Manual two-way stepwise variable selection was employed for model building

All these findings highlight the need for further investigation to identify high-risk areas for disease spread in the event of a future disease outbreak in California or nationwide. My overall thesis goal entailed evaluating the risk of STEC on DSSF and the risk of disease transmission in the spatial overlap of feral pigs and domestic swine raised outdoors in California. In Chapter 1, we collected fecal samples from cattle, goats, sheep and pigs raised on DSSF to estimate the prevalence of STEC on these unique operations. I used a multilevel logistic regression model to assess the association between risk factors and STEC presence in fecal samples. In Chapter 2, I used a species distribution modeling method Maximum Entropy to predict suitable habitat for feral pigs in California. Then I overlapped this MaxEnt model with the location of over 300 OPO in California to create a risk map that identified areas most at risk for disease transmission between these two growing swine populations. The increasing potential for contact between domestic swine raised outdoors and feral pigs provides an opportunity for the widespread transmission of diseases throughout California, as each pig could serve as a link in the transfer of pathogens between wildlife, livestock and humans. Additionally, the transmission of diseases to domestic pigs raised outside could negatively impact the sustainability of California’s agriculture economy. In Chapter 3, I designed a study to evaluate the prevalence of STEC in six counties at highest risk for contact between feral pigs and OPO, large plastic garden pots based on the risk map built in Chapter 2. I collected fecal samples from both feral pigs and domestic pigs raised outdoors and used a multilevel logistic regression model to assess risk factors associated with the presence of STEC in samples.

The results of these last two chapters fill critical information gaps regarding the epidemiology of STEC harbored in outdoor-raised pigs on DSSF located near suitable feral pig habitat. The increasing number of diversified small-scale farms that raise outdoor-based livestock in the United States , reflects growing consumer demand for sustainably-produced local foods, including animal products such as meat and eggs. California is the top producer of agricultural products in the US and also leads the country in organic food sales, which includes products from DSSF However, there is a lack of science-based information characterizing the risk factors associated with the prevalence of food borne pathogens, such as Shiga toxin-producing Escherichia coli , in livestock raised on DSSF. Diversified farms are most often small-scale and raise a combination of livestock and numerous types of produce or multiple livestock species, with the intent of selling specialized animal products directly to consumers. Some diversified farms integrate livestock and crop production by using their animals to graze crop residues or cover crops before planting a field to fresh produce. Grazing improves soil fertility and provides farm owners with another source of revenue through fiber or meat products. Many consumers perceive produce grown on small-scale farms and/or meat raised on pasture as more “natural” or safer than food grown on large-scale conventional farms or meat animals raised in confinement systems. However, livestock are asymptomatic reservoirs for food borne pathogens and without adequate mitigation strategies, these pathogenic microorganisms may enter into the food supply. Livestock are intermittent shedders of enteric pathogens and shedding may increase under certain conditions, such as during periods of stress , due to certain management practices or seasonally.

Foodborne pathogens survive in the soil for extended periods of time and can spread to humans directly through contact with livestock or indirectly via contaminated food or water.For instance, cattle grazing uphill from a produce field was likely the causative factor for the 2019 E. coli O157:H7 romaine lettuce outbreak. STEC remains one of the top enteric pathogens associated with food borne outbreaks in the US. The top seven STEC O-serogroups that cause the most severe illness in humans have been traced to consumption of produce consumed raw, such as spinach, tomatoes and melons. Fresh produce consumed raw , which has been contaminated by livestock or wildlife feces containing STEC, can become a vehicle for these pathogens to enter the food supply. The aim of this study is to a) describe the unique characteristics of DSSF, b) estimate the prevalence of STEC in livestock raised on DSSF and c) evaluate the association between risk factors and the presence of STEC in livestock raised on DSSF located in California. Farms were visited twice between May 2015 and June 2016, once during each of the following periods: summer/autumn or winter/spring, which reflect California growing seasons and the seasonality of STEC shedding. Livestock species sampled in this study included dairy and beef cattle, dairy and meat goats, pigs and sheep. Sample sizes were calculated using Epitools based on the total number of animals on each farm, with an assumed STEC prevalence of 5% and 10% precision error, stratified by each livestock species. Individual fresh fecal samples were collected from the ground. Samplers wore gloves and placed approximately 50- 100 grams of feces into each sterile whirl-pak bag . Bags were immediately placed into plastic coolers containing ice packs, transported to the laboratory at the end of the sampling day and most samples processed within 24 hrs.

