Rabies healthcare-seeking behaviors of urban and peri-urban residents: Results from a rabies knowledge, attitudes, and practices survey, Bangladesh, 2018

Rabies is one of the most lethal infectious diseases, with those living in Asia and Africa having the highest risk of dying from rabies. We conducted a knowledge, attitudes and practices survey in urban and peri-urban areas of Bangladesh to describe canine bite rates, rabies knowledge, and healthcare seeking behaviors and barriers to human and dog vaccination. A bite risk assessment score (BRAS) and healthcare-seeking behavior score (HSBS) was calculated for each bite victim. Respondents were given two hypothetical situations to assess potential behaviors after a bite and willingness to pay for rabies vaccine and immunoglobulin. In total, 2,447 households participated in the survey and 85 bite victims were identified. The BRAS identified that 31% of bites posed no risk of rabies transmission. Multivariate analyses showed that living in Chittagong (β = 1.4; 95% CI: 0.1, 2.7) was associated with a higher HSBS. Findings presented here provide useful information regarding bite occurrences, healthcare-seeking behaviors, and a need for strategies to increase rabies awareness.

approach is not without weaknesses. First, previous studies found that low rabies awareness and PEP use in Southern Asia are likely barriers to the elimination of dog-mediated human rabies deaths [10,11]. A study in eight Asian countries, including Bangladesh, China, India, Indonesia, Pakistan, the Philippines, Sri Lanka, and Thailand, found that among patients who reported bite exposures, 34% did not know about rabies before being bitten, 30% did not know where to find rabies prevention centers, 50% sought advice from a medical doctor, and only 22% received RIG (despite being recommended for 43% of patients) [10]. A different knowledge, attitudes and practices (KAP) survey also indicated low rabies knowledge, with 58% of respondents being unaware of the consequences of a dog bite. A survey of residents from Satkhira Sadar, a southwestern sub-district of Bangladesh, reported higher (>73%) awareness of rabies, its causes, and prevention measures was reported in the aforementioned study [11]. However, only 29% of bite victims received rabies vaccination and none received RIG in that survey study. Second, when the demand for the vaccine exceeds the supply of the subsidized vaccine, patients must purchase vaccine and RIG from pharmacies at a cost of $8.25 USD (701.04 BDT) and $12.50 USD (1062.19 BDT) per vial, respectively [9].
Animal rabies control programs are limited in Bangladesh, leading to a high risk for the transmission of rabies from biting dogs and a low capacity for veterinary response to reports of animal rabies exposure events. Accordingly, any animal bite is currently considered to be a potentially suspect rabies exposure. In contrast, Integrated Bite Case Management (IBCM) is an approach that links public health professionals and veterinarians to assess whether or not a dog bite is considered to be a likely rabies exposure [12]. The WHO and other global partners now emphasize dog vaccination and PEP provision conditional on IBCM-based risk assessments in lieu of a model where everyone with a recent bite receives PEP. Such techniques can reduce unnecessary use of PEP, improving cost-effectiveness of rabies control programs [13].
A 2012 study evaluated animal bites and rabies in humans among rural communities in Bangladesh [14]. The study focused only on rural sub-districts and estimated the number of bites and healthcare-seeking behaviors of bite victims using a snowball sampling technique. In this study, we explored rabies knowledge, attitudes towards dog bites, and practices of bite victims in urban and peri-urban settings in Bangladesh. The objective of this study was to: 1) estimate canine bite rates in urban and peri-urban settings in Bangladesh; 2) assess knowledge of rabies risks and protective measures; 3) evaluate health care seeking behaviors and barriers to human and dog vaccination. Ultimately, the results of the survey will deepen understanding of ways to develop cost-effective, risk-based PEP delivery to persons with probable rabies exposures to prevent unnecessary PEP shortages among urban and peri-urban populations.

Ethics statement
At each household, a voluntary verbal informed consent was obtained from each interviewee as well as the GPS coordinate of the household to validate study area. No names or other directly identifiable information were collected as a part of this survey. This study protocol was developed in collaboration with Bangladesh Directorate General of Health Services and was approved by the Centers for Disease Control Human Research Protection Office under Protocol ID 060118JB.

