|
nif:isString
|
-
RabiesEcon is a spreadsheet-based tool (S1 Appendix) that incorporates a mathematical (deterministic) model of dog-dog and dog-human rabies transmission to estimate dog and human rabies cases averted, and the cost per human rabies death averted and per year of life gained (YLG) due to dog rabies vaccination programs. We used RabiesEcon to estimate the cost-effectiveness of dog rabies vaccination programs in an illustrative East Africa human population of 1 million in a mixture of urban and rural settings. Because there are insufficient data from a single country in Africa for every input in RabiesEcon, we used data from a number of African countries, primarily Chad, Malawi and Tanzania (Table 1). We estimated, based on published measurements of dog ownership in East Africa [2, 20], that the modeled population has approximately 82,000 dogs (36,500 in urban setting, 45,700 in rural setting) (Table 1). We chose East Africa as an example because recently published studies demonstrated the feasibility of conducting dog rabies vaccination programs in this region [20–23]. We built RabiesEcon to include a separate sub-model for each sub-region, urban and rural. Each sub-model calculates the number of dog rabies, human deaths and impact of dog vaccinations and PEP for that sub-region, using data relevant to the sub-regions (Table 1). The results from each sub-region are then summed and presented as a total for the entire area being studied.
Table data removed from full text. Table identifier and caption: 10.1371/journal.pntd.0006490.t001 Main demographic and epidemiological model inputs to estimate the cost-effectiveness of an illustrative dog rabies vaccination programs in East Africa. a Calculated using the RabiesEcon tool. Please see Supplemental material.b The numbers of humans-per-dog for Blantyre were obtained from Gibson et al. [20]; the estimate for rural areas was based on Knobel et al.’s estimate for Africa [2].c The urban dog birth rate was obtained from a dog population household survey in N’Djamena, Chad [7]. For the rural scenario, we used data from Machakos District, Kenya [28].d Life expectancy at birth was 3.5 and 2.4 years for male and female dogs in Kenya [29].
We compared three different dog rabies vaccination options: no vaccination, annual vaccination of 50% of all dogs, and semi-annual vaccination of 20% of dogs. We included, for each vaccination option, two dog rabies transmission scenarios: low (1.2 dogs infected per infectious dog) and high (1.7 dogs infected per infectious dog) (see later for further details). We used several published sources of demographic, epidemiological, and economic data (Table 1). We used a government perspective (government-as-payer). We assessed the impact of the interventions over a 10-year period, and we discounted all future costs and benefits (including lives saved) at a rates of 3% and 16% [24]. The later discount rate was derived from the weighted average yield to maturity for 10-year Bank of Tanzania Treasury bonds in October 2017. (https://www.bot.go.tz/financialmarkets/aspSmartUpload/TBondsResults.asp: accessed May 10, 2018). A user of RabiesEcon can alter almost all the input values.
Our illustrative East African example includes urban and rural settings, using a population of approximately 1 million, with 2/3 of that population in an urban setting and 1/3 in a rural setting (Table 1). We set the total area occupied by this population at approximately 2,000 sq. km., with approximately 200 sq. km. being urban (Table 1). These urban and rural settings allow for differences in human and dog population densities, and resultant differences in risk of rabies transmission (Table 2). We used, based on published studies, a rate of human to dog population of 18.1:1 for the urban areas and 7.4:1 for the rural areas [2, 20, 30] (Table 1).
Table data removed from full text. Table identifier and caption: 10.1371/journal.pntd.0006490.t002 Values to estimate dog-to-dog rabies transmission in East Africa. a The number of dogs infected per infectious dog is sometimes termed as the basic reproduction number, R0. The biting behavior of rabid dogs during the course of infectious periods in rural Tanzania was highly variable (mean bites per rabid dog = 2.15, standard deviation: 2.37) [8].
