Survey of animal bite injuries and their management for an estimate of human rabies deaths in N’Djaména, Chad




To estimate the annual human rabies incidence as a baseline prior to mass dog vaccination campaigns in N'Djaména, Chad.


Survey of animal bites, involving 50% of all healthcare providers in N'Djaména, from September 2008 to April 2009. Of 86 people exposed to a suspected rabid animal, 50% received post-exposure vaccination and a further 8% had their wound cleaned. We estimated annual incidence of bites from suspected rabid animals of 12.9/100 000 and an incidence of 0.7 human rabies deaths/100 000, resulting in 7 estimated deaths (95% confidence interval 4–10 deaths) per year in N'Djaména. 14% of bite victims sought help from veterinarians.


Closer cooperation between physicians and veterinarians warrants more effective rabies control. The high proportion (42%) of potentially exposed people without post-exposure vaccination or wound treatment necessitates urgent attention.



Estimer l'incidence annuelle de la rage humaine comme base de référence avant les campagnes de vaccination de masse des chiens à N'Djaména, au Tchad.


Surveillance sur les morsures d'animaux, impliquant 50% de tous les prestataires de soins de santé à N'Djaména, de septembre 2008 à avril 2009. Sur 86 personnes exposées à la morsure d'un animal suspect de rage, 50% ont reçu la vaccination post-exposition et 8% additionnel ont eu leur plaie nettoyée. Nous avons estimé l'incidence annuelle des morsures d'animaux suspectés enragés à 12.9/100 000 et une incidence de 0.7 /100 000 décès humains par rage, représentant une estimation de 7 décès par an (IC 95%: 4–10) à N'Djamena. 14% des victimes de morsures on recherché de l'aide auprès des vétérinaires.


Une coopération plus étroite entre les médecins et les vétérinaires est recommandée pour une lutte plus efficace contre la rage. La forte proportion (42%) des personnes potentiellement exposées sans vaccination post-exposition ou sans traitement de la plaie, nécessite une urgente attention.



Calcular la incidencia anual de rabia humana como línea de base antes de las campañas de vacunación masiva de perros en N'Djaména, Chad.


Estudio sobre la mordida de animales, involucrando un 50% de los proveedores sanitarios en N'Djaména, entre Septiembre 2008 y Abril 2009. De 86 personas expuestas a animales con sospecha de rabia, un 50% recibió una vacunación post-exposición y a un 8% más se le limpió la herida. Hemos calculado la incidencia anual de mordeduras de animales con sospecha de rabia en 12.9/100,000 habitantes y una incidencia de 0.7 muertes por rabia humana/100,000 habitantes, resultando en 7 muertes estimadas (IC 95% 4 – 10 muertes) por año en N'Djaména. Un 14% de las víctimas de mordeduras buscó ayuda de veterinarios.


Una mayor cooperación entre médicos y veterinarios garantizaría un control más efectivo de la rabia. La alta proporción (42%) de personas potencialmente expuestas sin una vacuna post-exposición o tratamiento de la herida requiere de una atención urgente.


Rabies remains an important zoonosis in many developing countries – an estimated 55 000 human deaths occur each year (Knobel et al. 2005). Most human rabies cases that result from rabid dog bites are reported in Asia (e.g. Indonesia and India); however, in Africa, there is vast under-reporting of rabies in dogs (Kitala et al. 2000) and in humans (Cleaveland et al. 2002).

Without post-exposure prophylaxis (PEP), a person bitten by a rabid animal can develop clinical disease, which is fatal. Providing PEP – repeated injections of rabies vaccine and immunoglobulins to the animal bite victim as soon as possible after exposure, combined with local treatment of the wound (immediate mechanical cleaning of the wound with water and soap followed by disinfection) – can prevent the development of rabies (WHO 2010). However, PEP, especially the immunoglobulins, is expensive and often not available in low-income countries.

In N'Djaména, the capital city of the Chad, the annual incidence risk of rabies in dogs in 2002 was 1.4 per 1000 unvaccinated dogs. One rabid dog or cat exposed 2.3 persons on average; dogs accounted for 95% of exposures (Kayali et al. 2003a). Zinsstag et al. (2009) have used a dog–human mathematical transmission model combined with an economic assessment (based on previously collected data during surveys in N'Djaména) to show that the most cost-effective intervention against dog rabies is to combine parenteral dog vaccination campaigns and PEP, which leads to a break-even in costs after 5 years compared with costs of PEP alone.

