Prevalence Of Brucellosis
07 November 2023
1 Sampling
The study adopted a Cluster Sampling approach to identify the target districts within Uganda. A comprehensive selection process resulted in the choice of 12 specific districts for inclusion in the study.
Subsequently, for each of the selected districts, the Cochran method for sample size determination was utilized. The sample size calculation is as follows:
\[n = \frac{Z^2 \cdot P(1 - P)}{E^2} * DEFF\]
Where:
- \(n\) is the sample size for a district.
- \(Z = 1.96\) is the critical value corresponding to the 95% confidence level.
- \(P = 0.097\) (9.7% ~ 10%) is the assumed prevalence rate of Brucellosis (Mwebe et al., 2011) .
- \(E = 0.05\) is the margin of error.
- \(DEFF = 2\) is the assumed design effect from clustering.
\[n = \frac{1.96^2 \cdot 0.097(1 - 0.903)}{0.05^2} * 2\] \[n = 269.1917 \]
\[n = 270 \]
Therefore, in each district, about 270 Bovine were screened; any other species of Caprine and Ovine found on the farm were screened
2 Summary Of the Analysis Methodology
A combination of quantitative and qualitative methods were employed, involving data collection, processing, analysis, and modelling. The following subsections describe the tools and techniques used in each stage of the research.
2.1 Tools Used
A combination of tools to process were used to process, analyze, and model the data, ensuring a comprehensive and rigorous approach. R version 4.3.1 in R studio was the main software used for data preprocessing, statistical analysis, and visualization was utilised, mainly using the packages of tidyverse, highcharter, leaflet, qanteda, xgboost, and caret. These packages enabled a robust analytical process, encompassing data preparation, statistical analysis, and the creation of predictive models.
2.2 Data Processing
The data collected from the survey underwent a rigorous data processing stage to ensure the quality and integrity of the analysis. The data processing involved the following steps:
Data entry and validation: The data from the questionnaires and the laboratory results were entered into a Microsoft Excel spreadsheet and checked for errors and inconsistencies. The data were then imported into R version 4.3.1 in R studio for further processing and analysis.
Data cleaning: The data were cleaned to address missing values, outliers, and duplicates. The duplicates were identified using the unique identifiers of the questionnaire farm ID were removed to avoid double counting.
Data transformation: The data were transformed to make them suitable for analysis and modelling. The categorical variables were encoded using dummy variables. The dependent variable, brucellosis prevalence, was calculated as the proportion of positive samples out of the total samples tested for each species on a given farm.
2.3 Data Analysis
The data analysis involved descriptive, inferential statistics.
Descriptive statistics: Descriptive statistics, such as means, standard deviations, minimum and maximum values for continuous variables and frequency tables for categorical variables, were computed to summarize the data. The descriptive statistics were presented using tables and graphs.
Bivariate analysis: Bivariate analysis investigated the relationships between given variables and brucellosis prevalence with respect to species. This involved correlation analysis of prevalence among species and numeric variables, using the Pearson’s correlation coefficient and scatter heat plots. ANOVA tests were employed to assess the effect of given variables on the prevalence of species, using the F-statistic and the p-value.
Multivariate analysis: Multivariate analysis focused on determining the combined effects of multiple variables on brucellosis prevalence. A machine learning technique, gradient boosting model (XGBoost algorithm), was applied. This technique is a powerful and flexible method that can handle complex and nonlinear relationships among variables, and can handle missing values and outliers.The model was also used to identify the most influential factors contributing to brucellosis prevalence among livestock, using the feature importance score.
Geospatial Analysis
The geospatial analysis involved the use of geographic data to visualize and explore the spatial distribution and patterns of brucellosis prevalence among livestock over districts. The geospatial analysis involved the following steps:
Geographic mapping: Geographic data were obtained from the Uganda Bureau of Statistics (UBOS) and were used to create thematic maps that visually represented the spatial distribution of brucellosis prevalence among livestock in R. The maps also displayed the location and name of each district, and the prevalence of each species in each district.