We used a previously published model [7] as a basis for our mathematical model of rabies transmission incorporated into RabiesEcon (for equations, see S2 Appendix). We provide in Tables 1 and 2, and S2 Appendix (Table 1), a list of inputs used in the transmission model. The model uses one-week time steps. The introduction of rabies into a previously uninfected dog population initially results in large oscillations in the estimated weekly number of rabid dogs. We therefore, to make it easier to facilitate comparisons between no vaccination and dog vaccination programs, programmed into RabiesEcon a process to calculate a “steady state” of a near-constant number of annual cases of canine rabies in a “no vaccination” scenario. We did this by programming RabiesEcon to run an initial 10,000 weeks (S2 Appendix and Table 1 shows the specific parameters used). Because the risk of dog rabies transmission depends on a number of variables, such as the density of dogs and bites per rabid dog when attacking susceptible dog, we included in our analyses of each vaccination program two scenarios, low and high, of rates dog-to-dog rabies transmission [8,12,30]. We calculated the number of dogs infected per infectious dog as follows: Number of dogs infected per infectious dog (Ro) = Number of bites from infectious dog to susceptible dog x risk of infection per bite from infectious dog. Based on data from Tanzania, we used a range of 2.4–3.8 bites per infectious dog [8]. We then, to provide a range of Ro values from 1.1 to 1.7 (Table 2), assumed a value of 0.45 as the risk of infection per bite from infectious dog (S2 Appendix. Table 1). The range of values of Ro used closely follows the range reported by Hampson et al [8], when they reviewed the literature of canine rabies transmission dynamics. The number of dogs infected by an infectious dog (Ro value) is likely impacted by factors such as dog density and percentage of dogs that are unconfined (free roaming). The relationship between those and Ro is not well measured. Thus, any value chosen or calculated becomes a proxy for the impact of those other factors. We note that deterministic models, of the type used to build RabiesEcon, allow for the number of infectious dogs to be reduced to less than 1 (e.g., 0.5 infectious dog), but still able to transmit. This can result in “pop up” outbreaks of dog rabies in later years. We retained this factor for two reasons; It can be interpreted as mimicking, to a degree, the risk of importation of a rabid animal from outside, or the incomplete recording of all rabid dogs within, the dog rabies control area. And, users of RabiesEcon can easily ignore those “pop-up” outbreaks that occur in years well beyond the chosen analytic horizon (e.g., if the user runs a scenario in which dog rabies is eliminated by year 6, “pop up” of cases in, say, year 16 can be assumed to be due to the mechanics of the model).
As stated earlier, we compared a no vaccination option to two dog vaccination options (annual vaccination of 50% of all dogs, and semi-annual vaccination of 20% of dogs) (Table 3). The 50% annual coverage rate reflects, approximately, the average rate found by Jibat et al when they reviewed dog rabies vaccination coverage in Africa as reported in 16 published papers [31]. The 20% rate for semi-annual vaccination represents a potentially cheaper alternative (i.e., 10% less dogs are vaccinated). However, because the high turnover of dog populations (due to a combination of short life expectancy and high dog birth rate–Table 1), an annual vaccination program may result in up to 1/3 of vaccinated dogs dying in the interim between vaccinations programs. A smaller, but more frequent, semi-annual vaccination program may result in almost the same percentage of vaccinated dogs as with the annual program.
Table data removed from full text. Table identifier and caption: 10.1371/journal.pntd.0006490.t003 Characteristics of the mass dog vaccination and neutering programs, and post-exposure prophylaxis. a.
Frequency of vaccination: number of times the vaccination is given in a year.b. Vaccination coverage: percent vaccinated each time the vaccine is given. Option 1 considers annually vaccination covering 50% of the dog population. In Option 2 considers biannual vaccination covering 20% of the dog population during each vaccination program.c. Assumed that rabies vaccine in dogs is the same level of effectiveness as in humans [4].d. We assumed that the percentage of dogs neutered would be half that observed in 150 dog owning households Machakos, Kenya [29]. See text for further details.
These dog vaccination Options are illustrative, and can be readily changed by a user. We examine, in the sensitivity analysis, the impact of increasing the vaccination rate to the World Health Organization recommended level of 70% [2, 3,12]. We assumed that dog rabies vaccine, when correctly administered, was 95% effective, similar to the effectiveness in humans [4]. Following Zinsstag et al. [7], we included waning immunity in dogs vaccinated against rabies (Table 3). Because dog birth rate greatly influences dog-to-dog rabies transmission [7, 29], we included in the dog vaccination options concurrent dog population control programs, in which annually 7.5% of the intact male dogs were neutered (Table 3). We assumed that, for a user-defined percentage of male dogs neutered, there will be an equal percentage reduction in the number of dog litters, and thus a reduced dog population. We based this percentage on half the percentage of castrated male dogs observed in a survey of 150 dog-owning households in Machakos, Kenya [29]. We halved the percentage observed in Machakos because that was a relatively small survey, and our experience is that dog neutering programs in Africa are frequently under-resourced and thus do not impact large portions of the dog populations. We altered this assumption in our sensitivity analysis (see later). We assumed, based on recent data from Haiti (which faces rabies control resource constraints similar to many countries in Africa), that dogs with rabies symptoms would be immediately euthanized, and a small percentage (0.7%) of the brains from those animals would be laboratory tested for rabies (Table 3). We further assumed that 5% of all dog-human bites would be investigated for potential rabies transmission [19]. Finally, we assumed that 21% of dog bite victims would start post-exposure prophylaxis (PEP) (see later, Table 4). We assumed a 95% efficacy when PEP is given as per recommended protocols, [4]. We altered in our sensitivity analyses the percentage of dog bite victims who receive PEP (see later).
