect symptomatic cases is due to the fact that several zoonotic infections remain asymptomatic and have variable incubation periods. More expensive and intrusive methods may be available, but are often not accessible or not usable in the field. For example, both echinococcosis and neurocysticercosis rely heavily on advanced imaging to make a definitive diagnosis in humans. These techniques can incur more than minimal risk to the subjects, which limit their use in large scale surveys. Regardless of the diagnostic test(s) used, error will occur and must be accounted for in the analysis of frequency data. Latent class analysis methods, including Bayesian methods, have been applied to obtain disease frequency estimates based on imperfect tests (Dorny et al. 2004; McGarvey et al. 2006; Carabin et al. 2005). However, such approaches can only be used if some prior knowledge is available on the accuracy of the test specific to the region where it is being used. Indeed, due to cross-reactions with other agents, the specificity of some tests can vary extensively from region to region. To better estimate the risk (or cumulative incidence) of infection over time or the incidence rates of infection per person-time at risk, a cohort study is needed. In this design, a group of individuals initially free of infection are followed-up for a set period and new infections are identified. Unfortunately, these types of studies are very expensive and may be cost-prohibitive in the communities where they are most needed. The additional need to follow both animals and humans further reduces their use in the context of zoonotic infections.

In order to assess losses associated with health impact assessment of zoonotic conditions, productivity losses must be measured in addition to estimates of disease frequency (e.g. mortality and morbidity). An example of productivity losses, in humans, is the inability of an individual to work due to the symptoms associated with the sequelae of an infection such as cystic echinococcosis or neurocysticercosis. For individuals currently in the work force, actual estimates of salary losses can be used. However, for individuals, and especially women, who are not formally employed, losses in productivity due to illness are much more difficult to estimate. Nevertheless, it is important to account for productivity losses for all affected members of society. For example, Majorowski et al. (2005) estimated that women not formally economically active and retired individuals were considered 30% and 10% as productive as a formally employed individual, respectively. In a study estimating the monetary burden of cysticercosis in the Eastern Cape Province of South Africa, a sensitivity analysis was conducted to estimate the effect of three alternative methods to value time of unemployed or unofficially employed individuals (Carabin et al. 2006). In animals, productivity losses can be measured in terms of reductions in milk or wool production, reproduction indices and growth (Majorowski et al. 2005; Benner et al. 2010).

Non-financial methods have been developed to estimate human disease burden and are often referred to under the umbrella term of Health Adjusted Life Years (HALYs) (Gold et al. 2002). HALYs are summary measures of population health that combine the impact of morbidity and death. HALYs aim at comprehensively measuring the impacts associated with all aspects (domains) of an ill-health condition. Hence, in contrast to measures of morbidity or mortality, they are designed to capture the impact of all aspects of an infection, including more subtle manifestations which may not be diagnosed with usual methods. However, the use of HALYs does have a number of disadvantages and may not always be appropriate when estimating the societal burden of zoonoses. For one, no HALYs are applicable to animal disease.

Disability Adjusted Life Years (DALYs) are the WHO’s preferred measure of disease burden and were utilized in the Global Burden of Disease (GBD) Study (Murray 1994; Murray and Lopez 1996). The goal of the original and subsequent versions of the GBD Study was to consistently assess burden across diseases, risk factors and regions. In other words, the goal of DALYs was to assess the burden of human diseases, assuming that these diseases caused the same level of ‘disability’ across cultures. While the assumption that the same level of condition-specific disability exists regardless of culture or socioeconomic status is a key feature of the DALY, it is also one of the largest points of contention. It has been argued that an individual’s ability to adapt to a disability can vary greatly based on the socioeconomic status of the region. For example, Allotey et al. (2003) discussed the differences in the ability to adapt and cope with a condition such as paraplegia for an individual living in Australia as compared to an individual living in Cameroon. Their conclusion was that the DALY undervalues the burden of disease in less developed countries. However, even with this and other shortcomings, the DALY continues to be an important and widely used burden of disease metric.