STEC was isolated from fecal samples as described previously with modifications. In brief, 10 grams of fecal material was placed in 90 ml Tryptic Soy Broth and homogenized before and after. Samples were then incubated for 2 hrs at 25°C with 100 rpm agitation, followed by 8 hrs at 42°C with 100 rpm agitation, and held overnight at 6°C, using a Multitron programmable shaking incubator . For detecting E. coli O157:H7, immunomagnetic separation using Dynal anti-E. coli O157 beads was performed on TSB enrichment broths with the automated Dynal Bead Retriever per the manufacturer’s instructions. After incubation and washing, 50 µL of the resuspended beads were plated onto Rainbow agar O157 with novobiocin and tellurite . Fifty µL of the resuspended beads were also plated onto MacConkey II Agar using sorbitol supplemented with potassium tellurite and Cefixime ; plates were streaked for isolation and incubated for 24 hrs at 37°C. Suspect E. coli O157:H7 isolates were confirmed using traditional PCR for the rfbE gene. To detect non-O157 STEC, 1 mL of pre-enrichment broth was incubated in mEHEC selective media for 12 hrs at 42°C followed by plating and incubating on Chromagar STEC . Up to 8 presumptive STEC positive colonies were confirmed for the presence of stx1 and/or stx2 genes by real-time PCR. Confirmed STEC isolates were then characterized for virulence genes using conventional PCR. After PCR testing, one colony from each positive sample was submitted to the Pennsylvania State University E. coli Reference Center to confirm O-serogroups.A 41-question questionnaire, raspberry plant pot consisting of mostly closed-ended questions, was administered to farm owners at the end of the study period. The questionnaire included sections regarding farm demographics, animal health, farm management practices, and water sources . Variables analyzed for model building included risk factors from the farmer questionnaire, sample day factors and variables that were created using known information about each farm, for instance, whether a farm raised multiple types of livestock or if they integrated livestock within produce fields before planting. Variables from the questionnaire included whether farmers allowed different livestock species to share the same barn and if livestock had contact with wild areas . Weather data from the nearest California Irrigation Management Information System weather station within a similar microclimate, provided environmental factors for possible model inclusion . Also, USDA plant hardiness zones, which are based on the average annual minimum winter temperature, were included as a proxy for the many microclimates in California . Only three zones were necessary to categorize our participating farms: zone 7b , 9a and 9b .Descriptive statistics, were calculated for all data. STEC was estimated for the overall study and per livestock species . Generalized linear mixed models were used to assess the association between STEC presence in fecal samples and risk factors. The binary outcome of interest was whether each fecal sample was STEC positive or negative. Univariate analysis assessed the distribution of variables. During bivariate analysis, variables with low variability, small cell sizes , or large standard errors were either modified, collapsed if appropriate, or discarded from model building. Correlation of numeric variables was measured with Spearman’s rank correlation coefficient; those variables that were correlated 0.80 or more were highlighted during the model-building phase to evaluate for multicollinearity issues. To identify possible confounders, each variable was evaluated using a directed acyclic graph and then added to the model to determine whether the variable affected the odds ratios of the other variables by more than 10%. The glmer function was used from the lme4 package in R to build models, with farm added as a random effect to account for possible farm-level clustering effects when analyzing individual samples.

Variance inflation factors identified possible multicollinearity and variables in the model that had a VIF over 5 were assessed for removal. Top models were compared, and a final model chosen based on the lowest Akaike Information Criterion , smallest deviance, relative to the other models. Intraclass correlation . Diagnostic plots from the DHARMa package in R were used to assess the final model and included fitted vs binned residuals, a Q-Q plot and the Kolmogorov-Smirnov test statistic. Odds ratios and 95% confidence intervals were calculated for variables in the final model. All data analysis was performed using R Statistical Software version 1.4.1036 ©.Of the 16 participating farms, two were not included in model building, as their questionnaires were not completed, leaving a total of 502 fecal samples for model building. The mean, median and range of selected numeric variables assessed for inclusion during model building are shown in Table 1.4. Farms in this study ranged from two to 500 acres and had been farming two to 30 years. Stocking rate was calculated by dividing the total number of livestock, excluding poultry, by the total number of farm acres . Selected categorial variables, stratified by whether they were STEC positive or negative are presented in Table 1.5.P -values were reported for chi-square test or Fisher’s Exact test if cell sizes were less than five. For instance, 28.99% of positive samples came from farms that mixed livestock species within a barn, whilst only 15.70% of negative samples came from farms with shared barns . Moreover, 72.46% of positive samples were from farms that allow livestock to have contact with bordering wild areas .This is one of the first studies to describe the unique characteristics of diversified small scale farms in California, while ascertaining significant associations between risk factors and the prevalence of STEC. This study detected STEC on more than half of the enrolled farms and in all the livestock species sampled. Moreover, O-serogroups isolated in this study included ones that cause serve illness in humans, including O157:H7, O26, O103 and O111. Significant risk factors associated with the presence of STEC included the daily maximum temperature, whether multiple livestock species shared a barn, the livestock source of the collected fecal sample, and whether livestock had contact with wild areas.Overall STEC prevalence measured for the 16 farms in this study was 13.62%. Six of the 16 farms had 0% STEC prevalence; however, due to the intermittent shedding of STEC which may be affected by many factors, this result does not necessarily indicate that they are free from STEC. Although STEC prevalence in livestock raised on large farms has been measured frequently in past studies, evaluation of STEC prevalence and associated risk factors estimated on DSSF is less common. However, a study conducted by USDA-APHIS collected fecal samples from dairy cows in 21 states and stratified E. coli O157:H7 prevalence between large dairies and small dairies and reported that small ranches had 29.4% E. coli O157:H7 and large dairies had 53.9% prevalence. Although this USDA-APHIS study indicated that small farms have less E. coli O157:H7 than large farms, the 29.4% prevalence they detected on small dairies is still larger than the 18.18% we identified in dairy cattle. Risk factors for the transmission of food borne pathogens on large farms may be different, especially if they only raise one crop or livestock type, instead of a diversity of species. One of our studies published in 2018 measured a 4.17% STEC prevalence in sheep raised on a mixed crop-livestock organic farm in California, which was lower than the 13.4% prevalence in sheep identified in this current study.