With the daily influx of newborn calves, changes in incidence may be observed within months of implementation of the prevention or control program. Dairies in California differ in many ways from dairies in other parts of the United States. For example, the average dairy herd size for California in 2016 was 1,249 milking cows, with herds up to 10,000 milking cows, whereas the latest report on dairy cattle management practices published by USDA categorizes herd sizes with more than 500 cows as large . Other differences include California’s climate, which allows for year-round outdoor housing of calves and cattle, which is impractical in many parts of the country . Climate and dairy management practices in California are not homogeneous across the state, however. As an example, in 2011, California held 22.6% of the nation’s organic dairy cows, most of which were found in the northern part of the state . It is clear, that the development of a risk-assessment tool for California dairies cannot be based on national averages but should be based on data collected from within this unique industry. The objective of our study was to identify factors associated with BRD in preweaned calves on California dairies by correlating management practices on California dairy farms to prevalence of BRD in calves exposed to these practices. Our hypothesis was that BRD is associated with management practices related to risk factors for BRD, such as the amount of colostrum and milk fed, housing type, and vaccinations.
The research described here is a further step in the development of a risk-assessment tool for BRD in preweaned calves on California dairies following the survey of management practices related to BRD in preweaned calves on California dairies .A cross-sectional study was designed to collect data on management practices and estimate BRD prevalence in preweaned dairy calves on a convenience sample of California dairies. In 2013,trim tray pollen managers of 1,523 California dairy herds were surveyed using 2 mail questionnaires followed by a reminder card after each mailing. The survey aimed to identify the variety of management practices known to be associated with the incidence of BRD in preweaned calves with the purpose of characterizing these practices, comparing them across geographic regions of the state, and determining the principal components that explain the variability in management between herds . Respondents to the survey were given the option of providing contact information if they were interested in participating in a follow-up study. In addition to respondents who showed willingness for a follow-up herd visit, we sought local dairies, veterinarians, consultants, farm advisors, and dairy owners through personal contacts to enroll farms. Emphasis was placed on inclusion of a wide range of herd sizes, geographical locations, and management styles into the sample. The study protocol was approved by the University of California at Davis Animal Care and Use Committee .The 11 principal components identified in the 2013 survey formed the basis for the on-farm questionnaire . Briefly, the questionnaire collected information about demographics , maternity pen management , colostrum management , calf management , vaccinations , and dust abatement the visit in a simple random sample of calves at the dairy that were either housed individually or housed in groups provided they were fed milk as part or all of their diet. Calves were scored using the standardized and validated California scoring system for BRD in preweaned calves based on a point system .
In the California scoring system, points are assigned for the presence of each of these clinical signs: 2 points for cough, ocular discharge, dyspnea, or a rectal temperature of ≥39.2°C ; 4 points for nasal discharge; and 5 points for a head tilt or ear droop. A calf that scores 5 points or more is scored as positive for BRD.Dairies were visited between May 2014 and April 2016 and in-person interviews with owners or herd managers were conducted. The 10 individuals conducting the interviews consisted of veterinarians, graduate students, research assistants, animal scientists, extension and outreach specialists, and farm advisors who were all familiar with the California dairy industry. All interviewers were trained on the use of the California BRD scoring system and were accompanied by a more experienced team member during their first farm visit. Interviewees were instructed to answer questions based on management practices in place only as far back as when the oldest calf eligible to be enrolled was born. The number of calves on the premises was determined as part of the questionnaire. Once the questionnaire was completed, study personnel proceeded to score a random number of calves based on the total number of calves on the farm to estimate the calf herd’s BRD prevalence. The random sample was drawn from all individually housed calves and calves in groups on a milk diet and included calves that were recently weaned if still housed in a hutch. Sample sizes of calves to be scored for N = 100 , 250 , 500 , 750 , 1,000 , and 1,500 calves on the farm were calculated per equation 1 and sheets with random numbers between 1 and N prepared in a spreadsheet . The sheet with the next larger N was used for selecting the sample of calves. Scoring started typically with the youngest calf followed by counting of hutches according to the random number sheet until the next calf on the list was reached and scored and so on until the oldest eligible calf was reached.
