• View in gallery

    Bovine respiratory disease case count (black bars) and cumulative BRD incidence rate (gray line) over time for 70 weaned commercial beef steers during the first 28 days in a feedlot. For the purpose of this study, BRD was defined as a CIS > 1 and a rectal temperature > 40°C.

  • View in gallery

    Box-and-whisker plots of the number of seconds each calf described in Figure 1 spent in contact with all other calves (ALL;A), calves assumed to be shedding BRD pathogens (SHED; B), and the 1 BVDV-PI (PI; C) and histogram of the number of calves assumed to be shedding BRD pathogens (D) during each day of the study period. Calves with BRD were assumed to shed BRD pathogens for the 2 days before and 5 days after BRD diagnosis. For each box-and-whisker plot, the lower and upper limits of the box represent the first and third quartiles, the horizontal line within the box represents the median, and the whiskers delimit the minimum and maximum values (excluding outliers). See Figure 1 for remainder of key.

  • View in gallery

    Receiver operating characteristic curves for the undersampling (dotted line), oversampling (dashed line), both under- and oversampling (both; dotted and dashed line), and ROSE (solid black line) strategies used to compensate for the imbalanced binary data set during creation of the random forest classification algorithm models used to predict BRD for the calves of Figure 1 on the basis of animal-to-animal and community social patterns. The solid gray line represents the line of identity (x = y), which would be consistent with a sampling strategy that provides no diagnostic discrimination. The ROSE strategy had the largest AUC (0.821), which indicated that strategy maximized the capability of the random forest algorithm to predict BRD, and was selected for all subsequent random forest modeling. AUC = Area under the ROC curve.

  • 1. Griffin D. Economic impact associated with respiratory disease in beef cattle. Vet Clin North Am Food Anim Pract 1997;13:367377.

  • 2. Snowder GD, Van Vleck LD, Cundiff LV, et al. Bovine respiratory disease in feedlot cattle: environmental, genetic, and economic factors. J Anim Sci 2006;84:19992008.

    • Search Google Scholar
    • Export Citation
  • 3. Cernicchiaro N, White BJ, Renter DG, et al. Associations between the distance traveled from sale barns to commercial feedlots in the United States and overall performance, risk of respiratory disease, and cumulative mortality in feeder cattle during 1997 to 2009. J Anim Sci 2012;90:19291939.

    • Search Google Scholar
    • Export Citation
  • 4. Taylor JD, Fulton RW, Lehenbauer TW, et al. The epidemiology of bovine respiratory disease: what is the evidence for predisposing factors? Can Vet J 2010;51:10951102.

    • Search Google Scholar
    • Export Citation
  • 5. Babcock AH, Cernicchiaro N, White BJ, et al. A multivariable assessment quantifying effects of cohort-level factors associated with combined mortality and culling risk in cohorts of US commercial feedlot cattle. Prev Vet Med 2013;108:3846.

    • Search Google Scholar
    • Export Citation
  • 6. Cernicchiaro N, Renter DG, White BJ, et al. Associations between weather conditions during the first 45 days after feedlot arrival and daily respiratory disease risks in autumn-placed feeder cattle in the United States. J Anim Sci 2012;90:13281337.

    • Search Google Scholar
    • Export Citation
  • 7. Hay KE, Barnes TS, Morton JM, et al. Risk factors for bovine respiratory disease in Australian feedlot cattle: use of a causal diagram–informed approach to estimate effects of animal mixing and movements before feedlot entry. Prev Vet Med 2014;117:160169.

    • Search Google Scholar
    • Export Citation
  • 8. Hay KE, Morton JM, Clements AC, et al. Associations between feedlot management practices and bovine respiratory disease in Australian feedlot cattle. Prev Vet Med 2016;128:2332.

    • Search Google Scholar
    • Export Citation
  • 9. Theurer ME, White BJ, Larson RL, et al. Relationship between rectal temperature at first treatment for bovine respiratory disease complex in feedlot calves and the probability of not finishing the production cycle. J Am Vet Med Assoc 2014;245:12791285.

    • Search Google Scholar
    • Export Citation
  • 10. Hay KE, Barnes TS, Morton JM, et al. Associations between exposure to viruses and bovine respiratory disease in Australian feedlot cattle. Prev Vet Med 2016;127:121133.