To date, human rabies deaths had never been estimated in N'Djaména. We found 3 reported human rabies cases in Chad in the RABNET database (hosted by WHO) – and they were all for 1997. In Chad, there is no capacity for human post-mortem rabies diagnosis by immunofluorescence. Clinical human rabies cases can be misdiagnosed. In Malawi, 3 of 26 fatal cases initially assigned to cerebral malaria were diagnosed as rabies during post-mortem examinations, as were 14 of 133 fatal cases assigned to central nervous signs (Mallewa et al. 2007).

Cleaveland et al. (2002) found that active detection of human rabies deaths is difficult due to low incidence and the need to set-up specific detection studies such as collection of verbal autopsy data from household surveys. Passive surveillance may be insufficient, leading to vast under-reporting of human rabies cases. However, animal bites can be surveyed given their rather high incidence and the likelihood of victims to seek professional care. Cleaveland et al. (2002) used a probability decision tree to estimate human mortality from information provided by animal bite victims. After validation of the approach in field studies, they showed that, in rural Tanzania, the true incidence of human rabies was 10–100 times higher than the reported incidence.

In N'Djaména, mass dog vaccination should show a reduction in human exposure to suspected rabid animals. However, human clinical rabies in N'Djaména can only be estimated. We aimed to adapt Cleaveland et al.'s (2002) probability tree model to the urban setting of N'Djaména, Chad, to estimate indirectly human rabies incidence from an animal bite survey in health facilities and describe the treatments that animal bite victims received.

Materials and methods

Collaborating healthcare providers

Healthcare providers for animal bite victims in 10 municipal city subdivisions (arrondissements) of N'Djaména were contacted and registered as follows: (i) hospitals, treatment and health centres; (ii) health clinics; (iii) pharmacies and pharmaceutical shops and (iv) veterinary clinics. Bite victims sought counselling and first-aid treatment at any of these healthcare providers. We did not differentiate between private and public health centres or clinics. Our coverage of operational private healthcare providers (clinics, pharmacies and pharmaceutical shops) was compared with an existing list of authorised and active healthcare providers in N'Djaména (Gami 2008) and to the reported number of operational health centres in the Health Statistical Report of 2008 (Ministère de la Santé Publique du Tchad 2009). The heads of 10 municipal arrondissements were visited in order to inform them on the study and with the request to distribute the information to their heads of districts.

Data collection

We developed a data sheet (in French) to collect data on the age of the bite victim, site of the bite injury (indicated on drawings of an adult and a child), type of treatment (including post-exposure vaccination) and whether the animal was suspected by the bitten person of being rabid and/or being vaccinated. We also asked about wound treatment and whether the bite victim was referred from or to another healthcare provider, to avoid double inclusion. The healthcare providers completed the information sheet together with the bite victim.

From September 2008 to April 2009, all collaborating healthcare providers were visited every other week to collect the forms. In the same period, all dogs put under observation at the veterinary clinic of N'Djaména were recorded. If a suspected rabid animal died or showed signs of rabies (the latter were euthanised), its brain tissue was tested at the Laboratoire de Recherches Vétérinaires et Zootechniques de Farcha (LRVZ) with the direct fluorescent antibody test – the WHO-recommended gold standard of rabies diagnosis (Dean et al. 1996; World Health Organization 2005).

Ethical considerations

We obtained ethical clearance from the Ethics Committee of Basel (EKBB) and the Chadian Ministry of Health. Adapted leaflets and posters (Durr et al. 2009) were handed out to bite victims, which provided information on correct management of a bite wound, a contact number of a physician and the information that post-exposure vaccination is available for free at the Centre de Support en Santé Internationale, where a stock of rabies and tetanus vaccines was stored for the study. Note that no antirabies immunoglobulins were available in Chad.

