The Statistical model of foot and mouth disease outbreak in Chiang Mai

Abstract

Foot and mouth disease (FMD) is considered a highly contagious transboundary disease of cloven-hoofed animals as cattle. Over the previous decade, FMD has become endemic to northern Thailand. FMD outbreaks were reported in Mae On, San Kamphaeng and San Sai districts in Chiang Mai Province from 2015 to 2016. Hence, three analytical epidemiology methods including two data sources capture-recapture (CR) method, modeling FMD risk factors and spatio-temporal analysis, and social network analysis were applied in this study. The objectives of this study were to 1) to deterministically estimate all FMD cases using CR method in three districts, Chiang Mai province, 2) to describe FMD distribution and identify the most relevant risk factors, investigate the spatio-temporal patterns of FMD outbreak and also to predict the probability of FMD occurrence identified local epidemic risk using modeling FMD risk factor and spatiotemporal analysis, and 3) to analyze the connectedness of the FMD attributes using social network analysis. First analytical epidemiology method, two independent FMD outbreak data sources including data from the Department of Livestock Development of Thailand and questionnaire survey data were analyzed using CR method. The estimate numbers of FMD outbreak farms in 2015, 2016 and 2015-2016 were 54 (95% CI = 39, 67), 213 (95% CI = 203, 223) and 264 (95% CI = 250, 277) farms, respectively. The estimated farmlevel prevalence for the 2015-2016 outbreak was 58.79 % (95% CI = 53.79, 63.79). The estimations provide an upper bound for the true number of FMD outbreak farms in the study area, thereby compromising the under-reporting of the disease. For the second analytical epidemiology method, this study was divided into 2 parts. Part 1, a retrospective space-time scan statistic including a space-time permutation (STP), Poisson, and Bernoulli model were conducted to detect areas of a high incidence of FMD. Results have shown that the cluster sizes of the Poisson and Bernoulli model (>20 km) were greater than this of STP model. The cluster periods from 2015 to 2016 of the Poisson model were approximately 12 months, while 1 month were identified by the Bernoulli and STP models, respectively. For part 2, the study was conducted via a faceto- face interview questionnaire survey at 140 FMD outbreak farms and 307 control farms using logistic and autologistic regression models. Univariable and multivariable logistic regression analyses were used to determine the association between potential risk factors and FMD outbreaks. The final logistic regression model identified factors were related to farm biosecurity, FMD vaccination administration and distance from the farms to risk areas including the purchasing of a new cow without following quarantine protocol (odds ratio (OR) = 2.41, 95% CI = 1.45, 4.05), farms located near shared cattle grazing areas in a 10 km radius (OR = 1.83, 95% CI =1.11, 3.02), FMD vaccination administration by non-official livestock personnel (OR = 2.52, 95% CI = 1.39, 4.58), farms located in a 5 km radius of cattle slaughterhouse (OR = 1.83, 95% CI = 0.99, 3.40) and history of FMD outbreaks over the previous 12 months in districts where farms were located (OR = 0.44, 95% CI = 0.22, 0.86). The logistic regression model was modified with the incorporation of an autocovariate variable to incorporate any spatial autocorrelation between geographic units. Five significant risk factor s from the final logistic regression model was modified with the incorporation of an autocovariate variable to incorporate any spatial autocorrelation between geographic units. The significant risk factor from the final autologistic regression model was the purchasing of a new cow without following the quarantine protocol (OR = 2.55, 95% CI = 2.02, 3.07). Based on the prediction results of the autologistic regression model, the probability of FMD occurrence identified a local epidemic risk that located in northern of San Sai and central of San Kamphaeng district. Apart from the third analytical epidemiology method, the cattle trader-cattle farmer (CTCF) network and the dung trader-cattle farmer (DTCF) network were created to represent direct and indirect contact of FMD transmission, respectively. A structured questionnaire was used to identify sequences of farms routinely visited by each trader. Two static weighted directed one-mode networks were constructed, and the network metrics were measured. In CTCF network, a total of 622 participants were interviewed, including 282 dairy farm owners (45.3%), 240 beef farm owners (38.6%), 3 live cattle market owners (0.48%), 55 cattle collectors (8.84%), 31 cattle traders (4.98%), 4 slaughter men (0.64%) and 7 restaurants (1.13%). The result showed 32% of cattle traders had other careers such as cattle collectors, dung traders, dairy cooperative staff members and artificial inseminators (10/31). The regular traders possessed the highest value of inand out-degree centralities (51, 63), betweenness centralities (23,072.5), and k-core values (3). In DTCF network, a total of 611 participants were interviewed, including 407 dairy farmers (45.3%), 154 beef farmers (38.6%), 15 regular traders (8.84%), 10 dealers (1.13%), 14 final purchasers (0.64%) and 11 Trader who owns cattle farm (4.98%). The regular tradesman possessed the highest value of in- and out- degree centralities (71 and 4), betweenness centralities (114.5), and k-core values (2). Two networks showed that less values of clustering coefficient and network density. These mean that the network topology was random network. Interestingly, there was no recording system for the movement of the dung trader. In addition, such traders traveled across the study area without any movement control. The advantage of this study could be used to enhance or develop the effective control measure for FMD which was dense in a movement activity. In conclusion, the application of three analytical epidemiology methods in this study provided a better understanding of the epidemiology of FMD by identifying the true status and risk factors of FMD as well as determining clusters of the disease. The finding from this study can be used for developing effective control programs of FMD in northern Thailand.