    • Search Google Scholar
    • Export Citation
  • 11. Martin SW, Bateman KG, Shewen PE, et al. A group level analysis of the associations between antibodies to seven putative pathogens and respiratory disease and weight gain in Ontario feedlot calves. Can J Vet Res 1990;54:337342.

    • Search Google Scholar
    • Export Citation
  • 12. Martin SW, Darlington G, Bateman K, et al. Undifferentiated bovine respiratory disease (shipping fever): is it communicable? Prev Vet Med 1988;6:2735.

    • Search Google Scholar
    • Export Citation
  • 13. Chen S, Sanderson MW, White BJ, et al. Temporal-spatial heterogeneity in animal-environment contact: implications for the exposure and transmission of pathogens. Sci Rep 2013;3:3112.

    • Search Google Scholar
    • Export Citation
  • 14. Chen S, White BJ, Sanderson MW, et al. Highly dynamic animal contact network and implications on disease transmission. Sci Rep 2014;4:4472.

    • Search Google Scholar
    • Export Citation
  • 15. Porto SMC, Arcidiacono C, Giummarra A, et al. Localisation and identification performances of a real-time location system based on ultra wide band technology for monitoring and tracking dairy cow behaviour in a semi-open free-stall barn. Comput Electron Agric 2014;108:221229.

    • Search Google Scholar
    • Export Citation
  • 16. White BJ, Goehl DR, Amrine DE. Comparison of a remote early disease identification (REDI) system to visual observations to identify cattle with bovine respiratory diseases. Int J Appl Res Vet Med 2015;13:2330.

    • Search Google Scholar
    • Export Citation
  • 17. White BJ, Anderson DE, Renter DG, et al. Clinical, behavioral, and pulmonary changes in calves following inoculation with Mycoplasma bovis. Am J Vet Res 2012;73:490497.

    • Search Google Scholar
    • Export Citation
  • 18. Theurer ME, Anderson DE, White BJ, et al. Effect of Mannheimia haemolytica pneumonia on behavior and physiologic responses of calves during high ambient environmental temperatures. J Anim Sci 2013;91:39173929.

    • Search Google Scholar
    • Export Citation
  • 19. Shane DD, White BJ, Larson RL, et al. Probabilities of cattle participating in eating and drinking behavior when located at feeding and watering locations by a real time location system. Comput Electron Agric 2016;127:460466.

    • Search Google Scholar
    • Export Citation
  • 20. Grissett GP, White BJ, Larson RL. Structured literature review of responses of cattle to viral and bacterial pathogens causing bovine respiratory disease complex. J Vet Intern Med 2015;29:770780.

    • Search Google Scholar
    • Export Citation
  • 21. Pons P, Latapy M. Computing communities in large networks using random walks. J Graph Algorithms Appl 2006;10:191218.

  • 22. Evans NA. Tulathromycin: an overview of a new triamilide antibiotic for livestock respiratory disease. Vet Ther 2005;6:8395.

  • 23. National Animal Health Monitoring System. Feedlot 2011. Part IV: health and health management on US feedlots with a capacity of 1,000 or more head. Fort Collins, Colo: USDA APHIS Veterinary Services Center for Epidemiology and Animal Health, 2013. Available at: www.aphis.usda.gov/animal_health/nahms/feedlot/downloads/feedlot2011/Feed11_dr_PartIV.pdf. Accessed Jun 8, 2018.

    • Search Google Scholar
    • Export Citation
  • 24. Pillen JL, Pinedo PJ, Ives SE, et al. Alteration of activity variables relative to clinical diagnosis of bovine respiratory disease in newly received feedlot cattle. Bovine Pract 2016;50:18.

    • Search Google Scholar
    • Export Citation
  • 25. Jackson KS, Carstens GE, Tedeschi LO, et al. Changes in feeding behavior patterns and dry matter intake before clinical symptoms associated with bovine respiratory disease in growing bulls. J Anim Sci 2016;94:16441652.