Survey and study population
A cross-sectional household survey was conducted in four districts of Bangladesh, two of which represented urban areas and two peri-urban areas, using a stratified, multi-stage cluster design with random sampling. Survey sites were selected by Ministry of Health officials to coincide with city corporations and peri-urban upazillas (sub-districts) randomly where canine vaccination campaigns were being conducted. A structured questionnaire with Bengali translation was modified from questionnaires previously conducted in other countries to collect data through a face-to-face interview with participants in the selected sites [15,16]. Knowledge, attitudes, and practices (KAP) survey questions collected information pertaining to (i) demographics, (ii) dog bite rates, (iii) attitudes about dog bites and dog ownership, (iv) willingness to pay for human rabies vaccination, (v) healthcare access knowledge, (vi) rabies disease and vaccination knowledge, (vii) dog ownership practices, (viii) biting dog health status, and (ix) campaign awareness and barriers to rabies vaccinations for dogs (S1 Text). Interviewers used cell phones with the digital data collection tool KoBoCollect installed to record responses (Harvard Humanitarian Initiative, Cambridge, Massachusetts, USA). Interviewers selected by ministry officials were community health workers fluent in English and Bengali with prior experience conducting interviews. Prior to the survey, interviewers participated in a 2-day training held by the US Centers for Disease Control (US-CDC), Bangladesh's Department of Livestock Services (DLS), and Directorate General of Health Services (DGHS).
Two divisions, Dhaka and Chittagong, were selected for survey administration. From these divisions, the districts Chittagong (Chittagong Division) and Narayanganj (Dhaka Division) were selected to represent urban areas, and Sreepur (Dhaka Division) and Meghna (Chittagong Division) were selected to represent peri-urban areas (Fig 1). The first household was selected as the central point of the respective area (e.g. zero-point, cross street, or largest structure, etc.) and the direction was determined by the spinning of a pen. In urban areas every 40 th household was interviewed and in the peri-urban area every 16 th household was interviewed. If the interviewer reached an apartment building, each apartment in the building was counted as a household unit (counting began in the apartments from the ground floor). If the selected household was empty or when no one was home, the next closest household was selected for interview. Reaching any intersection, spinning of a pen indicated the next direction. If the interviewer came at the end of the assigned zone, the spinning pen directed the new direction of the survey after backtracking to the central point. One member of the household, who was either the head of household or met the inclusion criteria when the head of household was unavailable, was interviewed. Inclusion criteria for survey participation were: 18 years or older, a member of the selected household, and willing and able to provide informed consent. Information regarding dog bite occurrences and following healthcare-seeking behaviors was reported retrospectively by the survey respondent for all household members who had known dog bites within the 12 months prior to the survey date, even if the household member was no longer alive. Interviewees also provided responses regarding their knowledge of rabies and willingness to pay for rabies vaccination. The demographics of surveyed household respondents were compared by sub-district (Table 1). Our sample size was calculated for each field site based on an expected bite rate of 5% using the cohort methodology described by Fleiss et al with continuity correction, resulting in a sample size of approximately 660 surveys per field site, or 2,640 for the full study [17].
An annual dog bite rate per 100 residents was calculated for each sub-district. Characteristics of the biting dog were also assessed; dog ownership status was compared to the age of the bite victim and dog survival.