The DALY is estimated as the sum of the number of years of life lost due to mortality (YLL) and the number of years lived with a certain level of disability (YLD) associated with a specific disease or infection. To quantify DALYs, the frequency of symptoms and sequelae associated with each infection must be known. Those symptoms and sequelae are assigned specific disability weights for as long as they persist, and, therefore, their duration must also be known. Disability weights were developed to capture relative ‘disabilities’ of health states on a 0 to 1 scale. For example, a healthy individual has a disability weight of 0 and no loss of DALYs, whereas a fatal condition has a disability weight of 1 (Table 4.1).

While the DALY is the most well known and most employed burden of disease metric, other HALYs do exist. These methods arose due to a concern that measuring morbidity or mortality alone did not capture all aspects of a health condition. Another important rationale for the development of quality of life measures was the possible side effects of treatment for diseases which may outweigh their benefits, but could not be measured through morbidity alone. Health outcome measures first appeared in the late 1960s and early 1970s (Klarman et al. 1968; Fanshel and Bush 1970; Bush et al. 1972; Torrance et al. 1976), with the term quality adjusted life year (QALY) coined by Weinstein and Stason in 1977. QALYs are often considered the precursors to DALYs and are conceptually similar to DALYs, but use different scales. In general, index scores are calculated with QALY standard questionnaires and these are linked to a ‘utility’ value, which aims at evaluating the ‘value’ of a health status and were originally designed to be the equivalent of the concept of utility of goods in economics. Thus, interventions would aim to minimize DALYs but maximize QALYs.

The effectiveness of intervention strategies can be calculated as the number of DALYs estimated to occur due to a given condition in the absence of intervention minus the number of DALYs expected if control measures were implemented. However, this disregards any additional benefits of disease control to agriculture (i.e. anthelmintic treatment of dogs reduces cystic echinococcosis incidence in both humans and sheep). Nevertheless, if a total societal financial analysis is undertaken, the true cost-effectiveness of control, in terms of DALYs saved, can be estimated by implementation of cost sharing between sectors proportional to each sector’s overall benefit (Roth et al. 2003).

Official burden of disease estimates, in terms of DALYs, for zoonotic diseases are limited. For example, until recently, human cysticercosis, cystic echinococcosis, and rabies were not officially evaluated as part of the GBD Study or other international project to evaluate non-monetary disease burden. Research is, however, underway to overcome this deficit and include all three conditions in the upcoming re-evaluation of the GBD Study. In addition, cystic echinococcosis and T. solium cysticercosis are both being evaluated as part of the WHO’s assessment of the global burden of food-borne diseases (WHO 2010).

Monetary losses associated with zoonoses can be calculated by the following expression (Majorowski et al. 2005). Equation 1.

This equation corresponds to the additive societal costs for all affected species (S) across all age groups (A). For the age-species-specific population of size (N  _a,s), with the age-species-specific annual incidence (β  _a,s), there is an age-species proportion (π  _x,a,s,) of infected individuals with symptoms X. The treatment and consequences of each of these symptoms have a cost of C  _x,a,s. Ideally, the whole spectrum of symptoms and losses in humans and animals is included.

Both direct and indirect costs should be included in monetary estimates of human disease burden. Direct costs include costs associated with resources expended for health care (e.g. diagnostics, medications, surgical treatment, etc.). Indirect costs include costs of the resources forgone either to participate in an intervention or as the result of a health condition (e.g. earnings lost because of loss of time from work, costs of transportation to and from treatment, earning losses for a family member while taking care of the sick individual). As discussed earlier, indirect losses associated with productivity losses can be difficult to assess for those not formally employed.

Animal-associated monetary losses can also be divided into direct and indirect costs. Direct costs can stem from monetary losses due to condemnation of an infected carcass or parts thereof. For example, the condemnation of a pig carcass heavily infected with T. solium cysticerci or the condemnation of a sheep liver infected with cystic echinococcosis are direct costs. Indirect costs would encompass other disease-related production losses. For example, decreased carcass weight linked to a reduction in growth, decreased hide value, decreased milk production, and decreased fecundity due to cystic echinococcosis infection in a sheep. Even though estimates of these losses are sometimes uncertain and have very little available data, studies incorporating estimates of indirect costs for livestock-associated losses have shown that these losses are likely to be significant. For example, monetary losses associated with cystic echinococcosis in Spanish livestock contributed to an estimated 10.4% of cystic echinococcosis total costs, with indirect costs contributing to 10.3% and direct costs contributing to only 0.12% of losses (Benner et al. 2010).