    • Search Google Scholar
    • Export Citation
  • 26. Loneragan GH, Thomson DU, Montgomery DL, et al. Prevalence, outcome, and health consequences associated with persistent infection with bovine viral diarrhea virus in feedlot cattle. J Am Vet Med Assoc 2005;226:595601.

    • Search Google Scholar
    • Export Citation
  • 27. O'Connor AM, Sorden SD, Apley MD. Association between the existence of calves persistently infected with bovine viral diarrhea virus and commingling on pen morbidity in feedlot cattle. Am J Vet Res 2005;66:21302134.

    • Search Google Scholar
    • Export Citation
  • 28. Grandin T. Assessment of stress during handling and transport. J Anim Sci 1997;75:249257.

  • 29. Francisco CL, Cooke RF, Marques RS, et al. Effects of temperament and acclimation to handling on feedlot performance of Bos taurus feeder cattle originated from a rangeland-based cow-calf system. J Anim Sci 2012;90:50675077.

    • Search Google Scholar
    • Export Citation
  • 30. Dohoo I, Marten W, Stryhn H. Veterinary epidemiologic research. 2nd ed. Charlottetown, PEI, Canada: VER Inc, 2009;374375.

Advertisement

Evaluation of animal-to-animal and community contact structures determined by a real-time location system for correlation with and prediction of new bovine respiratory disease diagnoses in beef cattle during the first 28 days after feedlot entry

View More View Less
  • 1 Department of Diagnostic Medicine and Pathobiology, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 2 Beef Cattle Institute, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 3 Department of Clinical Sciences, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 4 Beef Cattle Institute, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 5 Department of Clinical Sciences, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 6 Beef Cattle Institute, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 7 Beef Cattle Institute, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 8 Department of Diagnostic Medicine and Pathobiology, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.
  • | 9 Department of Clinical Sciences, College of Veterinary Medicine, Kansas State University, Manhattan, KS 66502.

Abstract

OBJECTIVE To determine whether animal-to-animal and community contact patterns were correlated with and predictive for bovine respiratory disease (BRD) in beef steers during the first 28 days after feedlot entry.

ANIMALS 70 weaned beef steers (mean weight, 248.9 kg).

PROCEDURES Calves were instrumented with a real-time location system transmitter tag and commingled in a single pen. The location of each calf was continuously monitored. Contact between calves was defined as ≤ 0.5 m between pen coordinates, and the duration that 2 calves were within 0.5 m of each other was calculated daily. Bovine respiratory disease was defined as respiratory tract signs and a rectal temperature > 40°C. Locational data were input into a community detection program to determine daily calf contact and community profiles. The number of BRD cases within each community was determined. A random forest algorithm was then applied to the data to determine whether contact measures were predictive of BRD.

RESULTS Probability of BRD was positively correlated with the number of seconds a calf spent in contact with calves presumably shedding BRD pathogens and number of calves with BRD within the community on the day being evaluated and the previous 2 days. Diagnostic performance of the random forest algorithm varied, with the positive and negative predictive values generally < 10% and > 90%, respectively.

CONCLUSIONS AND CLINICAL RELEVANCE Results indicated that direct transmission of BRD pathogens likely occurs among feedlot cattle. The relative contribution of animal-to-animal contact to BRD risk remains unknown and warrants further investigation.

Abstract

OBJECTIVE To determine whether animal-to-animal and community contact patterns were correlated with and predictive for bovine respiratory disease (BRD) in beef steers during the first 28 days after feedlot entry.

ANIMALS 70 weaned beef steers (mean weight, 248.9 kg).

PROCEDURES Calves were instrumented with a real-time location system transmitter tag and commingled in a single pen. The location of each calf was continuously monitored. Contact between calves was defined as ≤ 0.5 m between pen coordinates, and the duration that 2 calves were within 0.5 m of each other was calculated daily. Bovine respiratory disease was defined as respiratory tract signs and a rectal temperature > 40°C. Locational data were input into a community detection program to determine daily calf contact and community profiles. The number of BRD cases within each community was determined. A random forest algorithm was then applied to the data to determine whether contact measures were predictive of BRD.

RESULTS Probability of BRD was positively correlated with the number of seconds a calf spent in contact with calves presumably shedding BRD pathogens and number of calves with BRD within the community on the day being evaluated and the previous 2 days. Diagnostic performance of the random forest algorithm varied, with the positive and negative predictive values generally < 10% and > 90%, respectively.

