Olaf Berke

Exploratory disease mapping: kriging the spatial risk function from regional count data

BackgroundThere is considerable interest in the literature on disease mapping to interpolate estimates of disease occurrence or risk of disease from a regional database onto a continuous surface. In addition to many interpolation techniques available the geostatistical method of kriging has been used but also criticised.ResultsTo circumvent these critics one may use kriging along with already smoothed regional estimates, where smoothing is based on empirical Bayes estimates, also known as shrinkage estimates. The empirical Bayes step has the advantage of shrinking the unstable and often extreme estimates to the global or local mean, and also has a stabilising effect on variance by borrowing strength, as well. Negative interpolates are prevented by choice of the appropriate kriging method. The proposed mapping method is applied to the North Carolina SIDS data example as well as to an example data set from veterinary epidemiology. The SIDS data are modelled without spatial trend. And spatial interpolation is based on ordinary kriging. The second example is included to demonstrate the method when the phenomenon under study exhibits a spatial trend and interpolation is based on universal kriging.ConclusionInterpolation of the regional estimates overcomes the areal bias problem and the resulting isopleth maps are easier to read than choropleth maps. The empirical Bayesian estimate for smoothing is related to internal standardization in epidemiology. Therefore, the proposed concept is easily communicable to map users.

Spatial distributions of male and female strawberry poison frogs and their relation to female reproductive resources

In many species with a resource-based mating system, males defend resources to increase their attractiveness to females. In the strawberry poison frog, Dendrobates pumilio, suitable tadpole-rearing sites appear to be a limited resource for females. Territorial males have been suggested to defend tadpole-rearing sites to increase their access to females. In this study we investigate the spatial association between tadpole-rearing sites and the sexes as well as the spatial association of males and females. If strawberry poison frogs have resource defense polygyny, we expect males and females to be associated with tadpole-rearing sites and that females will deposit their offspring in tadpole-rearing sites inside the territories of their mates. To test this hypothesis, home range and core area sizes were calculated for both sexes and the association patterns were compared in two areas that differed in their abundance of tadpole-rearing sites. Home ranges and core areas of females were much larger than male home ranges. Females showed a clumped distribution in the vicinity of tadpole-rearing sites. Males were not clumped and were less associated with tadpole-rearing sites. Females generally did not use tadpole-rearing sites in the territory of their mates and we therefore conclude that males did not defend tadpole-rearing sites for females. Our data are consistent with the general assumption that female distribution is influenced by resource distribution and that male distribution depends on female distribution. Nevertheless, the distribution of D. pumilio females was also influenced by male spacing patterns. Males probably initially establish their core areas where female density is high and then females move among territories to sample males. Males compete vigorously for places with high female density, the defense of which is likely important for enhancing their mating success. In general, the spacing patterns did not differ between populations but the sex ratio was strongly female biased in the habitat with more tadpole-rearing sites, reflecting the direct reliance of females on these resources.

A farm-level study of risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, 2001 – 2004

BackgroundFollowing increased rates of human campylobacteriosis in the late 1990s, and their apparent association with increased consumption of fresh chicken meat, a longitudinal study was conducted in Iceland to identify the means to decrease the frequency of broiler flock colonization with Campylobacter. Our objective in this study was to identify risk factors for flock colonization acting at the broiler farm level.MethodsBetween May 2001 and September 2004, pooled caecal samples were obtained from 1,425 flocks at slaughter and cultured for Campylobacter. Due to the strong seasonal variation in flock prevalence, analyses were restricted to a subset of 792 flocks raised during the four summer seasons. Flock results were collapsed to the farm level, such that the number of positive flocks and the total number of flocks raised were summed for each farm. Logistic regression models were fitted to the data using automated and manual selection methods. Variables of interest included manure management, water source and treatment, other poultry/livestock on farm, and farm size and management.ResultsThe 792 flocks raised during the summer seasons originated from 83 houses on 33 farms, and of these, 217 (27.4%) tested positive. The median number of flocks per farm was 14, and the median number of positive flocks per farm was three. Three farms did not have any positive flocks. In general, factors associated with an increased risk of Campylobacter were increasing median flock size on the farm (p ≤ 0.001), spreading manure on the farm (p = 0.004 to 0.035), and increasing the number of broiler houses on the farm (p = 0.008 to 0.038). Protective factors included the use of official (municipal) (p = 0.004 to 0.051) or official treated (p = 0.006 to 0.032) water compared to the use of non-official untreated water, storing manure on the farm (p = 0.025 to 0.029), and the presence of other domestic livestock on the farm (p = 0.004 to 0.028).ConclusionLimiting the average flock size, and limiting the number of houses built on new farms, are interventions that require investigation. Water may play a role in the transmission of Campylobacter, therefore the use of official water, and potentially, treating non-official water may reduce the risk of colonization. Manure management practices deserve further attention.