As mentioned above, several parameters that need to be used to estimate the monetary losses of zoonotic infections are poorly described. This leads to a large level of uncertainty as to what the true value of several of the components of the monetary value in Equation 1 actually are. One approach which has been used to reflect this uncertainty is that of stochastic models where the range of possible values for each uncertain parameter is repetitively sampled using a Latin Hypercube or Monte Carlo sampling technique (Majorowski et al. 2005; Knobel et al. 2005; Carabin et al. 2006; Benner et al. 2010).

The various measures of disease burden discussed above all have their positive and negative aspects. Table 4.2 provides comparisons between mortality, morbidity, QALYs, DALYs and monetary losses in terms of what is being assessed and the current availability of quality data.

Decision trees are a useful aid to help estimate frequency of infection or other outcomes when data are not readily available (Haddix et al. 2003). A decision tree is a decision support tool that uses a tree-like graph of decisions and their possible consequences, including chance event outcomes, resource costs, and utility. Figure 4.1 is an example of a decision tree for estimating the monetary burden of cystic echinococcosis in Tunisia (Majorwoski et al. 2005). A decision tree commonly consists of 3 different types of nodes:

The value at the end of each ‘branch’ is an estimation of the prevalence of that particular end-point.

Once frequency data from official sources and data obtained using tools such as decision trees have been collected, an uncertainty analysis can be conducted as described above. Sensitivity analyses can also be conducted to determine which parameters have the largest impact on the overall estimates.

An important aspect of undertaking disease burden studies is to compare the ability of alternative control strategies to reduce such burden. Public health and monetary resources are often scarce and hence intervention strategies should be designed to maximize cost benefit or cost utility of a control programme. In purely human health terms, cost utility (e.g. the cost per DALY averted or QALY saved) can be undertaken. For example, the WHO defines interventions with the best incremental cost utility ratios as those interventions costing less than US$25 per DALY averted. Incremental cost utility ratios of less than US$125 per DALY averted (WHO 1996) are considered as second tier. The impact of zoonotic diseases can be underestimated when solely human health metrics, such as DALYs or QALYs, are used since these measures ignore the often considerable economic impact of these diseases in animals. Such economic impact may be a considerable additional societal burden and can be very important in assessing the cost effectiveness of disease intervention.

Animals usually have an economic value and hence monetary losses can be calculated for animal diseases. A cost benefit analysis can then be performed in terms of money saved per money invested. The challenge occurs when attempting to obtain the overall societal benefit in terms of improvement of both human and animal health. The methods to assess the monetary burden of a zoonotic disease discussed above can be used to obtain an overall societal burden, including animal health losses. From this information, a cost benefit analysis can be conducted for comparing alternative intervention strategies. Cost sharing ideas can also be used to assess the cost utility of control where the costs are shared between public health and agricultural sectors in terms of relative benefits. If cost sharing of the control of echinococcosis in western China were to be undertaken, the cost effectiveness to the public health sector would be in the most cost effective band at less than US$25 per DALY averted. In addition, the programme would also have a positive cost benefit in terms of reduction in agricultural losses (Budke et al. 2005). Brucellosis control in Mongolia was also predicted to be highly cost effective (in terms of DALYs averted per cost invested) for the public health sector if cost sharing was undertaken with the agricultural sector (Roth et al. 2003).

