Introduction Rabies is an acute viral infection which causes horrifying neurological symptoms that inevitably result in death. Although human rabies encephalitis remains untreatable [1], the infection is entirely preventable, both by post-exposure prophylaxis (PEP) of bite victims, and by population-level vaccination of the zoonotic reservoir, which across most of Africa and Asia is the domestic dog [2]. Modern cell culture vaccines used in combination with rabies immunoglobulins are virtually 100% effective in preventing human deaths if administered promptly to rabies-exposed patients following appropriate wound management [3] and mass vaccination of domestic dogs has successfully eliminated or controlled domestic dog rabies in many parts of the world [4],[5]. It is therefore inexcusable that an estimated 55,000 human deaths from rabies occur annually [6], of which over 99% are in developing countries where the disease is endemic in domestic dog populations [7]. Recent estimates of human rabies mortality are based upon a probability decision-tree model [6], because current surveillance systems have been shown to substantially underreport the number of deaths from rabies. For example, in Tanzania more than 100 human rabies deaths are estimated to occur for each officially reported case [6]. Hospital studies further suggest that clinical diagnosis of human rabies may be hindered by confusion with common neurological syndromes, such as cerebral malaria [8]. These and other studies on rabies incidence and exposure risk rely on bite victims reporting to hospital, yet not all rabies-exposed individuals seek medical attention. To investigate the validity of methods being used to estimate the burden of rabies we established a contact-tracing study. Data collected using these methods provides a more comprehensive picture of the reality facing communities in regions where canine rabies is endemic. Using these data we quantify the risk of disease and exposure and attempt to understand why human deaths from canine rabies still occur and thus how this number can be reduced. Methods Contact-tracing Data was collected from two rural districts in northwest Tanzania: Serengeti, which is inhabited by multi-ethnic, agro-pastoralist communities and high-density dog populations, and Ngorongoro, which is inhabited by low-density pastoralist communities and lower density dog populations. Contact tracing of potential rabies-exposures was initiated using data from hospitals and medical dispensaries on patients with animal-bite injuries, and case reports from livestock offices and community-based surveillance activities. Visits were made to investigate incidents that occurred between January 2002 and December 2006 involving potentially suspect rabid animals. Interviews were conducted to assess the case history and identify the source of exposure and other contacts if known. The same procedure was followed for all resulting exposures and preceding cases where identified, and UTM coordinates were recorded at each household and at the location of the exposure event (where possible). Interviews were conducted by veterinary or livestock field-officers, often with a community leader in attendance. This created an active local reporting network. Animal cases were diagnosed on epidemiological and clinical criteria adapting the ‘six-step’ method through retrospective interviews with witnesses [9]. Wherever possible brain samples from animals that caused bite injuries were collected and tested for case confirmation [10]. Questionnaires A structured open-ended questionnaire was administered to bite victims at 3 designated district hospitals (in Magu, Misungwi and Tarime, n = 166) to obtain information on intervals between exposure and reporting to hospital for PEP, and ways used to raise funds to pay for PEP. Information was collected on household socioeconomic status, using indicators sensitive to local determinants of wealth, previously identified through Rapid Rural Appraisal approaches [11]. Specifically numbers of cattle and housing quality were chosen as independent wealth indicators because individuals may own many cattle and hence be considered to be wealthy but they may not necessarily own “modern” houses. Individuals with houses constructed from cement/baked bricks, which have cement floors and corrugated roofs were categorized as belonging to high socioeconomic status and those owning houses constructed from other materials were classified as low socioeconomic status. Regardless of housing quality, individuals owning >50 heads of cattle were categorized as high socioeconomic status; those with 97% of animals that caused bite injuries were classified as suspected rabid (648) or normal (406). The status of animals that bit the remaining 2.5 percent (26) of cases visited was unclear. Approximately 75% of samples from suspected rabid animals tested positive, indicating that recognition of rabies is accurate and that classification using the case history description is valid [12]. Over twenty-five percent of visited cases bitten by suspected rabid animals (180) were identified through contact tracing alone because the victim did not seek medical attention. Of 1322 bite injury records from medical facilities over the same period, 57% (760) were successfully traced, 9% (118) were not visited because the record indicated the animal was healthy and the remaining 444 cases were either impossible to trace, not present to interview, or have yet to be visited (139 were from 2006). At least 50 of these exposures were by suspected rabid animals. Conservative estimates suggest around 63/100,000 people in Serengeti and 17/100,000 in Ngorongoro are bitten by suspected rabid animals annually. Including animals of undetermined status raises those figures to 100 and 30 exposures/100,000 respectively. The risk of being bitten by a suspected rabid animal varied through time (approaching 150/100,000 during the epidemic peak), but was consistently higher in Serengeti, the more populated district (Table 1). Most suspected rabies exposures were due to domestic animals (89%), particularly dogs (Table 2). A higher proportion of bites by suspected rabid animals were from wild animals in Ngorongoro district compared to Serengeti district (∼20% versus 100,000 Tsh (∼US$85) to free (for limited periods), although courses were