Household wealth and rabies knowledge
We calculate a composite wealth score and rabies knowledge score for each household based on responses to select survey questions. Following previous work, we calculated the composite wealth scores based on scoring of four variables: (i) owned household items, (ii) type of toilet facilities, (iii) source of drinking water, and (iv) household wall construction (Box 1) [18]. Household items were assigned an ordinal point value of one to five, based upon their relative monetary worth. We summed values of household items and then standardized to a scale of zero to one. The factors related to sanitation, type of drinking water, and complexity of toilet construction were evaluated based on their reflection of economic status on a scale of 0 (less advanced toilet construction / drinking water from pond, river, lake) to 1 (advanced toilet construction / drinking water piped inside). Housing construction materials were assessed by the trained surveyor; households with higher quality construction material, such as brick, were given a score of one. The final wealth score was derived from an equally weighted sum of the four variables, standardized to a scale of 0 to100. For respondents who did not provide a response, a score was imputed from the average of the district. We calculated rabies knowledge scores based upon responses to five rabies knowledge questions among respondents that acknowledged that they had heard of the disease "rabies" (Box 2). Those who had not heard of rabies received a default score of zero and were not asked the five rabies knowledge questions. Correct responses received 20 points, those who indicated that they did not know the answer received 10 points, and incorrect answers received zero points to account for potentially harmful outcomes based on incorrect knowledge. The highest possible score a respondent could receive was 100. Responses that were excluded in the final dataset include those who did not respond to the question about prior knowledge of rabies and those who had heard of rabies but did not provide responses to three or more questions.

Bite risk and health behaviors
We also calculated a composite bite risk assessment score (BRAS) for each reported dog bite victim to approximate the WHO-recommended post-bite rabies risk assessment criteria and associated PEP recommendations [1]. If the dog was alive 10 days following the bite, the bite was automatically considered no risk. If the dog died within the 10 days following the bite, the bite was automatically considered high risk. If the health status of the dog after 10 days in unknown, we used the anatomical bite location, the bite severity, and the familiarity of the biting dog to classify the risk as "low", "moderate" or "high" (Box 3). For anatomical bite location, locations such as hands and feet were considered low risk and locations such as the head considered high risk [1,19]. In cases where the dog bit the victim in more than one anatomical location, the highest risk anatomical location was used in the final calculation. Bite severity is based upon the number of anatomical bite locations. Bite location and bite severity variables were scored one (low risk), two (moderate risk), or three (high risk) while familiarity with the dog was scored zero (no risk-dog owned by the family), one (low risk-known, owned dog), two (moderate risk-known, stray dog), or three (high risk-unknown dog). The final bite risk score for those bitten by a dog with an unknown 10-day health outcome was derived from an equally weighted average of the three BRAS variables adjusted to a scale of 0 to 100 and categorized as low (0-33), medium (34-66), or high (>66). The BRAS in conjunction with the bite

Box 2. Knowledge Score Calculation
Question First preferred answer (points)
Bite (20) Scratch (20) Contact with saliva (20) Decline to answer (10) Blank (10) Observing the animal ( Health-seeking behavior score (HSBS) was calculated to grade post-exposure healthcareseeking decisions for dog bite victims. The HSBS is a composite value of preferred actions recommended by WHO after a bite: washing of the wound, seeking medical care, initiating PEP, completing PEP, and receiving RIG. Bite victims were awarded one point each for indicating each of these behaviors, zero for not performing the behavior, or a half-point for missing responses, resulting in a range of zero to five for this composite variable.
We also presented respondents with two hypothetical situations to assess potential behaviors after a bite. Respondents were asked what post-exposure healthcare-seeking behaviors they would exhibit in hypothetical scenarios where 1) they were bit by a dog they recognized, and 2) they were bit by a dog they did not recognize (S1 Text).