Cost analyses can also be used to examine ongoing disease control programmes. For example, as purely a public health intervention strategy, the present bovine tuberculosis control programme in the UK and the Trichinella surveillance programme in the European Union (EU) are not considered as having good cost utility ratios in terms of cost per DALY averted (Torgerson and Torgerson 2010; Speybroek et al. 2010). In contrast, diseases such as toxoplasmosis result in a considerable human disease burden, but potential food safety interventions are not implemented (Kijlstra and Jongert 2009). Cost utility, in terms of DALYs averted, is one tool to drive public health policies to maximize health with the resources available.

Cysticerocis and neurocysticercosis (NCC) are caused by the taeniid-type zoonotic tapeworm Taenia solium. The parasite has a human definitive host and a pig intermediate host. Pigs contain cysticerci in different parts of the body (cysticercosis). Humans become infected with the tapeworm form (taeniasis) by ingesting undercooked pork containing T. solium cysticerci. Eggs and/or mature proglottids of the tapeworm are regularly excreted by human tapeworm carriers. Humans can also act as accidental hosts by ingesting food or water contaminated with the eggs of T. solium leading to cysticercosis and/or NCC, when cysts are located in the central nervous system. Common clinical manifestations associated with NCC include epilepsy, severe headaches, dementia, and stroke. A more complete overview of T. solium cysticercosis can be found in Chapter 51 of this volume.

The first estimates of both the monetary and non-monetary burden of disease, due to T. solium cysticercosis, were carried out in the African country of Cameroon (Praet et al. 2009). Parameters for costs associated with cysticercosis were estimated for both humans and pigs. Decision tree analysis was used to identify the proportion of the population with epilepsy due to NCC, with or without injury and with or without treatment by a medical doctor or traditional healer. Costs for visiting a physician or traditional healer, drugs, and salary losses were estimated from available data. In addition, losses due to porcine cysticercosis were estimated based on tongue inspection, with the average value on an adult pig and price reduction of pigs diagnosed with cysticercosis estimated based on available local data. Parameters for the estimation of DALYs lost were based on GBD Study estimates for disability weights associated with epilepsy and average duration of disability by age and gender. Monte Carlo methods were then used to estimate the annual socioeconomic costs and numbers of DALYs lost due to T. solium cysticercosis in Cameroon.

Based on a 3.6% prevalence of epilepsy, 50,326 (credibility interval (95% CR) 37,299–65,929) people were estimated to be suffering from NCC-associated epilepsy, which equals approximately 1% of the population. The total estimated cost associated with T. solium cysticercosis, was estimated at €19,255,202 (6,889, 048–14,754,044), with 4.7% of this value due to pig-associated losses and the remaining due to direct and indirect human losses. The average number of DALYs lost was 9.0 per thousand persons per year (95% CR 2.8–20.4).

Cystic echinococcosis is caused by the taeniid-type zoonotic tapeworm Echinococcus granulosus. It has a cosmopolitan distribution, but is most prevalent in resource-constrained livestock rearing areas. The parasite has a dog definitive host and a variety of livestock (e.g. sheep, goats, cattle) intermediate hosts. Parasitic cysts develop in infected intermediate hosts, especially in the liver and lungs, which can result in substantial livestock production losses. Humans become infected upon ingestion of parasite eggs shed in the faeces of infected dogs, with humans also developing slow-growing cystic lesions, most commonly in the liver or lungs, which can be fatal if not treated with the appropriate chemotherapeutic and/or surgical therapy. A more complete overview of cystic echinococcosis can be found in Chapter 53 of this volume.

A monetary and non-monetary burden of disease assessment was performed for Shiqu County, which is located in western Sichuan Province, China. Shiqu County has a population of approximately 63,000 with the vast majoring of inhabitants being ethnically Tibetan (Budke et al. 2005). DALYs were applied to human cystic echinococcosis cases in Shiqu County in a stochastic manner utilizing disability weights for conditions with similar symptomology (e.g. hepatic carcinoma). Based on human ultrasound prevalence, an estimated 17, 995 (95% CI 14, 268–22,128) DALYs were lost due to cystic echinococcosis (Budke et al. 2004). Since this region is also endemic for the zoonotic parasite Echinococcus multilocularis, the causative agent of alveolar echinococcosis, the number of DALYs lost was also estimated for this condition. Based on human ultrasound prevalence, an additional 32,978 (95% CI 25, 019–42, 422) DALYs were lost due to alveolar echinococcosis. This equates to approximately 0.81 DALYs lost (or 0.81 healthy year of life lost) per resident of Shiqu County due to echinococcosis, illustrating the substantial health impact of echinococcosis on this community.