Data analyses

Data were entered into Microsoft Access and analysed in Stata 10 (StatCorp LP). Age was categorised regarding age-dependent treatment procedures, and height at age in adults (≥13 years) and children (<13 years). The sites of the reported bite injuries were categorised as (i) head (head, face and neck); (ii) arm (hands, arms and shoulders); (iii) trunk; and (iv) legs. When more than one bite location was registered, the site with the higher probability of developing rabies (head or upper extremity) was retained (Hattwick & Gregg 1975). The reported severity of the wound was classified as deep wound, lacerated wound, superficial wound or scratch. We considered animals as vaccinated against rabies if the reported vaccination took place less than 12 months prior to the bite. Chi-square test (Fisher's exact test where appropriate) was used to assess possible associations between age, site of the bite, severity of injury and sex of bite victims.

To estimate the incidence of human cases of rabies, we used the decision tree designed by Cleaveland et al. (2002). The starting point was P1, the probability of a suspected animal being rabid. P2–P5 were defined by the proportions of bite injury sites. P6–P9 were the probabilities of developing rabies after a bite from a rabid animal depending on its location. P10 was the probability of receiving post-exposure vaccination (Table 1). Finally, the probability of dying of rabies (Pdeath) after being bitten by a rabid suspect animal was calculated as follows:

Table 1. Parameters to derive Pdeath and Ndeath estimated with own data P1–P5, P10 and psuspected as well as P6–P9 from literature
Variable Estimate
  1. *The minimum value was taken from Cleaveland et al. (2002), the maximum value from our data (12/14) of suspected animals tested in Chad.

P1Suspected dog being rabidPert: minimum = 0.68, mode = 0.75, maximum =  0.86
P2Bite injury to the headBinomial: p = 0.03 n = 86
P3Bite injury to the armsBinomial: p = 0.35 n = 86
P4Bite injury to the trunkBinomial: p = 0.05 n = 86
P5Bite injury to the legsBinomial: p = 0.57 n = 86
P6Development of rabies following a bite injury to the headPert: minimum = 0.3, mode = 0.45, maximum = 0.6
P7Development of rabies following a bite injury to the armsPert: minimum = 0.15, mode = 0.275, maximum = 0.4
P8Development of rabies following a bite injury to the trunkPert: minimum = 0, mode = 0.05, maximum = 0.1
P9Development of rabies following a bite injury to the legsPert: minimum = 0, mode = 0.05, maximum = 0.1
P10Probability of receiving post-exposure prophylaxis if bitten by a suspected rabid dogBinominal: p = 0.5, n = 86
p suspected Probability of a suspected anima bite.Binomial: p = 0.154, n = 557
display math

Pdeath considered that a proportion of people bitten by a suspect rabid animal had received post-exposure vaccination and thus would not develop rabies. The total number of deaths due to rabies per year was calculated as

display math

where i was the incidence of suspected animal bites per year per 100 000 inhabitants, and Q the total population at risk. The population at risk Q was taken from the census of 2009 as Q = 993 447 people (Ministère de l'Economie et du Plan 2009). The annual incidence i of suspected animal bites was calculated from the observed proportion of suspected bites (psuspected) multiplied with the number of suspected animal bites (nbites) during 8 months and adjusted to 1 year and 100 000 people at risk as given below.

display math

We have taken a binomial distribution for P2–P5 and a Pert distribution for P1 and P6–P9 (Table 1). The Pert distribution is often used on parameters with normal-shaped distribution but with defined upper and lower limits, as is the case with expert opinions (Malcolm et al. 1959). The number of suspected animal bites bbite was expressed as binomial distribution (Table 1). Pdeath and Ndeath were estimated by Monte Carlo simulations with 10 000 iterations using ‘Ersatz’ software (EpiGear International Pty Ltd Brisbane, Australia, 2009-2012,


Ninety public and private healthcare providers in N'Djaména were contacted, and all agreed to be enrolled in the study. Because bite victims sometimes contacted a veterinarian, we also included 5 veterinary clinics in the survey. These received 77 of the 557 bite victims registered (Table 2). In 2008, a study by Gami established in parallel to our survey a list of private health centres/clinics and pharmacies in N'Djaména. According to this list, 50 private treatment/health centres and clinics were operational in N'Djaména. We enrolled 12 private health centres and 19 private health clinics, representing 62% of private healthcare facilities. There were 28 pharmacies and 71 pharmaceutical stores, of which we contacted 17 pharmacies (61%) and 10 pharmaceutical stores (14%). We also enrolled 27 (71%) of the 38 public health facilities (6 hospitals and 32 health centres) (Ministère de la Santé Publique du Tchad 2009). Overall, we enrolled approximately 50% of the public and private healthcare providers.