From explanation to prediction: a model for recurrent bovine tuberculosis in Irish cattle herds.

There is a good understanding of factors associated with bovine tuberculosis (BTB) risk in Irish herds. As yet, however, this knowledge has not been incorporated into predictive models with the potential for improved, risk-based surveillance. The goal of the study was to enhance the national herd scoring system for BTB risk, thus leading to improved identification of cattle herds at high risk of recurrent BTB episodes. A retrospective cohort study was conducted to develop a statistical model predictive of recurrent bovine tuberculosis episodes in cattle herds in the Republic of Ireland. Herd-level disease history data for the previous 12 years, the previous 3 years, the previous episode, and the current-episode were used in survival analyses to determine the aspects of disease history that were predictive of a recurrent breakdown within 3 years of a cleared BTB episode. Relative to herds with 0-1 standard reactors in the current BTB episode, hazard ratios increased to 1.3 and 1.6 for herds with 2-5 and >5 standard reactors, respectively. Compared to herds with <30 animals, hazard ratios increased from 1.8 to 2.5 and then to 3.1 for herds with 30-79, 80-173, and >174 animals respectively. Relative to herds with <4 herd-level tests in the previous 3 years, herds with 4-5 and >5 tests had 1.1 and 1.4 times greater hazard of a BTB breakdown. Herds that did not have a BTB episode in the 5 years prior to their 2001 episode were 0.8 times less likely to breakdown in the next 3 years than herds that did. Herds breaking down in the spring or summer were 0.8 times less likely to suffer a recurrent breakdown than herds breaking down in autumn or winter (this was likely due to seasonality in testing regimes). The presence of a confirmed BTB lesion was not predictive of increased risk of recurrent BTB. Despite the availability of detailed disease history, the predictive ability of the model was poor. One explanation for this was that herds suffering a recurrence of BTB on their first test after clearing a BTB episode were different from herds that broke down later in the period at risk. Future research might need to include additional variables to identify which subsets of herd BTB episodes, if any, have identifiable features that are predictive of recurrent breakdowns.

Choropleth mapping of regional count data of Echinococcus multilocularis among red foxes in Lower Saxony, Germany

In this paper, choropleth maps display the geographical distribution of Echinococcus multilocularis in red foxes in Lower Saxony. Areas of high prevalence of the infection in foxes and with high fox population density pose high risk for the human population for alveolar echinococcosis. Spatial statistical methods were used to analyse regional count data obtained from 5365 hunted or found-dead foxes. Spatial smoothing (of raw estimates of regional prevalences based on count data before mapping) was used because raw estimates can give an erratic impression of the spatial pattern of the infection. For smoothing, empirical Bayesian methods are used as an explorative spatial epidemiological tool. The resulting map showed the geographical variation of the estimated prevalences around a median at 9% and indicated the presence of spatial-trend effects. Based on this finding, conditional autoregressive spatial modelling for Freeman-Tukey transformed data was used as an inferential spatial epidemiological tool. There were significant additive linear and quadratic spatial-trend effects with elevated prevalences in the north and south and extreme values (prevalences>38%) for the south of Lower Saxony.

Factors associated with whole carcass condemnation rates in provincially-inspected abattoirs in Ontario 2001-2007: implications for food animal syndromic surveillance

BackgroundOntario provincial abattoirs have the potential to be important sources of syndromic surveillance data for emerging diseases of concern to animal health, public health and food safety. The objectives of this study were to: (1) describe provincially inspected abattoirs processing cattle in Ontario in terms of the number of abattoirs, the number of weeks abattoirs process cattle, geographical distribution, types of whole carcass condemnations reported, and the distance animals are shipped for slaughter; and (2) identify various seasonal, secular, disease and non-disease factors that might bias the results of quantitative methods, such as cluster detection methods, used for food animal syndromic surveillance.ResultsData were collected from the Ontario Ministry of Agriculture, Food and Rural Affairs and the Ontario Cattlemens Association regarding whole carcass condemnation rates for cattle animal classes, abattoir compliance ratings, and the monthly sales-yard price for various cattle classes from 2001-2007. To analyze the association between condemnation rates and potential explanatory variables including abattoir characteristics, season, year and commodity price, as well as animal class, negative binomial regression models were fit using generalized estimating equations (GEE) to account for autocorrelation among observations from the same abattoir. Results of the fitted model found animal class, year, season, price, and audit rating are associated with condemnation rates in Ontario abattoirs. In addition, a subset of data was used to estimate the average distance cattle are shipped to Ontario provincial abattoirs. The median distance from the farm to the abattoir was approximately 82 km, and 75% of cattle were shipped less than 100 km.ConclusionsThe results suggest that secular and seasonal trends, as well as some non-disease factors will need to be corrected for when applying quantitative methods for syndromic surveillance involving these data. This study also demonstrated that animals shipped to Ontario provincial abattoirs come from relatively local farms, which is important when considering the use of spatial surveillance methods for these data.

