Retrospective Study of Human Rabies Exposure and Associated Risk Factors in North-West Ethiopia

2. Materials and Methods

2.1. Description of the study area

Amhara region is located between 8°45'N and 13°45' N latitude and 35° 46' and 40° 25' E longitude with about 170,000 square kilo-meters of land area in the northwestern part of Ethiopia [19]. The region is known for its diversified topography, climatic conditions, agro ecology and animal population. Regarding the climatic condition of the Amhara region, 31% of the region with an altitude of 1500 m.a.s.l. is categorized as Kola (hot zone) 44% of the region with an altitude of 1500-2500 m.a.s.l is classified as Woyina Dega (warm zone) and 25% with an elevation of 2500-4620 m.a.s.l is classified as Dega (cold zone) [20].

The mean annual temperature of the region is between 15 °C and 21 °C, while in the valleys and marginal areas the temperature exceeds 27 °C. Relatively high temperatures are observed in some valleys and the marginal regions exhibiting arid climates. There are two distinct seasons, a short rainy season lasting three to four months and a long dry season of eight to nine months [21].

The total human population of the region is estimated at around 22,877,365 of which about 88% are rural, and the rest 12% are urban dwellers.[19]. From which the study area has a total population of 5,464,079 humans, 5,474,545 cattle; 2,484,400 sheep; 2,461,942 goats; 44,860 horses, 847,780 donkeys and 6,609 dogs respectively [22, ]. The current study was conducted in four zones of Gondar. From these zones, we selected seven districts (East Dembia and Tach Armachiho from Central Gondar, Genda wuha and Metema from West Gondar, Fogera and Debre Tabor from South Gondar and Debark from North Gondar zones) of Amhara region, northwest Ethiopia (Figure 1).

2.2. Study population

The study populations were total human populations in the Central, West, North, and South Gondar zones of the Amhara region, northwest Ethiopia.

2.3. Sampling methods

Three-stage cluster sampling technique was used to select sampling units. First, the four zones of Gondar were selected purposively because rabies is common in these zones. Then districts or towns that we think are representative of the zones were selected: East Dembia and Tach Armachiho from Central, Debark from North, Metema and Genda-wuha from West and Debre Tabor and Fogera from South Gondar zones. Finally, four kebeles were randomly selected from each district making a total of 28 kebeles [23, 24].

2.4. Design of the questionnaire and data collection

A survey questionnaire was developed using the literature and our own experience. The questionnaire aimed to collect data on the occurrence and/or exposure of rabies, associated risk factors for exposure, and post-exposure treatment in the human population. The checklist in the questionnaire had three categories. 1) about the demography of participants: such as “age, gender, religion, educational level, and family size”; 2) about knowledge and management of dogs related to rabies such as “knowledge of rabies infection, rabies vaccination history, awareness of post-exposure vaccination, and understanding of the rabies virus (disease awareness, transmission to humans, and dog symptoms)” and 3) “about previous dog bite exposure”.

The original questionnaire was prepared in English and then it was later translated into the local language (Amharic). The translated questionnaires were then back-translated into English to check consistency. The English version of the questionnaire is attached as a supplementary file (Supplementary File 1). Face-to-face interviews were administered in the final questionnaire. Data was collected by asking respondents whether they had rabies exposure in one or more of their family members in the last two years.

2.5. Sample size determination

Multistage cluster sampling was used to select study participants. The sample size was determined using the sample size estimation formula for random sampling in an infinite population from Thrusfield [25].

$$.{n=1.96}^2\mathrm{*}{\displaystyle \frac{p\mathrm{*}\left(1-p\right)\mathrm{*}D}{d^2}}$$

Where:

n= number of the household

P =proportion of the population taken as 50% since it is unknown in the study area

d²= desired level of precision (margin of error) taken as 5%

D= design effect

Based on these assumptions, a sample of 385 households was required. However, the actual sampling method used was multistage cluster sampling. Therefore, the sample size was adjusted using the following formula $D=1+\left(b-1\right)ICC$.