Table data removed from full text. Table identifier and caption: 10.1371/journal.pntd.0006490.t004 Human and animal costs related to treating suspected rabies exposures and dog population managementa. a.
We used a 3% discount rate [24]. All costs adjusted to 2015 US dollars [32].b. Percent of exposed humans who receive PEP and are fully compliant with PEP treatment regime such that they are protected against developing rabies. The percentage receiving PEP is regardless of dog vaccination option considered. The 21% estimate comes from a recent study in Haiti [19], where out of the 54% of bite victims who sought medical care, only 39% began PEP.c. Vaccine efficacy estimated at approximately 95%, if guidelines for dose schedule are followed [4].d. Costs per patient receiving PEP (Table 3). Cost of PEP includes costs of materials (needles, swabs, etc. ), tissue-culture vaccine, RIG (7% of patients receiving PEP receive RIG), and costs of 5 visits to a public health facility.e. Weighted average cost calculated as follows: (probability of bite investigation x $ of bite investigation) + (probability of laboratory testing x $ of laboratory testing). Probabilities from Table 3.f. The material costs of $2.22 per castrated dog [38]. We added, for each castrated dog, $0.65 for human resources, $0.24 for awareness programs, and $0.29 for transportation [37]. See Table 3 for description of coverage of neutering programs We used, when modeling the dog vaccination strategies, the following three assumptions. Dog rabies is endemic (i.e., near steady state) in the region being analyzed. Second, mass vaccination campaigns last 10 weeks, each year (or 10 weeks twice per year if bi-annual). Third, the dog population can only increase to a maximum of 5% per year, which is near the lower limit measured by Kitala et al. in Machakos District, Kenya [28]. Kitala et al stated that the dog population in Machakos was growing at a rate faster than normally encountered in Africa.
We calculated the cumulative 10-year totals of the number of rabid dogs, human rabies deaths and YLG with and without the rabies vaccination programs. We also estimated the 10-year total cost of each program. To calculate the cost-effectiveness over 10 years of each vaccination option per human death averted, we used the following formula: Costperhumandeathaverted=Costsofdogvaccinationprogram−costsincurredwithnovaccinationprogramNumberofhumandeathswithoutvacciantionprogram−humandeathswithvaccinationprogram For estimates of cost per case averted over more than 1 year (e.g., 10 years), each component of the formula was first summed, then the overall result calculated (e.g., for a 10 year cost of human death averted, the 10 year cost for dog vaccination program was summed separately, then added into the formula). When discounting was applied, each component was individually discounted to year 1. We used a similar formula to calculate the cost per YLG, assuming that the average age of dog-rabies related death is 10 years of age [28], and that life expectancy at age 10 is approximately 53 years [27] (Table 1) (Additional details in S2 Appendix, Note #2).
We included, when estimating the costs of dog vaccination programs, the costs of treating humans suspected of rabies exposure, cost of dog population management, and the costs of mass dog vaccination. As previously stated, we used a government perspective (government-as-payer), and thus we did not include costs borne by the patient, such as co-paid medical bills or time lost from work. In addition to the previously mentioned discounting, we adjusted all costs to 2015 US dollars using US gross domestic product implicit price deflators [32].
Costs associated with suspected rabies exposures: We, assumed, based on data from Haiti, that just 21% of exposed persons receive PEP [19]. There are very few studies reporting the probability that a dog bite victim receives PEP [1]. Hampson et al. estimated the probability of receiving PEP as function of the Human Development Index (HDI) [1]. An exposed person in a country with an HDI of 0.3–0.5 (on a scale of 0 to 1, with 1, with 1 being the ideal) had an approximate probability of receiving PEP of 0.4 to 0.8. However, data from Haiti indicate that only 1/3 of those who receive PEP are fully compliant [19]. We also conducted sensitivity analyses in which we examined the impact of assuming the 99% of all potential dog rabies exposures receive PEP (see later). We estimated an average cost of $83.65 per person receiving PEP due to suspected rabies exposure (Table 4). This cost includes materials (needles, swabs, etc. ), tissue-culture vaccine, and cost per outpatient visit to a public health facility (Table 4). The use of rabies immunoglobulin (RIG) in most countries with high burdens of rabies is negligible due to high relative costs and limited supply [1,33, 34]. We assumed that 7% of patients receiving PEP would receive RIG. This assumption was based on Knobel et al.’s estimate of 1% of PEP patients received RIG usage [2], and data from Haiti that 13% of patients receiving PEP also received RIG [19].