Table 2. Enrolled healthcare providers and numbers of collected reports of animal bite victims per type of provider in N'Djaména, Chad
 N enrolled healthcare providersN providers submitting at least one reportTotal reports submitted
Hospitals, treatment and health centres3929262
Health clinics19 6 56
Veterinary clinics 5 2 77

We collected a total of 557 completed data sheets from 62 healthcare providers (Table 2). Two bite victims were registered twice, and the data of their first visit were excluded from analyses. The home Arrondissement was known for 528 bite victims. Figure 1 shows the distribution of the healthcare providers and reported animal bite injuries per 100 000 inhabitants in the 10 Arrondissements of N'Djaména, with the highest rates in Arrondissements 6 and 7. More records per month were collected during the hot, dry season months (March and April), than during the dry season months (September–February) (Figure 2).

Figure 1.

The distribution of the healthcare providers (all enrolled: light grey bars and those with reports: dark grey bars) and reported animal bite injuries per 100 000 inhabitants (diamonds) in the 10 Arrondissements of N'Djaména.

Figure 2.

Number of reports received per month (Dark grey: total number; light grey: reports of bites from suspect rabid animals) and average temperatures (dots).

Of the biting animals, 505 were dogs, 29 cats, 20 monkeys and 1 was a donkey. A total of 219 dogs (43%), 3 cats (10%) and 12 monkeys (60%) were reportedly vaccinated against rabies, 62.9% within the past 12 months. 40.7% of all animals were not vaccinated, and for 13.6% the status was unknown. The victims reported that 16% of dogs, 7% of cats and 5% of monkeys were suspected of being rabid (n = 86 within the eight months of the survey). Another 74.9% of animals were declared non-suspicious and, for 9.7%, no declaration could be recorded. 39% of the suspected dogs were animals unknown to the bite victim; this proportion was lower (12%) among non-suspected dogs (Table 3). An animal biting a child was significantly more often declared suspicious than an animal biting an adult (21.2 vs. 14.3%). Reports on further bite victims were significantly more frequent for rabies-suspected than non-suspected dogs (49% vs. 16%), and the mean number of bite victims was greater (2.1 vs. 1.3 people).

Table 3. Characteristics of biting animals (with details on dogs and cats) recorded in the animal bite survey in N'Djaména
 All biting animalsaDogsCats
SuspectNon-suspectNot categorisedSuspectNon-suspectNot categorisedSuspectNon-suspect
N % N % N % N % N % N %  N % N %
  1. a

    505 were dogs, 29 cats, 20 monkeys and 1 was a donkey.

Vaccinated < 1 year
Yes 4 713334 117 4 712134 125 0 0 2 9
Sex of biting animal
Male68813338018866480308831594 21001248
Relation of victim to animal
Own animal151714134 0 0131611230 0 0 1 502080
Neighbours animal384422053 939374521357 528 0 0 2 8
Unknown animal3338 531314613239 46121372 1 50 312

During the study period, 27 suspected rabid animal carcasses were brought for examination to the Laboratoire de Recherches Vétérinaires et Zootechniques de Farcha. A total of 23 dogs had bitten 37 persons (an average of 1.6 persons), and four cats had bitten 6 persons (an average of 1.5 persons). Bite victims of 75% of these suspected animals had been registered as a bite victim in this study. A total of 13 carcasses were confirmed rabid (12 dogs, 1 cat), 4 tested negative and 11 could not be tested due to decomposed brain samples. Of all tested samples, 76% (95% CI 50–93%) were confirmed rabid (86% of dogs, 50% of cats).