Apart from the clinical signs that are part of the BRD score, the calf’s ear tag number, sex, and date of birth were recorded. Breed was recorded as Holstein, Jersey, or other, the latter including other purebred breeds as well as crossbred animals. Hutch type for each calf was described in terms of floor and wall materials, whether a roof was partial, full, or absent, whether extra shade was provided, and the dimensions of the hutch or pen. In addition, the elevation of the hutch floor from the ground, the presence and type of a flush system for manure,trim bin tray and the number of calves in the enclosure were recorded as well as whether calf-to-calf contact with other calves was possible. Further, the distance from the hutch to the nearest dust area was measured. Maternity pens were visited, dimensions of pens were measured, and hygiene scores for 15 cows were assigned according to a 4-point system. A score of 1 indicated no manure soiling above the pastern; score 2 indicated manure soiling on the lower leg; score 3 indicated manure soiling on the upper leg, udder, and abdomen; and score 4 indicated manure soiling on udder or abdomen and toward spine . If fewer than 15 cows were present in the maternity pen, all cows present were scored. The dimensions of hutches as well as maternity pens and group pens were measured in feet and inches and transformed to metric measurements for analysis.Of 5,034 calves that were originally enrolled, 398 calves on 4 dairies were removed from analysis due to missing information on their date of birth. Several dairies kept records with approximate ages of calves , but these 312 calves were also excluded from the final model once age was confirmed as an important confounder. Calves that had missing data for any of the predictors in the final multivariable model were excluded from analysis as well, resulting in 4,253 calves in the final model . No imputation of missing data was performed. Survey-weighted proportions of calves missing from the analysis were compared with calves in the final model in terms of prevalence with Pearson’s chi-squared test.All statistical analyses were performed using a commercial software package . Calf-level explanatory variables included age, breed, and sex; housing-related variables, such as the type of hutch the calf is housed in, including the wall and floor materials; whether extra shade was available; whether the hutch was elevated or not; whether or not calf-to-calf contact was possible; whether the calf was individually housed or in groups; if and what type of flush system for manure was in place; and what type of surfaces were close to the calves. In addition, milk-feeding related variables were assigned at the calf level by recording each calf’s age and relating it to the source and amount of milk being fed, and whether or not the milk was pasteurized or anything was added to the milk. Vaccine-related variables were at the calf level by determining whether or not a calf had received a certain vaccine based on the age of the calf and the vaccination protocol discussed while filling in the questionnaire.
All other variables, such as all variables relating to maternity pen management, colostrum management, vaccinations of cows and heifers, or herd demographics, were recorded at the herd level. All analyses were performed at the calf level with the unit of analysis being the calf and values of P < 0.05 considered significant. Relational Database Design. Individual calves scored during visits were matched to questionnaire answers by dairy identification in a database . Calf records were linked to survey responses with a query in Access, matching the age, sex, and breed demographics of the calf to the respective management based on the survey. Most differences in management were related to the amount, source, or number of milk feedings by age, but also included different timing of movement to group housing by breed or management differences concerning diet or colostrum by breed or sex. All records were checked for completeness, plausibility, and accuracy. Descriptive Statistics. Characteristics of dairies and calves in the sample were described using proportions for categorical and means for continuous variables as well as standard errors and 95% confidence intervals of these estimates. Estimates at the calf level, including prevalence estimates, were survey adjusted, where each observation was weighted by the inverse of the sampling fraction , with fj being the number of calves scored divided by the number of calves on the jth premises. Comparisons between levels of categorical variables were performed with chi-squared tests and BonferroniDunn adjustments for multiple comparisons. Strata with single sampling units were omitted from proportion estimates because they prevented calculation of a standard error. Multivariable Logistic Regression Model. The dependent variable for all analyses was the presence or absence of BRD in study calves, as assessed via the California BRD scoring system. Continuous variables were assessed for linearity using the Box Tidwell test . Two-by-two tables of categorical variables with case status were inspected for cell frequencies and categories were collapsed as necessary to achieve cell sizes of 5 or greater. Survey-adjusted univariate generalized linear mixed models with a logit link function with a random effect for dairy were fitted to each variable with the melogit command in Stata. Variables were then added one at a time to a survey-adjusted multivariable generalized linear mixed model with a logit link function with dairy as a random effect. Variables were assessed in the multivariable model in the groups colostrum management, housing, precalving management and immunity, and feed. Variables were added in order of increasing P-value obtained in univariate analysis. The variables age, breed, sex, season, herd size, and region were considered possible confounders and model fit was assessed frequently during model building by adding these covariates as confounders. Biologically relevant interaction terms for colostrum, feeding, and housing management with the variables breed, herd size, sex, and age were tested during model building. The variables for herd size, breed, and region were forced into the final model to adjust for their potential confounding effects. Variables that either had a significant association with the outcome in the univariate analysis or that had a significant association in the multivariable model but were eliminated after adjusting for another variable were tested again in the final multivariable model. Akaike information criterion was used to assess model fit with models producing lower values considered better fit.Colostrum source was categorized as 100% individual cow or nursing from dam , 100% pooled from multiple cows, or a mix of those sources including colostrum replacer. Colostrum from individual cows and colostrum nursed from the dam were combined into a single category because the quality of the colostrum from these sources was regarded as equal. For those dairies that heat-treated their colostrum, the time between colostrum harvest and heat-treatment was categorized into either less than or more than 6 h.