Study area and period
This study was conducted in Jimma Town from October 2016 to October 2017. The town is located in the Jimma zone of Oromia Regional State, South Western Ethiopia (Fig. 1). Jimma town is situated at a distance of 356 Km, South West of Addis Ababa, the capital city of Ethiopia, between 7º41"N latitude and 36º50"E longitudes and has an altitude of 1704 meters above sea level. The climate of the area is a tropical humid climate characterized by heavy rainfall which ranges from 1200–2000 mm per annum. With the mean annual minimum and maximum temperature ranging from 6ºC and 31ºC respectively, the overall average temperature is approximately 18.5ºC. Jimma zone is one of the largest in livestock populations in Ethiopia with cattle population estimated 2,212,962 heads (CSA, 2016). Dairy cattle are more under production in Jimma town and the surroundings small towns but more than 95% of the cattle populations are under extensive management which are used for mixed dairy and meat production as well as cash income generation for the rural communities.
Target and study population
The target population was apparently healthy crossbred dairy cattle kept under intensive and semi-intensive management systems and local breed cattle which are kept under extensive management system. These involved smallholder dairy farms and Jimma Dairy Development Enterprise (JDDE) and the local breed of male cattle presented to slaughter house aged between 3 and less than 10 years.
Sample size determination
The sample size to arrive at the study population was determined using the formula described by (Dohoo et al., 2009). The conservative estimate of 50% prevalence, 95% level of confidence and 5% absolute precision was used. Accordingly, the estimated sample size of 384 animals was obtained. The calculated sample size was oversampled by 10% to account for possible problems with non-response or missing data (Naing et al., 2006). This allowance was added summing up to the total of 422 samples. These samples were approximately halved to be distributed to dairy farms and slaughter house for blood sample collection. The proportion of required number of samples from each dairy farm was obtained by multiplying 28.3% expected prevalence of C. burnetii in cattle reported from Kenya (Knobel et al., 2013) to the total number of cattle in each dairy farm. Then, 9 animals were sampled from each dairy farm on average.
Study design and sampling strategy
Two cross sectional studies were designed to achieve the objectives of this study. First, a slaughterhouse survey was designed in the following way: on each day of visit to the slaughter house a representative percentage of 25% of animals were picked by simple random sampling technique from the lairage during ante mortem inspection. The sampling frame was constructed by listing the total number of animals in the lairage of each visiting day. The total numbers of slaughtered animals in Jimma slaughter house ranged from 55–85 per day. On average, 14 samples were sampled per day to attain the total samples required and after sampling, animal level data like age, sex, tick infestation, breed, body condition score, production system were recorded.
Second, a farm-level survey was designed to measure Q fever exposure in the following way: a list of all 61 dairy farms and their contact details and location (ie. Kebele) was obtained from Jimma town livestock and fisheries resources development office. Thus a total of 25 dairy farms were selected by simple random sampling technique out of the 61 farms on the list to satisfy the total sample required from dairy farms. All targeted farms are business oriented dairy farms with crossbred and/or pure exotic breeds of dairy cattle (Holstein- Friesian). Based on Mulisa (2011) herd size was categorized as small (if the animals number in the herd were 3–10 animals), and large (if the animal number in the herd were 11 and above). A questionnaire to the farm owners was used to collect risk factors data for Q-fever infection, these included individual-level data and farm-level data. For individual-level data animals’ age in year by the means of dentition as described by (Lawrence et al., 2001) and also asking the owners, sex, body condition score (BCS) categorizes as (poor, good and very good) (Roche et al., 2004) and breed. For farm-level data these included multi species mix, multi age mix, tick infestation status of the animals and farms, history of contact with other herd, herd size (continuous scale), production system (intensive and semi-intensive, extensive), presence of nuisance animals in the farm (dogs, cats, rodents and others), parity, and abortion status were included in the questionnaire/check list (Appendix 1).
Specimen collection procedure
About 10 ml of blood sample was collected from the jugular vein of each selected cattle using plain vacutainer tubes and multipurpose disposable blood collection needle 21Gx1 1/2" plus needle holder (Zhejiang Kanshi) Medical Devices Co. Ltd. (HENSO). Before and after sample collection, 70% ethanol alcohol was applied as disinfectant. Each specimen was labeled with unique identification number. The tubes were transported to Jimma University College of Agriculture and Veterinary Medicine laboratory in an icebox and the tubes were put in an oblique position of 45˚, for overnight at room temperature, to allow clotting of blood, the next morning sera was gently pipetted into cryovials and stored in deep freezer at -200C, until diagnosis was made in the laboratory of National Veterinary Institute (NVI) at Debre-Zeit, Ethiopia.
Laboratory Analysis and Interpretation
All serum samples were tested using Indirect Enzyme-Linked Immunosorbent Assay (i-ELISA) from ID Screen®Q-Fever Indirect Multi- Species kits (ID.vet, 310; rue Louis Pasteur–Grabels–France) for the detection of antibodies against C. burnetii. All reagents were prepared and results were interpreted according to the manufacturer’s instructions. Briefly, the optical densities (OD) were read at 450 nm in a micro-plate photometer (Multi Skan Ex, Thermo Electron Corporation, Finland). Negative control (NC), and positive control (PC) were run as duplicates in the micro – plate wells A, B and C, D respectively whereas sera were run as a single spot in the remaining micro plate wells. Interpretation of the result for each sample was obtained as the percentage of the ratio between the sample Optical Density (OD) and positive control OD, according to the formula. The negative and positive samples were determined based on the laboratory test thresholds–values for its status (Table 1). The sensitivity (Se) and specificity (SP) of the test was claimed 100% as described by the manufacturer using serum from confirmed infected animals but other authors cited the test sensitivity and specificity for serum as 100% and 95%, respectively, compared to PCR (García-Pérez et al., 2009).
The coloration quantity depends on the presence of antibodies in the specimen; positive sample will remain colored after addition of stop solution, while the light yellow negative sample will be colorless or white (Fig. 2).
Data management and statistical analysis
All data collected during the sero-surveys were entered into MS Office Excel 2010. Data were analyzed separately for cattle sampled in dairy farms and cattle sampled at the slaughterhouse. The overall prevalence was calculated as a total number of positive samples for C. burnetii divided by the total number of samples tested multiplied by 100. For each prevalence, binomial ‘exact’ 95% confidence interval (CI) was calculated using Epitools (Sergeant, 2019). To statistically test the difference between the overall prevalence in dairy farms and slaughter house, a test for two sample proportions was calculated using the proportion test calculator in the statistical software STATA versio 13 (StataCorp., 2013)
Univariable mixed effect logistic regression analysis was used to select individual explanatory variable that may predict individual C. burnetii sero-positivity. Variables with a p-value ≤ 0.25 at the univariable screening were taken forward to a multivariable mixed effect generalized linear model (farm as random effect) with Bernoulli family with a logit link. A separate multivariable binomial generalized linear model was used to model herd level prevalence data. Slaughterhouse data was analyzed using logit generalized linear model. Furthermore, multicollinearity was also assessed for any correlation between the explanatory variables with Spearman’s rank correlation and between management system and contact with other herds shows there is a correlation (Spearman's rho =-0.6001; P-value ≤ 0.0015). Interaction terms between explanatory variables were entered into the model to investigate the presence of effect modification. Statistical significance in the multivariable model was set at a P–value ≤ 0.05. All statistical analyses were performed in Stata statistical software version 13 (StataCorp., 2013).