Overlaying geospatial data: Geographic information was overlaid with prevalence data to explore correlations between land use, district characteristics, and brucellosis prevalence.
3 Overall Animal Prevalence
3.1 By Species
The results above focuses on the overall prevalence percentages to gain a comprehensive understanding of the distribution of this disease, which is crucial for formulating targeted interventions and policy decisions.
Bovine Prevalence: The data indicates that, on average, 4.09% of cattle and a farm are affected by tuberculosis. However, this average is influenced by a significant standard deviation of 12.662%, suggesting substantial variability in prevalence rates across different farms. The minimum reported prevalence is 0%, while the maximum reaches as high as 92.5%. This wide range underscores the importance of more focused and localized interventions. High standard deviation implies that certain areas experience much higher brucellosis prevalence in cattle and others very low close to none.
Caprine Prevalence: Among this species, the average brucellosis prevalence on a farm is 3.29%. While this is relatively lower than in Bovine, the standard deviation of 9.709 demonstrates a considerable variation in prevalence rates across the farms, The lowest reported prevalence in goats is 0%, while the highest reaches 45.45%.
Ovine Prevalence: Ovine exhibit the lowest average brucellosis prevalence at 0.35%. The standard deviation of 1.609 remains relatively low, suggesting a more consistent prevalence rate across the study areas. The data shows a minimum prevalence of 0% and a maximum of 7.69% in ovine populations.
In conclusion, this analysis of brucellosis prevalence in the livestock species reveals the significant variations in the disease’s occurrence across the study areas and species. The wide range in prevalence rates, particularly among Bovine and Caprine Species, highlights the importance of targeted interventions to address this disease effectively.
3.2 Correlation of Prevalence Amongst Species
A heatmap map showing correlation of Prevalence Of Brucellosis Amongst Species
The analysis reveals a weak positive correlation (cor = 0.3381) between Bovine Prevalence and Caprine Prevalence which suggests that as Bovine Prevalence increases, Caprine Prevalence also tends to increase. The p-value of 0.0085 indicates that this correlation is statistically significant at the 5% significance level, implying that the relationship between these two variables is unlikely to have occurred by random chance.
Futhermore, it was found out that there is an extremely weak negative (cor = -0.069) between Bovine Prevalence and Ovine Prevalence which suggests that as Bovine Prevalence increases, Ovine Prevalence tends to reduce; however, the p-value of 0.7198 indicates that this correlation is not statistically significant nd may hve occurred randomy; similar to the relationship of CAprine and Ovine Prevalence (corr = 0.0219, p-value = 0.8947)
3.3 By District
Bovine Prevalence: For the examined prevalence of brucellosis in Bovine populations across various districts. - The results reveal significant variations in brucellosis prevalence among different regions. - Apac district stands out with the highest prevalence at 20.51%, indicating a pressing need for interventions to control the disease in the area. Kasese follows with a prevalence of 13.62%, and Mbarara with 13.27%, highlighting the importance of implementing control measures and surveillance in these districts as well. In contrast, several districts such as Kiruhura, Pader, Soroti, Kayunga, Serere, Masaka, and Lira report low or zero prevalence in bovine populations. - The results suggest the importance of targeting areas with higher prevalence
Caprine Prevalence: - Our analysis reveals varying brucellosis prevalence among Caprine populations across the districts studied. - Kiruhura district reports the highest prevalence among Caprine species, with a concerning 19.09%. Kases also reports a relatively high prevalence of 18.18%.
Ovine Prevalence: The data related to Ovine po.ulations in the districts shows that brucellosis is relatively less common among sheep.
In summary, the district-level analysis of brucellosis prevalence among livestock species, including Bovine, Caprine, and Ovine, offers valuable insights on the degree of variations by district, It underscores the importance of tailoring efforts based on the prevalence rates in specific districts and species.