CONCLUSIONS AND CLINICAL RELEVANCE Results indicated that direct transmission of BRD pathogens likely occurs among feedlot cattle. The relative contribution of animal-to-animal contact to BRD risk remains unknown and warrants further investigation.

Bovine respiratory disease represents substantial financial and animal health challenges for cattle producers.1 It is a multifactorial disease complex caused by multiple viral and bacterial pathogens as well as environmental factors such as stress associated with commingling and transport.2–9 The role of social behavior on the BRD outcome for individual cattle has not been explored. The social behavior of cattle could be an important determinant in BRD outcome if BRD is truly a transmissible disease. Results of previous studies10,11 indicate that direct transmission of BRD-associated pathogens between cattle can occur in feedlots, and cattle that are seronegative for antibodies against those pathogens at feedlot arrival are more likely to develop BRD during the feeding period than cattle that are seropositive at feedlot arrival. However, results of another study12 suggest that BRD might not be a communicable disease, and there remains considerable controversy regarding whether BRD is or is not a communicable disease.

Real-time location systems have been used to study cattle contact structure in small groups of calves.13,14 Such systems have been used to indirectly monitor cattle behavior with adequate accuracy and precision.15 They have also been used to determine the amount of time cattle spend at locations of interest (eg, feeders or waterers), describe cattle behavior patterns, and detect behavior changes associated with BRD.16–19 Cattle have heterogeneous contact structures that may play an important role in the transmission of enteric foodborne pathogens,14 but the role of those contact structures in the transmission of BRD-associated pathogens has not been evaluated.

The primary objective of the study reported here was to determine whether an initial BRD diagnosis on a given day was correlated with animal-to-animal and community contact patterns (as determined by an RTLS) during previous days for beef cattle during the first 28 days after feedlot arrival. Secondary objectives included the integration of RTLS data into machine learning algorithms to determine whether calf-contact time was predictive of a future BRD diagnosis and to assess the role that the daily community pattern has on the risk of BRD diagnosis within the community, which might provide an indication of whether BRD is or is not a communicable disease.

Materials and Methods

Cattle and husbandry

All animal procedures were reviewed and approved by the Kansas State University Institutional Animal Care and Use Committee (protocol No. 3732). Seventy weaned commercial beef steers with a mean body weight of 248.9 kg (range, 202.7 to 306.4 kg) were procured for the study from the southeastern United States. The study was conducted and the steers were housed at a private feeding facility managed by the Veterinary and Biomedical Research Center in Manhattan, Kan. The study period was from 11:00 am on May 1, 2016, to 7:00 am on May 29, 2016.

The calves were housed in a single standard drylot pen with approximately 28 m2 of pen space available per calf. Calves were fed a complete mixed ration that consisted of prairie hay, dry distiller's grain, and a commercial blended grower ration for the first 14 days of the study. Cracked corn was added to the complete mixed ration for the last 14 days of the study. Feed was delivered once daily in a standard linear feed bunk that provided approximately 0.6 m of bunk space per calf. Water was available ad libitum from a standard float valve ball tank with enough room for 1 calf to drink at a time. A second oval-shaped plastic water tank with a float valve was added to the pen at approximately 3:12 pm on May 9, 2016, and remained assessable to the calves for the remainder of the study.

All calves were processed approximately 12 hours after arrival at the feeding facility (day 0). Processing included the application of duplicate identification plastic ear tags in both ears and an RTLS taga in the left ear. The identification ear tags were assigned to calves sequentially in the order in which they were processed. Calves were also individually weighed and administered a 5-way modified-live-virus respiratory disease vaccine,b clostridial bacterin,c and pour-on anthelmintic.d An ear notch biopsy specimen was collected from each calf and submitted for detection of BVDV RNA by means of a reverse transcription PCR assay. Specimens with a cycle threshold < 30 were considered positive for BVDV RNA. One calf had a positive RT-PCR assay result and was classified as a BVDV-PI.

Clinical observations

Calves were observed by a veterinarian once daily for clinical signs consistent with BRD. The veterinarian also assigned each calf a CIS (Appendix) on a daily basis. Calves assigned a CIS of 2 or 3 were removed from the pen so they could be examined and treated as deemed appropriate.