Willingness to pay for RIG and PEP
We assessed the willingness to pay for rabies vaccine and immunoglobulin based on responses to questions regarding a hypothetical dog bite to the leg, following methods described by Birhane et. al [21]. Respondents were first asked how much they were willing to pay for rabies vaccine starting from 2,500 Bangladeshi Taka (BDT), which is equivalent to approximately 22% of a monthly household income (Fig 2) [22]. Based on their response to this starting value, respondents were then asked willingness to pay at increments, or 'bids,' of 500 BDT in the positive or negative direction. If respondents did not provide a final minimum or maximum value within 3 bids, they were asked to specify a value. Subsequent to the vaccine cost bid, they were asked how much in addition to the previously stated amount they were willing to pay for rabies immunoglobulin (RIG), starting from 3,500 BDT and increasing or decreasing in increments of 1,000 BDT. The maximum amount that respondents were willing to pay for vaccine and RIG were combined for the maximum amount willing to pay for PEP. Willingness to pay was compared between respondents' wealth and rabies knowledge. Interviewees who chose not to respond to both questions regarding vaccine and RIG were excluded from this sub-analysis. Respondents' attitudes regarding rabies post-exposure healthcare-seeking behaviors were assessed as well as the influence of recognition of the biting dog on these behaviors. Survey proportions were used to estimate rates of dog bite victims employing individual healthcare-seeking behaviors. We calculated risk ratios, representing the probability of a particular healthcare seeking behavior following a bite from a known dog to its probability following a bite from a dog they do not recognize. We assessed whether the difference in behavior differed significantly by whether or not the dog was known using a chi-squared test, with p-values �0.05 considered statistically significant.
2.6.2 Practices. Factors associated with favorable post-exposure healthcare seeking decisions among dog bite victims (n = 85) were assessed using bivariate and multivariate linear regression. We regressed the bite victims' healthcare seeking behavior scores (HSBS) against multiple independent factors collected through the KAP survey. Independent variables of interest for these analyses included age of the bite victims, highest education level of a household member, cost of travel to the nearest location where rabies PEP could be received and ownership status of the biting dog. To categorize risk the status of the biting dog 10 days after exposure and anatomical location of the bite on the victim were also included. All variables were considered epidemiologically important to adjust for in the multivariable analysis as they were considered to be associated with the other independent variables of interest in the population, causal of the outcome, and not on the causal pathway between the independent variables and the outcome. District was included as an independent variable to account for the clustering of data within the different survey settings. Beta estimates reflecting the mean differences in HSBSs among those with and without the factor of interest, with corresponding 95% confidence intervals, were calculated for each independent variable, and p-values �0.05 were considered to represent a statistically significant association.

Survey and study population
Interviewers approached 2,544 households and completed surveys at 2,447 households representing 13,529 household members, 85 of whom had experienced a dog bite within one year prior to survey administration (Fig 3). Survey proportions were translated into population rates using 2017 population estimates from the CIA World Factbook (Table 1) [23].
Most respondents interviewed were male (70%) and the largest proportion of all respondents were between the ages of 18 and 30 (29%) ( Table 1). A secondary education was the highest educational level of any household member for 60% of surveyed households. While the average wealth score of a household was 77.2 out of 100, wealth scores of peri-urban households were lower than those of urban households (70.05 vs 84.87, p-value <0. 001 respectively). Overall, rabies knowledge scores were low, with the average score among all surveyed households being 33.1 out of 100. In urban areas, the average knowledge score was 35.8 out of 100 while the average knowledge score in peri-urban areas was 30.5 (p-value <0.001). Among respondents, 47% indicated that they were not familiar with rabies disease prior to this survey. Both Chittagong and Meghna, located in the Chittagong division, had the lowest scores for rabies knowledge of the four districts compared with Sreepur and Narayanganj, with means of 28.5 and 23.4 compared to 32.0 and 48.1, respectively (p-value <0.001).