typically 75,000 Tsh in Ngorongoro district (five doses) and 30,000 in Serengeti (3 doses), in comparison to monthly per capita expenditure and per household expenditure of 8,538 Tsh and 52,649 Tsh respectively in 2001 [11] (although in 2008 prices are now approaching ∼30,000 Tsh per dose). However, the probability of receiving PEP following exposure was very similar in the two districts (0.70 in Serengeti versus 0.68 in Ngorongoro). Rabies immunoglobulins were not offered to any bite victims. Most people who attended a medical facility did so shortly after exposure, but there was considerable variance in delays before receiving the first dose of PEP (Fig. 3); at least 25% of courses were started more than one week later. Distance from the nearest medical facility and socioeconomic status were both significant predictors of delays in PEP delivery (p 17%) were initially diagnosed with cerebral malaria, but as symptoms progressed and when the history of a bite was discovered, the diagnosis was changed to rabies. Exposed individuals who developed rabies generally lived further from medical facilities than those who did not, although this was not statistically significant (p = 0.08). Risks of (and trauma from) human-to-human transmission are also not inconsequential; three rabies-infected individuals (>10%) bit a family member and a fourth hit her mother, apparently due to disease-induced changes including aggression. Additionally a twenty-year old woman died of tetanus following a suspected rabid dog bite. She developed symptoms of tetanus before completing her third dose of PEP. Because she was pregnant it was assumed that she must have been previously vaccinated against tetanus. Discussion We investigated how risks of rabies exposure and onset of disease vary according to epidemiological and socioeconomic determinants and present evidence-based recommendations to reduce these risks in settings where canine rabies is endemic, addressing perspectives of both the health provider and patient [13]. Numbers of suspected rabies exposures varied considerably through time and across districts. The temporal variation was presumably due to the tendency of the disease to fluctuate on a timescale of approximately five years [4]. Assuming constant numbers of exposures per year may therefore be misleading if used as a basis for provisioning PEP. We suggest that exposure incidence, when used for indirect estimation of the burden of rabies, should be averaged over at least a five-year period because of inherent temporal variability. This study lasted five years, spanning one complete epidemic cycle and therefore the likely range of exposures through time. Our upper estimate of annual incidence of bite-injuries by suspect rabid animals in agro-pastoralist communities (100/100,000) is very close to previous estimates (104/100,000) [14]. However, vaccination of dog populations during the study substantially reduced the number of exposures and probably heightened awareness of the disease within study communities (several rabies-exposed individuals sought PEP after being interviewed). Our estimates therefore probably underestimate countrywide incidence, because mass dog vaccination campaigns are not routinely conducted across most of Tanzania. Heightened awareness may similarly explain our relatively low yet comparable estimates of annual rabies mortality (1.5 and 2.3/100,000 in Serengeti and Ngorongoro districts respectively) compared to previous estimates (4.9/100,00) [14]. The higher risk of exposure in the more populated areas was likely due to the higher incidence of rabies and longer duration of outbreaks (and less frequent fade-out) in larger domestic dog populations (dog density: ∼11.4/km2 in Serengeti district versus 4.2/km2 in Ngorongoro district, which is close to the critical threshold for persistence ∼4.5/km2) [15]. More abundant wild carnivores in Ngorongoro explains the high proportion of suspected rabies exposures caused by wild animals in the district [12]. Nevertheless, only the African 1b domestic dog associated rabies strain has been identified from the sequenced isolates (>50) in 9 species over the study area and evidence points to domestic dogs as the only population capable of rabies maintenance [16]. Control efforts should therefore be targeted towards domestic dog populations but education efforts should stress that the bite of any mammal can transmit rabies and should be treated promptly. One of the greatest challenges for controlling canine rabies has been raising the priority of the disease. It is widely recognized that rabies is grossly under-reported even though it is notifiable and the lack of accurate figures has rendered rabies a low public health and veterinary priority. Previous attempts to quantify the burden of rabies have relied upon hospital records and have pointed out the need to verify their methods and conduct active case detection studies [6],[14],[17]. The validity of these indirect assessments is dependant upon key assumptions, such as the assumption that all rabies-exposed patients are recorded in hospital records. We show that at least 20% of all rabies exposures do not seek medical attention. Our estimates of rabies mortality are still comparable to model predictions, probably because the proportion of rabies-exposed individuals that received PEP, if medical attention was sought, was higher than during the previous study (0.86 versus 0.56) [14], though still unacceptably low. Thus, our contemporary data suggest indirect estimates of rabies-exposures and mortality based on well parameterized decision tree models are reasonable, but could be improved by accounting for the fact that not all bite victims seek PEP. Our results highlight key aspects of health services that could be targeted to improve the treatment of patients reporting with animal-bite injuries. For instance, many bite victims had to travel to hospitals in neighboring districts (sometimes several) to obtain PEP, prolonging delays before PEP delivery, increasing the risk of disease and incurring considerable costs on victims and their families. Improved surveillance combined with timely reporting and centralized responses for vaccine distribution could prevent PEP shortages and reduce the need to travel to alternative clinics. Animal-bite injury records are an accurate indicator of rabies