Costs of dog management and laboratory testing: Recommendation for quarantining and testing dogs that have bitten a person vary depending on local rabies prevalence and national recommendations [4, 39]. We estimated, using the probabilities of laboratory testing of dogs suspected of having rabies and bite investigations (Table 3), an average cost of $1.08 per dog for laboratory testing and bite investigations (Table 4).
Cost of dog neutering and spaying: We calculated a cost of $3.40 per neutered male dog (Table 4). We based this cost on the cost of $2.22 for pinhole castration in Uganda [38]. To the Uganda-based cost data, we added $0.65/dog for human resources, $0.24/dog per awareness program, and $0.29/dog per transportation costs. We based these non-medical costs using data from a dog vaccination program in Chad [37]. For comparison, the costs associated with a standard surgical castration of puppies in Uganda were $6.02 [40]. Note, that although we did not incorporate in this example the spaying of female dogs, such an option can be selected in RabiesEcon. The cost of spaying, however, is typically greater than neutering (Table 4).
We used an average cost per dog vaccinated of $2.39 (Table 5). We based this on previous studies of mass dog vaccination programs in East Africa [21,23,37,40] (Table 5, and S2 Appendix, Table 2). Operating costs included training, public awareness and program information (e.g., media such as posters and advertisement), personnel costs (e.g., costs of supervisors, technicians, general staff), transportation (i.e. vehicles, gasoline), and other equipment. Medical supply costs included supplies such as dog rabies vaccines, syringes, needles, animal marking, and vaccination certificates. For comparison, Elser et al reviewed published costs of dog rabies vaccination, and found a range $1.13/ dog vaccinated in Bhutan to $5.41/ dog in Kwa-Zululand, South Africa, with upper limits at approximately $11–$16/ dog for different phases of vaccination programs in southeastern Tanzania [41].
Table data removed from full text. Table identifier and caption: 10.1371/journal.pntd.0006490.t005 Mass dog rabies vaccination program costs and average costs per dog vaccinateda,b. a.
See Table 3 for description of frequency and coverage of vaccination programs.b. Additional details in S2 Appendix, Table 2.c. Mass vaccination options are either once-per-year (Option 1) or twice per year (Option 2). See Table 3 and main text further description.d. N/A = not applicable.
In addition to presenting all our results based on two different scenarios of low and high dog-to-dog rabies transmission (Table 2), we conducted the following sensitivity analyses. First, we examined the impact on estimates of rabid dogs in the high transmission scenario by changing the percentage of dogs neutered during the vaccination programs from 7.5% (Table 3) to either 0% or 20%, assuming use of vaccination Option 1 (50% dogs vaccinated annually). Second, we calculated the number of rabid dogs if 0%, 20%, 50%, and 70% of the dog population were vaccinated annually, over a 30-year period. The 70% level is the World Health Organization (WHO) recommended minimum level of rabies vaccination needed to ensure dog rabies elimination [2, 3,12]. We also considered the value of increasing PEP coverage from the base case of 21% (Table 3) to 99%. Assuming that the effectiveness of PEP is 95% (Table 4), and that all those exposed comply with the full PEP regime, such a strategy would be designed to prevent almost all loss of human life to dog rabies, without the cost of large-scale dog rabies vaccination programs. Because such a strategy would have to continue without cessation due to the unceasing threat of rabid dogs, we calculated the results for both 10 years (as for the other analyses in this paper), and for 30 years. Finally, we noted that the rate of onward dog-to-dog transmission is a crucial factor in estimating the spread of dog rabies and the consequent benefits of vaccinating dogs against rabies. We therefore conducted a multivariable analysis in which we made simulations changes in the following 4 variables that most directly impact the number of rabid dogs in our scenarios (Table 1). Annual percentage dogs vaccinated (30%, 40%, 50%—baseline 50%); Dog birth rate (550 and 350/1,000 dogs–baseline 676/1,000); Dog life expectancy (3.0 and 2.5 years–baseline 3.0 years); and, initial rate of dog-to-dog transmission, Ro (1.2, 1.5, 1.8 –baseline 1.2). To simplify, when running this sensitivity analysis, we only used the values for the “urban” setting (Table 1) (i.e., “turned off” rural settings). The range of annual percentage of dogs vaccinated was based on observations that these are the levels of coverage need to begin to observe “notable” reductions, but not guaranteed elimination, of human rabies deaths [1]. The estimate birth rate of 550/1,000 dogs was based on the lower 99% confidence interval from N’Djamena, Chad [7]. The lower estimate of 350/1,000 dogs came from birth rates for young dogs (≤ 12 months of age) in rural Machakos District, Kenya [29]. The lower estimate of life expectancy is based on data from N’Djamena, Chad [7]. The Ro values examined are similar to those in Table 2, which we derived from the review by Hampson et al. [8].
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}