Healthcare-seeking behaviors of bite victims
The survey identified 85 dog bite victims resulting in an overall annual bite rate of 628 (95% CI 505-773) per 100,000 residents (Table 1). We found that the largest proportion of the dog bite victims were less than 14 years old (44%), located in Sreepur (49%), experienced a bite to the arm or hand (89%), and was bitten by a recognized, unowned community dog (61%) ( Table 2). Based on the calculated bite risk assessment scores (BRASs), 31% of bites posed no risk of rabies transmission, 24% were low risk, 34% were moderate risk, and 12% were high risk. The health status 10 days after a bite for unowned dogs was not known in 67% of the reported bites (Table 3).
Overall, 46% of the 85 bite victims washed the wound and 56% initiated PEP. Of those who initiated PEP 36% received the recommended 4 doses and 52% received rabies immunoglobulin (RIG) (Fig 4). Of the 26 bite victims with no risk of rabies transmission, 38% washed the wound and 54% initiated PEP with a 100% completion rate of the PEP schedule. Initiation of PEP did not appear to differ with bite risk. Among those whose dog bite was considered low risk of rabies transmission, 70% washed the wound and 45% initiated PEP with an 80% completion rate of the PEP schedule. Of those in the moderate risk category, 28% washed the wound and 67% initiated PEP with a 69% completion rate of the PEP schedule. Of the bite victims in the high-risk category, 70% washed the wound and 60% initiated vaccination with a PEP schedule completion rate of 100%. Based on the age-adjusted rate of reported PEP initiation in Bangladesh, 584,006 people initiate the series each year. Based on the bite risk assessment (BRAS) among this population, we estimate that 181,042 (31%) annual bite victims are at no rabies risk, yet still receive vaccination.
We identified three bite victims who sought medical care but did not go on to receive PEP. Using the BRAS, we determined the risk of these three bites to be one each of no risk, low risk, and high risk. When asked why these bite victims did not receive PEP responses included; the vaccine was too expensive (no risk bite victim), they did not feel it was necessary (low risk bite victim), and the vaccine was not available (high risk bite victim). All three bite victims were alive at the time of this survey with bites occurring within 1 month, 9 months, and 12 months prior to the survey, respectively.
Another three bite victims were reported to have died or were not able to be located prior to the time of survey administration. Deceased bite victim number one was 10 years of age and was bitten on the arm/hand by an unowned dog. This dog was reported to have been healthy after 10 days (no risk bite), so this death is not likely due to rabies from this dog bite. The second bite victim, presumed deceased as respondent could not confirm they were alive, was 12 years of age and was bitten in the arm/hand by an unowned dog. The health status of the dog after 10 days was unknown, and this bite was classified as being a moderate risk for rabies.  In bivariate analyses, being bitten on the chest/torso (β = 1.98; 95% CI: 0.38, 3.57), living in Chittagong (β = 0.94; 95% CI: 0.17, 1.71), and living 6-20 km from where rabies PEP could be received (β = 0.73; 95% CI: 0.03, 1.43) were positively and significantly associated with a higher HSBS among dog bite victims (Table 4). In multivariate analyses, only living in Chittagong (β = 1.30; 95% CI: 0.03, 2.57) is considered positively and significantly associated with higher HSBS scores (Table 4). While the age of the bite victim had a significant p-value (0.0001), this association is not considered to be significant since the confidence interval crosses zero.

Attitudes towards post-exposure healthcare-seeking behaviors
Attitudes about specific healthcare-seeking behaviors differed significantly based on the respondent's familiarity with the dog (Table 5). In hypothetical scenarios of a dog bite, respondents who were unable to recognize the dog were 1.2 times as likely to seek rabies PEP (pvalue <0.0001), 1.1 times as likely to wash their wound (p-value 0.0019), 1.4 times as likely to call a veterinarian (p-value <0.0001), 1.3 times as likely to isolate the dog for observation (pvalue <0.0001), and 1.6 times as likely to submit the animal for testing (p-value 0.0023) than respondents who recognized the hypothetical biting dog.

Willingness to pay for rabies PEP
The mean amount that respondents were willing to pay for a 4-dose course of rabies vaccine after an exposure is equivalent to approximately $25.65 USD (2,201.34 BDT). The average amount that respondents were willing to pay for RIG is equivalent to approximately $24.66 USD (2,116.38 BDT). The combined average that respondents were willing to pay for both vaccine and RIG was approximately $50.32 USD (4,318.58 BDT). The combined cost of both rabies vaccine and RIG would need to be less than $18 USD (1,544.80 BDT) in order for 70% of the study population to be willing to pay for PEP (Fig 5). Linear regression modelling was used to determine if there was an association between the maximum amount respondents were willing to pay for rabies vaccine and RIG and the knowledge and wealth scores of the respondents. Using t-tests, there was no significant association between the maximum amount respondents were willing to pay and rabies knowledge (p-value = 0.91) nor household wealth score (p-value 0.94).