4 Farm Prevalence
(Note:For Only Farms for which a species was screen). The section demonstrates the percentage of farms with and without at least one confirmed case of Brucellosis within each of these species.
4.1 By Species
As shown in the figure above, Among the farms from which Bovine Species was screened, about 18.72% reported the presence of Brucellosis cases, while 81.28% did not have any cases. In the case of Caprine Species, approximately 14.04% of the farms had Brucellosis cases, and 85.96% were Brucellosis-free. For Ovine Species, only 4.55% reported Brucellosis cases, while the vast majority, 95.45%, did not have any cases fro the farms where Ovine was screened.
4.2 By District
4.2.1 For Bovine
| Bovine Farm Prevalence | Total | ||
|---|---|---|---|
| Brucellosis Not Present | Brucellosis Present | ||
| DISTRICT | |||
| Apac | 8 (44%) | 10 (56%) | 18 (100%) |
| Kasese | 4 (36%) | 7 (64%) | 11 (100%) |
| Kayunga | 7 (78%) | 2 (22%) | 9 (100%) |
| Kiruhura | 9 (75%) | 3 (25%) | 12 (100%) |
| Kisoro | 15 (88%) | 2 (12%) | 17 (100%) |
| Lira | 32 (100%) | 0 (0%) | 32 (100%) |
| Masaka | 18 (100%) | 0 (0%) | 18 (100%) |
| Mbarara | 7 (70%) | 3 (30%) | 10 (100%) |
| Pader | 10 (71%) | 4 (29%) | 14 (100%) |
| Serere | 11 (79%) | 3 (21%) | 14 (100%) |
| Soroti | 12 (92%) | 1 (7.7%) | 13 (100%) |
| Wakiso | 32 (91%) | 3 (8.6%) | 35 (100%) |
| Total | 165 (81%) | 38 (19%) | 203 (100%) |
4.2.2 For Caprine
| Caprine Farm Prevalence | Total | ||
|---|---|---|---|
| Brucellosis Not Present | Brucellosis Present | ||
| DISTRICT | |||
| Apac | 6 (86%) | 1 (14%) | 7 (100%) |
| Kasese | 1 (13%) | 7 (88%) | 8 (100%) |
| Kayunga | 5 (83%) | 1 (17%) | 6 (100%) |
| Kiruhura | 4 (67%) | 2 (33%) | 6 (100%) |
| Kisoro | 17 (100%) | 0 (0%) | 17 (100%) |
| Lira | 27 (100%) | 0 (0%) | 27 (100%) |
| Masaka | 6 (100%) | 0 (0%) | 6 (100%) |
| Mbarara | 2 (67%) | 1 (33%) | 3 (100%) |
| Pader | 8 (100%) | 0 (0%) | 8 (100%) |
| Serere | 9 (100%) | 0 (0%) | 9 (100%) |
| Soroti | 8 (89%) | 1 (11%) | 9 (100%) |
| Wakiso | 5 (63%) | 3 (38%) | 8 (100%) |
| Total | 98 (86%) | 16 (14%) | 114 (100%) |
4.2.3 For Ovine
| Ovine Farm Prevalence | Total | ||
|---|---|---|---|
| Brucellosis Not Present | Brucellosis Present | ||
| DISTRICT | |||
| Apac | 3 (100%) | 0 (0%) | 3 (100%) |
| Kasese | 0 (0%) | 1 (100%) | 1 (100%) |
| Kayunga | 2 (100%) | 0 (0%) | 2 (100%) |
| Kiruhura | 5 (100%) | 0 (0%) | 5 (100%) |
| Kisoro | 0 (NA%) | 0 (NA%) | 0 (NA%) |
| Lira | 15 (100%) | 0 (0%) | 15 (100%) |
| Masaka | 1 (100%) | 0 (0%) | 1 (100%) |
| Mbarara | 0 (NA%) | 0 (NA%) | 0 (NA%) |
| Pader | 1 (100%) | 0 (0%) | 1 (100%) |
| Serere | 5 (100%) | 0 (0%) | 5 (100%) |
| Soroti | 8 (100%) | 0 (0%) | 8 (100%) |
| Wakiso | 2 (67%) | 1 (33%) | 3 (100%) |
| Total | 42 (95%) | 2 (4.5%) | 44 (100%) |
4.2.4 Visual Illustration
5 Estimating True Prevalence
To estimate the true prevalence of brucellosis in Bovine in Uganda, a Bayesian approach was employed. This method integrates prior information and data to derive posterior estimates and credible intervals, making it particularly useful for accommodating the imperfections in the diagnostic test, as well as capturing the variability and uncertainty in the data and the prior.