exposures (exposure status is not regularly recorded) and therefore have potential to be used as a surveillance tool, but to be of most value, records ought to be collated over catchment areas spanning several districts. The number of cases where patients reporting to medical facilities were misadvised is also unacceptable indicating that medical personnel require greater training in recognizing cases of rabies exposure and in judicious administration of appropriate PEP. The risk and burden of rabies falls disproportionately on the most vulnerable sectors of society: children and particularly those in marginalized pastoralist populations. The high proportion of childhood rabies deaths, a well-documented statistic [18],[19], increases the disability-adjusted life years lost and therefore the burden of the disease [20]. Similarly those that live furthest from health facilities and are in lower socioeconomic classes undergo longer delays before receiving PEP which increases the risk of developing rabies. The high costs of PEP contribute to this problem, as many people must sell livestock or other possessions to raise funds. But many families spend even larger amounts of money trying to obtain treatment for a family member with clinical rabies than the total cost of preventative PEP, suggesting that the danger posed by the bite of a rabid animal is not fully appreciated. The substantially higher risk of developing rabies following exposure in Ngorongoro compared to Serengeti district cannot be explained by the probability of seeking medical attention. A plausible explanation is the adequacy of first aid delivered after a bite. Immediate washing of the wound considerably reduces the risk of disease progression [21], and may be practiced more in Serengeti than Ngorongoro. We do not have data to test this, but 7 of 14 deaths in Ngorongoro were children bitten whilst they were alone herding cattle, likely in remote areas, who probably did not administer appropriate first aid. Contact tracing uncovered many exposures and deaths not recorded in official sources, showing that the proportion of people exposed to rabies that seek PEP is unacceptably low. This results primarily from patients' lack of knowledge, or resources (or ability to mobilize them) suggesting that education to raise awareness about rabies prevention, wound management (particularly immediate flushing of the wound with any available liquid), and prompt PEP administration, could substantially reduce numbers of rabies deaths. Zoonotic diseases are often neglected because the major burden falls within the health sector, yet the veterinary sector is usually responsible for their control. The two sectors typically operate independently and resources available to the medical sector are often much greater than those in veterinary departments. In reality rabies is a shared problem that can only be tackled by a multidisciplinary approach. Without laboratory confirmation and accurate diagnosis of animal rabies, public health authorities will not recognize rabies prevalence and without accurate information on human deaths and exposures from public health authorities the disease will not receive the attention it requires from the veterinary sector. One example that is a pervasive problem, evident in this and other studies [22], is the lack of diagnostic confirmation of human cases even though samples can be collected non-intrusively by supraorbital needle biopsy. Our results support previous findings that clinical diagnosis alone underestimates rabies incidence because of confusion with other neurological infections [8]. Nonetheless, the data we present provides a detailed picture of human rabies exposures and deaths during the last five years in a rural region of Tanzania; it leads to a number of practical recommendations for preventing future deaths ( Box 1) which should be valuable to medical practitioners and veterinarians alike. Misdiagnosis, incomplete understanding of how rabies is transmitted (for medical and veterinary workers and the general public), poverty and the lack of appropriate affordable treatment all result in needless human deaths. We highlight the practical problems that face people living in regions of endemic canine rabies and the tragically high prevalence of this disease which can be entirely controlled given sufficient political will. Box 1. Policy recommendations for reducing human deaths in canine rabies endemic regions In accordance with the Regional East African Community Health (REACH) initiative's mission to access, synthesize, package and communicate evidence required for policy and practice to improve population health and health equity (http://www.who.int/alliance-hpsr/evidenceinformed/reach/en/index.html) we provide recommendations for reducing human deaths from rabies following exposure. Awareness needs to be raised about the importance of immediately washing animal-bite wounds and reporting rapidly to medical facilities for PEP (irrespective of the size and severity of injury). Supply and distribution systems for PEP should be reviewed because shortages are frequent, regional disparities exist in prices and regimen, and treatment cannot always be accessed during evenings and weekends. Mechanisms should be sought to reduce the price of PEP and enable early initiation of treatment for patients who may be unable to quickly access sufficient funds to pay for PEP (e.g. use of economical intradermal PEP regimens [23] for multiple patients who present simultaneously could be evaluated) Improved training is needed for medical personnel to ensure awareness about the serious nature of rabies exposures and enable judicious decisions about PEP administration. Prophylaxis should be initiated immediately unless the patient is reporting more than ten days after exposure and is completely certain the biting animal is alive and healthy. Similarly PEP can be discontinued if the animal's good health can be established at subsequent hospital visits. Collaborative (veterinary and medical) programs should be established to control and eliminate rabies from domestic dog populations and improve surveillance and diagnosis in both animal and human populations. Supporting Information Alternative Language Abstract S1 Translation of the Abstract into Swahili by M. Kaare. (0.01 MB PDF) Click here for additional data file.