Discussion
Bangladesh continues to report one of the highest rates of human rabies in all of Asia, despite providing approximately 250,000 post-exposure prophylaxis (PEP) regimens to bite victims annually [9]. In an effort to understand barriers to rabies PEP and causes of the continued high mortality, a KAP survey (n = 2,447) was conducted in four urban and peri-urban communities in Dhaka and Chittagong divisions, representing over 13,000 residents. We found an estimated dog bite rate that ranged from 456.9 bites per 100,000 people to 1096.6 bites per 100,000 people by district, reflecting a comparable bite rate to other Asian counties such as India and Nepal, where bite rates are estimated to be 692.6 per 100,000 people and 335 per 100,000 people, respectively [3]. We found a higher dog bite rate among this urban and peri-urban study population compared to previous publications from Bangladesh [11]. Bite rates reported in our study are 2 to 10 times as high as those published in a 2012 study from only rural Bangladesh communities by Hossein et al [14]. None of the communities interviewed during the 2012 study were visited in this 2018 study. Urban/rural differences in dog ownership and bite rates have been noted in other countries, and the differences noted here reflect similar findings. Furthermore, this study shows higher frequencies of wound washing (46% vs 2.4%) and vaccine initiation (56% vs 28.5%) among bite victims compared to what has been seen in a previous study of rural Bangladesh [11]. Urban and rural settings, therefore, may have different rabies exposure characteristics and country-level strategies for rabies elimination (e.g. dog vaccination campaigns or PEP allocation) should seek to take into consideration these heterogeneities in dog bites and rabies exposures. Given differences across the urban-peri-urban gradient, as well as differences between districts sampled in this study, one set of parameter values for estimating human rabies burden applied to the entire country may not be accurate.

Strategies for bite prevention
Rabies knowledge among respondents was lower than expected, given the relatively high education level and widespread rabies endemicity throughout the country. Key concepts of rabies control that were lacking among respondents included awareness of the risk of rabies and proper wound washing after a bite. These key concepts should be incorporated into further comprehensive rabies prevention programs.
The majority of bite victims were children under the age of 14. Of the 37 bites reported in children 0-14 years, all but two of the children were bitten by dogs that did not belong to their family. The majority (73%) of bites came from known, community dogs, followed by unrecognized dogs (19%) and a dog owned by a neighbor (3%). Additional studies in Bangladesh on causes of childhood bites from unowned dogs are warranted as well as the evaluation of preventive measures to reduce rabies exposures and deaths among children. For instance, immediate preventative actions could include targeted education of children focused on appropriate interactions with dogs, particularly dogs that children do not recognize or are not owned by their families, in order to prevent bites [24].
Over 80% of dog bite victims were bitten by a free-roaming dog. This finding is consistent with other reports from Bangladesh that have found a large free-roaming dog population in relation to a relatively low dog-ownership rate [7,11]. While improved bite-prevention education is clearly warranted in Dhaka and Chittagong, efforts to manage the free-roaming dog population should also be considered. Dog population management is complex and must include consideration for the humane removal of unhealthy animals from the population. However, there are very few examples where sterilization programs have resulted in sustained reductions in the dog population [25][26][27]. Similar to the evolution of the current humane dog population management practices in the United States, Bangladesh should consider engaging with local animal welfare organizations to develop practices and policies that result in humane dog population management while adequately addressing the public health risks inherent in free-roaming dog populations [28].

Post-bite healthcare-seeking behaviors and deaths following bites
Post-bite healthcare-seeking behaviors and PEP completion rates among bite victims were much higher than other reports from Asia, where 50% sought advice from a medical doctor, PEP was completed in 85% of those receiving intradermal rabies vaccination, and 22% received RIG [10,29]. This may in part be due to the large network of PEP centers in Bangladesh, as well as the fact that PEP is provided nearly free-of-cost by the government [9].
We identified a large proportion of the bite victim sub-population (60%) with high-risk bites that did not complete PEP, even though they sought medical care. One bite victim did not initiate PEP despite being bitten in the arm by their family dog who died within ten days of the bite. When asked why this person did not receive PEP, the respondent stated that no PEP was available at the facility. Although this individual was still alive at the time of the survey 2 months after the bite, we were unable to follow up with this induvial as identifying information was not collected. Findings from this KAP survey should inform additional studies to characterize high-risk populations and the barriers they experience in seeking appropriate healthcare and rabies PEP.
This survey identified three individuals who were bitten and died or were thought to have been dead at the time of the survey, and were between the ages of 1 and 12. Both the 12-yearold and the 1-year-old bite victims are consistent with the probable rabies case definition provided by the WHO which is a suspected case plus a reliable history of contact [1]. Assuming that these two deaths were attributed to rabies, the study area had a 1-year rabies death rate of 14.8 per 100,000 persons (95% Confidence Interval 2.5-48.8). This figure is consistent with previous estimates and supports existing data that Bangladesh has one of the highest rates of human rabies deaths in the world [3].