The Bayesian analysis was performed in two steps :-
- The first step estimated the prior true prevalence of brucellosis with data obtained from a systematic review of the literature. The systematic review searched for relevant studies on brucellosis prevalence in livestock in Uganda, using various databases and keywords. The inclusion criteria were studies published between 2000 and 2022, studies that used a standardized diagnostic test, studies that reported the sample size and the apparent prevalence, and studies that covered different districts in Uganda. The data extracted from the selected studies were study characteristics (authors, year, region, species), diagnostic test characteristics (type, sensitivity, specificity), sample size, and apparent prevalence.
The meta-analysis method, based on a random-effects model, was utilized to estimate this prior true prevalence (Gelman et al., 2014). Meta-analysis method is a statistical technique that combines the results of multiple studies to obtain a pooled estimate. The meta-analysis method we used was based on a random-effects model, which assumes that the true prevalence varies across studies due to heterogeneity. The random-effects model accounts for heterogeneity between studies and within-study variability due to sampling error (Borenstein et al., 2009). The meta-analysis method was implemented using a custom function based on the metafor package in R .
The meta-analysis employed the following formula to estimate the prior true prevalence (θ):
\[θ = \frac{1}{\sum_{i=1}^n \frac{1}{\hat{θ}_i}}\]
Where:
- \(θ\) represents the true prevalence.
- \(\hat{θ}_i\) is the estimated prevalence from each study, indexed from 1 to n.
## The prior true prevalence is 1.0907The second step was for generating the posterior true Prevalence using data from this survey.
The data from our survey was used to build a Bayesian model for estimating the posterior true prevalence of brucellosis. The Bayesian model was based on a binomial likelihood function, which describes how the observed data (the number of positive and negative test results) depend on the true prevalence and the test characteristics (the sensitivity and specificity).
The posterior distribution for the true prevalence was obtained by applying Bayes’ theorem, which is a mathematical formula that updates the prior distribution with the likelihood function to produce a new distribution that reflects the updated knowledge after observing the data (Robert & Casella, 2004).
The posterior distribution was calculated using a numerical method called grid search, which involves creating a grid of possible values for the true prevalence, and then calculating the posterior probability for each value based on the prior, the likelihood, and the data. The grid search method was implemented using a custom function in R.
\[P(\text{data}|\theta) = \binom{n}{k} \theta^k (1-\theta)^{n-k}\]
Where:
- \(n\) is the sample size.
- \(k\) is the number of positive test results.
- \(\theta\) is the true prior prevalence.
The prior distribution for the true prevalence was obtained from the results of the meta-analysis in the first step. The posterior distribution for the true prevalence was calculated using Bayes’ theorem:
\[P(\theta|\text{data}) \propto P(\text{data}|\theta)P(\theta)\]
Where:
- \(P(\theta|\text{data})\) is the posterior distribution for the true prevalence.
- \(P(\text{data}|\theta)\) is the likelihood function.
- \(P(\theta)\) is the prior distribution.
Bayes’ theorem combines the prior information with the likelihood function based on the observed data to yield a new distribution that reflects the updated knowledge after observing the data
The results of the Bayesian analysis were summarized by calculating the posterior mean, and a 95% credible interval for the true prevalence.