In addition to non-monetary estimates of disease burden, an estimate of monetary losses, due to echinococcosis, was also performed for Shiqu County (Budke et al. 2005). Assuming that humans were treated for echinococcosis primarily via chemotherapeutic treatment, US$218, 676 (95% CI 189, 850–247,871) was estimated to be lost annually if only direct livestock-associated losses (i.e. only liver-related losses) were taken into account. This value could, however, reach as high as approximately US$1, 000, 000 if livestock CE-associated production losses (e.g. decreased hide value, decrease carcass weight, decreased fecundity) were assumed.

Brucellosis is a bacterial zoonoses with humans becoming infected via contact with an infected animal or animal product. There are a number of Brucella species that can infect humans including B. melitensis, which is commonly found in small ruminants, B. abortus, which is common in cattle, B. suis, which is found in pigs, and B. canis which is found in dogs. Brucellosis is considered endemic in livestock and humans in parts of the Mediterranean, Asia, Africa, and Latin America (Pappas et al. 2006). A more complete overview of brucellosis can be found in Chapter 7 of this volume.

The cost-utility and economic benefit of a brucellosis vaccination campaign was modelled for the country of Mongolia (Roth et al. 2003). The number of DALYs lost due to human brucellosis was estimated. In addition, monetary losses associated with human and livestock brucellosis were estimated as were the costs of a planned 10-year livestock mass vaccination campaign. The authors determined that a vaccination campaign that could reduce brucellosis transmission between livestock by 52% would result in 49, 027 DALYs being averted. It was also estimated that the proposed vaccination programme would cost US$3.3 million, but would have an overall benefit of US$26.6 million, resulting in a net present value of US$18.3 million and an average benefit-cost ratio of 3.2 (95% CI 2.3–4.4). This study went on to propose how a brucellosis prevention programme could be even more attractive to the various public health and agricultural agencies if the intervention costs were shared proportionally by the sectors which the condition impacts.

Rabies is a neuro-invasive viral disease most commonly transmitted via the bite of an infected warm-blooded animal. A more complete overview of rabies can be found in Chapter 35 of this volume.

The burden of canine-associated rabies has been evaluated for Africa and Asia in terms of monetary losses and DALYs lost (Knobel et al. 2005). First, the number of human rabies deaths was estimated using a dog-bite probability model. Costs associated with rabies were then estimated as were direct and indirect costs associated with post-exposure treatment, dog vaccination and population control, livestock losses, and surveillance. DALY estimates were conducted that not only included mortality due to rabies, but also mortality and morbidity due to the use of post-exposure treatments using crude nerve-tissue vaccines. Regional monetary and non-monetary burden estimates were estimated for both Africa and Asia using Monte Carlo sampling methods to model uncertainty.

Endemic rabies was estimated to cause 55, 000 deaths per year in Africa and Asia (95% CI 24, 000–93, 000). The annual costs of rabies was estimated at US$20, 500, 000 (95% CI 19, 300, 000–21, 800 ,000) for Africa and US$563, 000, 000 (95% CI 520, 000, 000–605, 800, 000) for Asia. This study also estimated 747, 558 (95% CI 217, 690–1, 448, 514) DALYs lost in Africa and 994, 607 (95% CI 257, 275–1,393, 125) DALYs lost in Asia due to canine-associated rabies.

Zoonotic diseases such as cystic echinococcosis, Taenia solium cysticercosis, brucellosis and rabies contribute to high levels of human morbidity, human mortality, and livestock losses in endemic regions. While control of these conditions should be prioritized, in many areas of the developing world, a lack of accurate estimates of disease burden hampers these efforts. In industrialised countries, such estimates can more easily be made and used to compare cost effectiveness of current and proposed disease control programmes.