38% of the bite victims were children (37% girls, 63% boys) and 62% were adults (42% women, 58% men). The median age was 18 years, with a range between 2 months and 79 years. All bites to the head were in children. The bite location differed significantly between children and adults (with children more frequently being bitten in the upper body), but not between males and females (the higher frequency of male bite victims is reflected in all bite locations). We did not see significant differences between the reported severity of the bite and its location (Table 4).

Table 4. Bite locations and their distribution between age classes, sex and severity of the bite
CategoryAll locationsHeadArmTrunkLegFisher's exact test P-valueb
N % N % Category N % Category N % Category N % Category
  1. a

    Of six bites to the head, for one victim, the age was missing.

  2. b

    For differences between bite locations.

>=13343620 05446164427369
Male331605 837361236423059 0.79
Female223401 174639133616341
Deep wound155282 333933102810426 0.92
Lacerated wound111201 172622 514 7920
Superficial wound177322 333428.5133612833
Scratch111201 172016.5 822 8221

One-third of the bite victims (182 of 550 with data) went to a healthcare provider on the same day (with a range of 0–120 days and a median of 1 day). 50% of the 86 people bitten by a suspected rabid animal received post-exposure vaccination (and among the 43 vaccinated, 1 person also had the wound cleaned at the healthcare provider). These proportions were significantly lower among people bitten by a non-suspected animal (Table 5).

Table 5. Comparison of wound management between bite victims of suspected and non-suspected rabid animals and per severity of wound
 Suspect, N = 86Non-suspect, N = 416Comparison suspect–non-suspect
Total %Deep woundLacerated woundSuperficial woundScratchTotal %Deep woundLacerated woundSuperficial woundScratchChi-square test P -value
Post-exposure vaccination5018.69.3148.<0.001
Tetanus shot60.520.<0.001
No treatment75.801.
Hospitals, treatment and health centres45.410. 
Health clinics11. 
Veterinary clinics18.616.32.30013. 

Incidence of human cases of rabies

With Monte Carlo simulation, we derived a probability of dying of rabies after being bitten by a suspected rabid animal [Pdeath] of 5.3% (95% CI 3.2–7.8%) in the city of N'Djaména.

The annual incidence of bites from suspected rabid animals was derived from the 86 people bitten by a suspect dog within 8 months, leading to an incidence of 12.9/100 000 over 12 months. Note that we recorded bites from suspected victims in both dry and rainy season. The annual total number of human rabies deaths [Ndeath] in N'Djaména was calculated as 6.6, thus 7 deaths (95% CI 3.9–10.4) or, expressed as annual human rabies incidence, 0.7 per 100 000 (95% CI 0.4–1.05).

A sensitivity analysis of the stochastic model calculated Spearman's rank correlation coefficients (RCC) for all parameters. The probability of death and the number of death were most sensitive (RCC >0.35) to the probability developing rabies from an arm injury, the probability of receiving PEP and the probability of being bitten in the arm.


Allmost all (97%) bite victims of a suspect rabid animal in N'Djaména were exposed to a dog, which is in agreement with observations in other cities (Wandeler et al. 1993) and previous studies in N'Djaména (Kayali et al. 2003a). Most of the bite reports were from Arrondissements 6 and 7, which matches with observations by Mindekem (2003), who reported that Arrondissements 6 and 7 had higher dog:person ratios than the others.

Several healthcare providers stated that more animal bites are seen during the hottest months, because animals are irritated by the heat and as a consequence bite more easily. Indeed, we received more report forms of bite victims in March and April (the very hot months), but the proportion of rabid suspected bites remained constant over the 8 surveyed months.

Animals were more often classified as suspect when they bit to the head or the upper extremity. Consistent with other studies, likely due to their height and their more careless approaching of animals, children were bitten twice as often in dangerous locations (head, shoulders or arms) than adults and thus were at higher risk of being infected and dying of rabies, because the probability of developing rabies is highest after a bite to the head or to the arms (Hattwick & Gregg 1975).

In our urban study, 16% of all dogs inflicting bite injuries were suspected of being rabid, which is much less than the 77% found in rural Uganda (Fevre et al. 2005) and the 75% in rural Tanzania (Cleaveland et al. 2002) – although only 35% of the owned dogs in our study were vaccinated. It seems more likely in an urban than a rural setting that bite victims, also those without suspected rabies exposure, seek health care (we cannot further quantify the proportions of all animal bite victims seeking professional help).