Therefore; the posterior distribution a peak means that this is the most likely value for the true prevalence given the data and the prior. The posterior mean, which is the average of all possible values weighted by their probabilities is the actual contained in the 95% of the posterior probability mass confidence interval, showing that there is a 95% chance that the true prevalence lies within this range given the data and the prior.
Using the respecitive Species sample prevalence and sensitivity (90%), and specificity (95%) of the diagnostic test to build their respective Bayesian models for estimating the posterior true prevalence of brucellosis, and then true Prevalence
5.1 Bovine True Prevalence
Based on the Bayesian analysis, we estimate that the true prevalence of brucellosis in Bovine Species in Uganda is 5.2%, with a 95% confidence interval interval of 4% to 6%
5.2 Caprine True Prevalence
Based on the Bayesian analysis with respect to Caprine Species sample data, we estimate that the true prevalence of brucellosis in Caprine Species in Uganda is 3.9%, with a 95% confidence interval interval of 2% to 5%
5.3 Ovine True Prevalence
Based on the Bayesian analysis with respect to Ovine Species sample data, we estimate that the true prevalence of brucellosis in Caprine Species in Uganda is 2.2%, with a 95% confidence interval interval of 1% to 4%
6 Maps
6.1 Maps for Prevelance
Showing The Prevalence Of Brucellosis Over Surveyed Districts
In Bovine
In Caprine
In Ovine
6.2 A Land Use Map
Of Surveyed Districts Showing of The Prevalence Of Brucellosis Relative to Specices
7 Uni Variate Analysis
The average prevalence of brucellosis among bovine animals was 4%, with a standard deviation of 12.67%. This means that the prevalence varied widely across the farm, ranging from 0% to 92.5%. The number of bovine animals screened was 16, with a minimum of 1 and a maximum of 93. The number of bovine animals that tested positive for brucellosis was 1, with a minimum of 0 and a maximum of 37.
The average prevalence of brucellosis among caprine animals was 3%, with a standard deviation of 9.71%. This means that the prevalence was also variable across the farm, ranging from 0% to 45.6%. The number of caprine animals screened was 13, with a minimum of 1 and a maximum of 53. The number of caprine animals that tested positive for brucellosis was 1, with a minimum of 0 and a maximum of 11.
The average prevalence of brucellosis among ovine animals was 0%, with a standard deviation of 1.61%. This means that the prevalence was very low across the farm, ranging from 0% to 7.7%. The number of ovine animals screened was 6, with a minimum of 1 and a maximum of 66. The number of ovine animals that tested positive for brucellosis was also very low, with an average, minimum, and maximum of 1.
The results also show the number of abortions in the last two years among the screened animals. The average number of abortions was 4, with a standard deviation of 5. This means that the number of abortions varied significantly across the farm, ranging from 1 to 35.
The results show that most farmers (48.4%) use paddocking as their management system, which means they keep their animals in fenced areas. The average herd size of shoats was 25, with most farmers (58.9%) having more than 25 shoats. Commonly farmers have have mixed breeds (43.9%), followed by cross breeds (32.5%).
The results also show that most farmers (69%) share their grazing ground with other livestock, which may increase the risk of disease transmission. Most farmers (58.2%) have individual access to water for their animals, which may reduce the risk of contamination. However, most farmers (60.4%) reported having shoats that aborted in the last two years, which may indicate brucellosis infection. The most common way of managing aborted material was doing nothing (40.5%), which may pose a health hazard for animals and humans.
The results also show that most farmers (63.9%) buy new stock from other sources, which may introduce new strains of brucellosis to the farm. The average number of new stock added in the last two years was 5, with most farmers (63.9%) adding 0 to 5 shoats. Most farmers (88.4%) do not feed milk to young ones directly, which may prevent the transmission of brucellosis through milk. However, most farmers (95.7%) have not vaccinated their animals against brucellosis, which may leave them vulnerable to infection.