Emergence of Echinococcus multilocularis among Red Foxes in northern Germany, 1991--2005.

The parasite Echinococcus multilocularis is currently of great concern in Europe because of its spreading behavior, which has public health implications. This study investigates the spatiotemporal distribution of Echinococcus multilocularis in Red Foxes. The infection status of 8459 foxes was sampled from 43 regions in the north German province of Lower Saxony over three investigation periods 1991--1994, 1994--1997 and 2003--2005. Linear empirical Bayesian smoothing within the binomial model was used to produce smoothed choropleth maps. Geostatistical kriging was applied to generate prevalence risk maps. Further geostatistical modeling of the prevalence difference between study periods facilitated spatiotemporal trend investigations. The spatial scan statistic was used for cluster detection analysis. The average prevalence risk for Lower Saxony increased from about 12 to 20% during 1991--2005. Specifically the increases from first to second and to third study periods were estimated by 3.3% (CI 95%: 0.6%-5.9%) and 8.5% (CI 95%: 5.2%-11.8%), respectively. Infections in foxes were clustering and a location stable disease cluster was detected in the south of the province. This study is the first showing evidence for steady emergence of Echinococcus multilocularis in Red Foxes. First cases of human Alveolar Echinococcosis were recorded recently.

Spatial Analysis of Sleeping Sickness, Southeastern Uganda, 1970-2003

Disease will likely spread into central Uganda.

The risk of a positive test for bovine tuberculosis in cattle purchased from herds with and without a recent history of bovine tuberculosis in Ireland.

A retrospective cohort study was conducted to assess if cattle sold from Irish dairy herds within 7 months of herd de-restriction (clearance to trade) from a bovine-tuberculosis (BTB) episode had an excess risk of testing positive for BTB during the following 2 years, and to determine other risk factors associated with this outcome. If possible, a predictive metric for herds at high risk of selling future BTB-positive cattle would be generated. The unexposed cohort included all cattle sold within 7 months of the annual herd test in a random sample of dairy herds that did not test positive for BTB in 2003. The exposed cohort consisted of all cattle sold within 7 months of the date of de-restriction in all dairy herds that cleared a BTB episode in 2003. Only cattle sold from herds that were initially found to test positive for BTB using the single intradermal comparative tuberculin test (SICTT)-and not due to discovery of a BTB-positive animal at slaughter-were included as exposed cattle. To aid in the development of a predictive metric, the exposed cohort was subcategorized based on the number of reactors to the SICTT in the herd of origin during the BTB episode immediately prior to sale. The final exposure categories of 0 (unexposed), 1-7, and >or=8 total reactors were considered the unexposed, mildly exposed, and severely exposed cohorts, respectively. A multivariable logistic regression model was fit to the final BTB status of the animal using a generalized estimating equation method (GEE), assuming an exchangeable correlation structure of animals within herds, and using robust standard errors. Exposure level and the other available herd- and animal-level information were modeled. After controlling for other risk factors including the size of the herd of origin and the sex and age of the animal, the three-level exposure variable significantly improved the model (based on a change in Quasi-Akaike Information Criteria of 2.2) and demonstrated a trend of increasing risk of a future positive BTB test with increasing exposure category. The severely exposed cohort of animals had significantly higher risk of a future positive BTB test than the unexposed cohort (OR=1.78, p=0.030).

Spatial prediction of the intensity of latent effects governing hydrogeological phenomena

Fluctuation of the groundwater level in karstic areas results from different cumulative effects, such as water infiltration from precipitation, human interference on the water resources of a certain area, the Moons tide effect, air pressure fluctuation, etc. It is practically impossible to measure these effects directly, generating the need for estimations by indirect methods, applied to the registered water level in monitoring wells. The conventional tool to determine latent or background effects governing variability or fluctuations is factor analysis. Since ordinary factor analysis has been elaborated for independent observations and the observed karstwater levels at a certain location represent realizations of time series, dynamic factor analysis has to be used instead. The obtained dynamic factors can then be identified as the above mentioned effects. Once the intensity of an effect has been determined at every observation site, its spatial structure and prediction is aimed to be given. To achieve this goal, universal kriging is used and analyzed for the spatial process of estimated loadings of dynamic factors. The present paper concentrates on the effect of infiltration. The obtained map of the loadings of the dynamic factor corresponding to infiltration provides valuable information on the intensity of this effect. Its importance from the environmental protection point of view is immense, because it helps to determine those potentially dangerous areas where infiltrating contaminated water can quickly, and in vast quantity, reach underground water resources. Copyright