We compared our results to the incidence estimate of 1.4 per 1000 dogs per year and 2.3 exposed persons per animal by Kayali et al. (2003a) using a total dog population estimate of 23 500 dogs in N'Djaména in 2001 (Mindekem et al. 2005). Thus, 75 persons per year were exposed to a suspected rabid animal in N'Djaména, resulting in an incidence of death (while considering that 50% were protected by post-exposure vaccination) of 0.5/100 000 (for a human population of 770 000 in 2002). The estimated human death incidence from our animal bite study (0.7/100 000) is 1.4 times higher than that calculated from the dog rabies incidence in 2001.

No human rabies cases are reported in Chadian official health records, and we thus cannot calculate an under-reporting of human rabies cases in N'Djaména. Our indirect estimate of human rabies incidence indicates that human rabies cases likely occur in N'Djaména and conflicts with the fact that no human rabies are reported in official health records. Given the endemic stability of rabies in N'Djaména (Zinsstag et al. 2009), we extrapolated our data to an estimate of an annual exposure incidence. Our estimate of the proportion of deaths among exposed people is comparable to Cleaveland et al. (2002) (5.3% in this study and 4.4% in Tanzania), reflecting only slight differences in bite localisations.

The coverage of enrolled healthcare providers in our study was approximately 50%. Therefore, we assume that the true number of people exposed to suspected animals could be higher (i.e. double) than that recorded, but the effective coverage of healthcare providers is not known. Among the bite victims of animals brought for rabies examination to the national veterinary laboratory during the same study period, we registered 75% at a healthcare provider. Assuming that our study covered 75% of the effective bite victims, human rabies mortality could be as high as 0.93/100 000. However, we also saw that 32% of rabid suspect animals were animals not known to the bite victim (and their ownership status was not known). For comparison, Kayali et al. (2003b) found that 85% of the rabid dogs were owned. We thus likely have underestimated the exposure frequency, but the reported 86 bites from suspected animals among 557 bite victims were underestimated to a lesser extent. We conclude that the estimate of 7 human rabies deaths per year in N'Djaména is conservative.

Only 50% of the victims of suspected rabid animal bites received post-exposure vaccination. In Chad, as in many developing countries, repeated vaccinations are given as sole treatment and not combined with injections of antirabies virus immunoglobulins. Post-exposure vaccinations often remain unavailable and are expensive (approximately 16 USD/vaccine, five vaccines needed). Where post-exposure vaccination is not available, it is even more crucial to clean the wound properly, but of those who were potentially exposed and did not receive vaccinations (43 victims), only 7 had their wound cleaned by the healthcare provider, which requires urgent attention.

Of note are the 14% of bite victims who have first sought health care at a veterinary clinic. We found in this group that a greater proportion of victims bitten by a suspected rabid animal received post-exposure vaccination than of bite victims who attended other healthcare providers (62% vs. 50%). This might be due to the fact that the human vaccine was available at the largest veterinary clinic in N'Djaména, but also due to a better information about rabies of veterinary than of medical staff.


This study suspects the presence of human rabies cases in N'Djaména based on an animal bite survey, which contrasts with officially reported rabies cases for animals, but not for humans. This is seen in several Sahelian cities. New approaches to estimate the human rabies burden should use the available information on rabies in animals. This survey will serve as a baseline to assess whether the government-funded mass dog rabies vaccination campaign starting in 2012 will have a measurable effect on communities in terms of lower animal bite frequency, post-exposure vaccinations and human rabies death incidences. Closer data exchange between health and veterinary clinics would benefit from a fuller description of the status of rabies in Chad. Given the estimate of 7 human deaths due to rabies each year in N'Djaména, and the broad fear of rabies in the population, mass dog vaccination campaigns covering the whole city are warranted.


Our great appreciation goes to the healthcare providers and the animal bite victims in N'Djaména, for their active participation in our study. We thank the Swiss Federal Veterinary Office for the financial support, and Prof. Christoph Hatz and Dr. Kaspar Wyss for their mentoring of the medical aspects.