The results also show that most farmers (89.6%) have other livestock on the farm, such as cattle, pigs, or poultry, which may increase the diversity and complexity of the farm system. The most common breeding system for shoats was using a buck or bull on own farm (54.2%), followed by communal bull or bucks (22.9%). Most farmers (56%) added new shoats in the last year, which may reflect the growth and productivity of the farm.
8 Bi Variate Analysis
The results of this section show the relationship between farm characteristics and prevalence of brucellosis among bovine, caprine, and ovine animals on a given farm. The results are based on an ANOVA test, which compares the mean prevalence of brucellosis across different categories.
The p-values indicate whether there is a statistically significant difference in the prevalence of brucellosis among the different chracteristics within each species. A p-value less than 0.05 means that there is a significant difference, while a p-value greater than or equal to 0.05 means that there is no significant difference.
8.1 Production & Management Practice
It is evident from the data that there is a significant difference in the mean bovine prevalence of brucellosis among different production systems (p-value = 0.028). The highest mean bovine prevalence was observed in farms with zero grazing (11.07%), followed by farms with communal grazing (5.75%). The lowest mean bovine prevalence was observed in farms with tethering (0%). This suggests that production system may influence the risk of brucellosis infection among bovine animals.
The results also show that there is no significant difference in the mean caprine and ovine prevalence of brucellosis among different production systems (p-value = 0.501,0.783) respectively. This suggests that production system may not impact brucellosis infection among caprine and ovine species.
8.2 Herd Size
The results show that there is a significant difference in the mean bovine prevalence of brucellosis among different herd sizes (p-value = 0.0175). The highest mean bovine prevalence was observed in farms with above 25 bovines (7.13%), followed by farms with 16 to 20 bovines (5.39%). The lowest mean bovine prevalence was observed in farms with 1 to 5, 6 to 10, and 21 to 25 bovines (0%). This suggests that larger herd sizes may increase the risk of brucellosis infection among bovine animals.
The results show that there is no significant difference in the mean caprine and Ovine prevalence of brucellosis among different herd sizes (p-value = 0.245, 0.893) respectively.
8.3 Breeds
From th results above, it is shown that there is a significant difference in the mean caprine prevalence of brucellosis among different shoats breeds (p-value = 0.0097**). The highest mean caprine prevalence was observed in farms with cross breed shoats (7.18%), followed by farms with indigenous breed shoats (5.44%). The lowest mean caprine prevalence was observed in farms with mixed breeds shoats (0.09%). This suggests that shoats breeds may influence the susceptibility of caprine animals to brucellosis infection.
Additionally, it was found out that there is no significant difference in the mean Bovine and ovine prevalence of brucellosis among different shoats breeds (p-value = 0.232, 0.318) respectively.
8.4 Water Access
The results show that there is no significant difference in the mean bovine prevalence of brucellosis among different water access categories (p-value = 0.150). The highest mean bovine prevalence was observed in farms with communal water access (6.22%), followed by farms with individual water access (2.83%). This suggests that water access may not affect the risk of brucellosis infection among bovine animals.
Additionally, there is also no significant difference in the mean caprine and ovine prevalence of brucellosis among different water access categories
8.5 Management Of Aborted Material
8.6 Having new shoats introduced to the heard with a year back
8.7 Breeding System
8.8 Vaccinated Animal Against Brucellosis
8.9 Feed Milk to young ones directly
9 Multi Variate Analysis
9.1 Influence of various factors to Bovine Prevelance
The results above show the most influential features towards bovine prevalence of brucellosis . The results are based on an XGBoost model, which is a machine learning technique that uses decision trees to identify the features that best explain the variation in the outcome variable.