Where:

b= the number of household responses at least one from the cluster

ICC= intra correlation coefficient

D =design effect

Design effect (D) accounts for the loss of precision in cluster surveys. It's calculated using the predetermined cluster size (28 kebeles) and the number of households selected per cluster; i. e. 23. In multistage cluster surveys, ICC reflects variability across all sampling stages. We used an ICC value of around 0.03 [26]. Considering these factors, the design effect for this sample size determination was calculated to be 1.67. This resulted in a final estimated sample size of 642 households.

2.6. Data analysis

The data were entered into Microsoft Excel. Completeness of the data was checked, followed by calculations of descriptive statistics such as percentages and rates. Risk factor analysis was conducted using binary logistic regression models. Univariable analysis was performed first, followed by pairwise correlation for variables significantly associated with the univariable analysis. Multivariable logistic regression models were then performed to identify factors associated with rabies exposure. Model building considered statistically significant variables (p < 0.05) in the univariable analysis. Any variables with a p-value ≤ 0.05 were considered significant in the final model. The Hosmer-Lemeshow test was used to assess the model's goodness of fit. This test evaluated the overall fit of the final logistic regression model. The presence of confounding was checked by adding one variable at a time in the model-building process. The presence of interactions was checked for variables that remained in the final model. Findings were presented with odds ratios (ORs), 95% confidence intervals (CIs), and p-values. Stata release 17 (Stata Corp LP, USA) was used for the statistical analysis

3. Results

3.1. Respondents’ awareness and management of rabies

A total of 642 respondents were interviewed in the survey. Majorities (94% to 99%) of respondents are aware rabies exists, is a zoonotic disease, and has one or more transmission methods of rabies. However, only a small proportion of them implement control measures such as vaccinating their dogs (41%) and taking post-exposure prophylaxis (23%). Table 2 summarizes the demography of respondents and their awareness and management of rabies.

3.2. Prevalence of rabies exposure and associated risk factors

The incidence of rabies exposure was 10.47%. Rabies exposure had shown large variability between districts, zones, respondents’ level of education and family size, and between respondents who owned and didn’t own dogs. The level of rabies exposure is illustrated in Table 3.

3.3. Sources of rabies exposure

The result showed that 87.73% of the victims were bitten by dog and 11.52% were bitten by equines (donkeys). In the case of ownership of the source animals, 48.70% were exposed by their animals (dogs and donkeys), while 26.39% were exposed by stray dogs, and the rest 24.91% were exposed by neighbors’ animals. Measures taken by respondents after rabies exposure 42.37 % of 269 victims washed the wound using plain water, 36.02% washed using water and soap, 12.29% applied other materials (antiseptics, benzene, kerosene, etc.) and 9.32% applied holy water (Table 4).

3.4. Risk factors associated with rabies exposure.

Twelve variables were tested for statistical association with rabies exposure. Nine of them were statistically significantly associated (P<0.05) with rabies exposure. Variables; like high educational background, being female and vaccinating their dogs reduce rabies exposure while high family size, presence of free contact of the dog, presence of a dog in the house and low level of education increased rabies exposure. The details of univariable associations of rabies exposure and risk factors are summarized in Table 5.

3.5. Multivariate logistic regression model for risk association of rabies exposure

The stepwise multivariate logistic regression model of rabies exposure removes the variables which were not statistically significant (p > 0.05). The multivariable regression analysis demonstrated that the variables ‘district difference, ‘family size’ and ‘dog ownership’ were strongly associated with rabies exposure in humans. In the final logistic regression model analysis only three were significant (P<0.05) from twelve variable i.e., residence(districts) family size, and dog ownership were all significantly associated with exposure of rabies in human. The logistic regression model exhibited a good overall fit (Hosmer and Lemeshow Chi-square value 43.76, P=0.064) (Table 6).