A Risk-based approach to post-exposure prophylaxis
As of 2017, WHO recommends a risk-based approach when deciding to administer rabies PEP to more efficiently utilize PEP stockpiles, particularly when costs or supply of PEP may lead to poor access for those that truly need it [1]. However, Bangladesh currently lacks national guidelines for a risk-based approach, instead recommending provision of PEP to all bite victims that are treated at bite centers. Practical experience has shown, and been supported with data presented here, that nearly everyone who attends a DRPCC clinic is provided intradermal PEP, if it is available. However, the sustainability of this massive distribution of PEP is questionably dependent upon government-supported budgets and prone to regional and national supply shortages as was the case for at least one bite victim in this study. Examples of large-scale implementation of these risk-based (i.e. integrated bite case management) systems are rare for low-and middle-income countries [30]. Findings from this survey support that a risk-based approach could be beneficial to triage bite victims based on risk-level and improved post-bite observation of dogs is likely feasible and could result in substantial PEP cost-savings of at least $1.4 million USD (1.2 billion BDT). This approach could also be helpful in reducing the burden on health care facilities. According to The World Bank, Bangladesh has 0.5 physicians per 1,000 people which is higher than what we have seen in other countries who have successfully implemented a risk based approach [13].
The findings from this study show great promise for the potential to implement risk-based PEP approaches, such as integrated bite case management (IBCM) programs. Other programs that have instituted these risk-based approaches have shown similar capacity to follow-up and assess biting dogs [30,31]. Despite a lack of any formal national guidance for the assessment of biting animals, bite victims are able to identify the health status of the biting dog. A study of IBCM implementation in a low-income country found that the complete cost of case investigation (staff, supplies, quarantine, testing) was $24 USD (2,059.74 BDT) per animal assessed [13]. If the WHO-recommended risk-based approach were implemented, and PEP could be safely avoided for persons assessed to have no risk for rabies, the resulting cost-savings could support as many as 56,568 IBCM animal case investigations each year [13]. Initial investment in high-quality surveillance and response programs can be daunting, but continued administration of high-volumes of PEP to persons with no clear risk of rabies poses an undue burden on the government, particularly when these funds could support additional critical rabies control infrastructure development.
There is some evidence to support that bite victims in Bangladesh can afford to contribute towards the costs for rabies PEP. However, any implementation of patient-costs for PEP should be very carefully evaluated. Policies that could negatively impact PEP access for persons with legitimate rabies exposures should be avoided. Fees based on level of risk and ability to pay should be considered [32]. When patients are assessed to have had a true rabies virus exposure, it is the responsibility of public health authorities to ensure that they have access to appropriate treatment.
As with similar KAP studies, this study has its limitations. Only one person from each surveyed household was interviewed who may or may not be the primary decision maker. As such their attitudes and knowledge of rabies may not be reflective of that of the whole household. This person also provided information regarding all bites that had occurred in the household in the year prior to the survey and therefore may be subject to differential recall bias. For instance, participants may be more likely to recall severe bites or bites from an unknown dog rather than mild bites or bites from their owned dog. Confirmation biases may also be contributing to improved health seeking behaviors.

Conclusion
Findings presented here provide useful information regarding bite occurrences, healthcareseeking behaviors, and a need for strategies to increase rabies awareness. This study confirmed a high human rabies death rate in Bangladesh despite a high rate of PEP access and completion rates. While PEP is widely offered throughout the country, results presented here show that the country is not immune to errors in distribution. Implementation of a risk-based approach with an educational component would very likely lead to more efficient PEP utilization, and the cost-savings could be diverted to support advancements in rabies surveillance through development of a program such as integrated bite case management [30].