The results show that the most important feature is DISTRICT, which has an importance score of 21.26%. This means that the district where the farm is located has a strong impact on the bovine prevalence of brucellosis. The second most important feature is Herd Size, which has an importance score of 14.06%. This means that the number of bovine animals on the farm also affects the bovine prevalence of brucellosis. The third most important feature is Caprine Prevalence, which has an importance score of 12.46%. This means that the prevalence of brucellosis among caprine animals on the farm is related to the prevalence of brucellosis among bovine animals.
The results also show that other features have lower importance scores, ranging from 12.23% to 0.39%. These features include the number of new stock added in the last two years, the management of aborted material, the feeding of milk to young ones directly, the production system, the ovine prevalence, the addition of new shoats in the last year, and some specific districts, such as Mbarara and Pader. These features may have some influence on the bovine prevalence of brucellosis, but not as much as the top three features.
The results suggest that there are various factors and practices that may contribute to the bovine prevalence of brucellosis on the farm. Therefore, further analysis is needed to understand how these factors interact and affect each other and to recommend appropriate interventions to reduce the spread of brucellosis.
10 References
Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari, A., & Rubin, D. B. (2014). Bayesian Data Analysis (3rd ed.). Chapman & Hall/CRC Press.
Joseph, L., Gyorkos, T. W., & Coupal, L. (1995). Bayesian estimation of disease prevalence and the parameters of diagnostic tests in the absence of a gold standard. American Journal of Epidemiology, 141(3), 263-272.
Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G., & The PRISMA Group. (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Medicine, 6(7), e1000097.
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to Meta-Analysis. John Wiley & Sons.
Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package. Journal of Statistical Software, 36(3), 1-48.
Robert, C. P., & Casella, G. (2004). Monte Carlo Statistical Methods (2nd ed.). Springer.
Joseph, L., Gyorkos, T. W., & Coupal, L. (2004). BayesDiagnosticTests: A Perl program to calculate Bayesian marginal posterior distributions via Gibbs sampling for the prevalence of a disease, and sensitivity, specificity, positive and negative predictive values and positive and negative likelihood ratios for each diagnostic test for that disease, in the absence of a gold standard. Retrieved from http://www.medicine.mcgill.ca/epidemiology/Joseph/software.html
Bugeza, J., Skjerve, R. A., Zanella, E., Muma, J., Nilsen, E., Oloya, J., & Djonne, G. (2011). Brucellosis seroprevalence in livestock in Uganda from 1998 to 2008. Tropical Animal Health and Production, 43(7), 1447-1452.
Makita, K., Kaneene, J. B., Kankya, C., Matovu, G., Opuda-Asibo, J., Kabasa, J. D., Erume, J., Mwiine, F. N., & Wampande, W. A. (2011). Brucellosis in humans and livestock in Uganda: Challenges and opportunities for control. In J. Zinsstag, E. Schelling, D. Waltner-Toews, M. Whittaker, M. Tanner (Eds.), One Health: The Theory and Practice of Integrated Health Approaches (pp. 161-172). CABI International.
Mwebe, R., Kankya, C., Skjerve, E., Nasinyama, G. W., & Godfroid, J. (2017). A serological survey of brucellosis in sheep and goats in four districts of Uganda. Tropical Animal Health and Production, 49(2), 371-376.
Nakavuma, J., Kankya, C., Basuta, G. M., Mwiine, F. N., Wampande, W. A., Erume, J., Kabasa, J. D., Opuda-Asibo, J., Makita, K., Kaneene, J. B., Godfroid, J., Skjerve, E., & Nasinyama, G. W. (2016). Seroprevalence of brucellosis among slaughterhouse workers in Kampala, Uganda: A cross-sectional study using a rapid immunochromatographic test (ICT) kit for diagnosis of human brucellosis in Uganda. African Journal of Infectious Diseases (AJID), 10(2), 93-99.
Aruho, E., Mwiine, F. N., Wampande, W. A., Kankya, C., Mwebe, R., Godfroid, J., & Skjerve, E. (2021). A serological survey of brucellosis in wildlife in four major National Parks of Uganda. BMC Veterinary Research, 17(1), 64.