BRD definition and treatment protocol

Calves assigned a CIS of 2 or 3 that had a rectal temperature > 40°C when examined were considered to have BRD and treated in accordance with a standard protocol. Calves received tulathromycine (2.5 mg/kg, SC), florfenicolf (40 mg/kg, SC), and oxytetracyclineg (4 mg/kg, SC) following the first, second, and third diagnoses of BRD, respectively. There was a 72-hour moratorium after each treatment during which a calf could not receive any additional treatments for BRD. Calves that were pulled from the pen and met the case definition for BRD > 3 times were observed but did not receive any additional antimicrobial treatments.

RTLS

The location of each calf in the pen was continuously monitored with a commercial RTLSa that was set up, installed, and managed by a private third-party company.h The RTLS consisted of ultra-wide band signal transmitter tags, one of which was placed in an ear of each animal, and signal sensors placed around the perimeter of the pen, which triangulate signals within the system area to determine calf location. Data were relayed to a central server and stored. The XY coordinate locations for each calf were logged for each second of the study period.

Data handling and management

Raw data were handled and processed in a commercial database by the RTLS data management company.h If an RTLS tag failed for any reason, the last time that tag detected movement > 2 m was considered the last valid locational data point, and data collected between that point and the application of a new RTLS tag were considered invalid, removed from the dataset, and coded as missing during analysis. The XY coordinates for each calf were compared with the XY coordinates for all other calves at each second of the study period. A Euclidean distance of ≤ 5 m between the coordinates for any 2 calves was defined as contact between those calves. A Euclidean distance of 0.5 m was selected as the cutoff for defining contact between calves by investigators on the basis of the resolution capabilities of the RTLS and expert opinion. The total number of seconds of contact between individual pairs of calves was aggregated on a daily basis, with a day defined as 12:00:00 am to 11:59:59 pm. Those data were transferred from the data management company and then manipulated with an open-source data analytics program.i

For the purpose of this study, calves with BRD were assumed to shed BRD pathogens for the 2 days before and 5 days after BRD diagnosis. That shedding window was selected on the basis of a review20 of the scientific literature, which describes the onset of viral shedding and clinical signs for BVDV and other BRD-associated pathogens. In cattle experimentally inoculated (challenged) with BVDV, the mean duration between challenge and peak rectal temperature and onset of clinical signs is 7 days, and viral shedding peaks approximately 2 days prior to and resolves approximately 6 days after the onset of peak clinical signs.20 We assumed that BRD-affected calves were likely detected near the time at which clinical signs and rectal temperature peaked. The assumed shedding period for each BRD-affected calf was added to the data set.

The primary objective of the study was to evaluate whether the duration of calf contact with other calves presumed to be shedding BRD pathogens in the days prior to initial diagnosis of BRD was correlated with or predictive of that diagnosis. We hypothesized that calves would be exposed to causative pathogens prior to BRD diagnosis, and that the duration of contact with BRD-affected calves would be positively correlated with the risk of BRD diagnosis.20 For each calf, the number of seconds it spent in contact with all other calves (ALL), calves assumed to be shedding BRD pathogens (SHED), and the BVDV-PI (PI) was aggregated on a daily basis. If a calf was in contact with > 1 calf at the same time, the number of seconds it was in contact with each of the other calves was summed together to calculate the calf-contact time at that moment. For example, if a calf was in contact with 2 other calves at the same time (second), the calf-contact time recorded would be 2 seconds. A calf-contact time of 0 seconds on any given day was considered biologically impossible, and that value was treated as a missing independent data point during analyses. The daily BRD status of each calf was added to the final data set as a binary (yes or no) outcome variable. Thus, the final data set consisted of a single record (row) for each calf for each study day.

Detection of social patterns

A community detection program algorithmj from an open-source collection of network analysis toolsk was applied to the data for individual calf-contact pairs (nonaggregated data set) with open-source statistical computing software.l The algorithm was developed to detect the structure of contact patterns for members of communities within a given time period (in this case, day).21 It computes communities in large networks on the basis of random walks, which were weighted on the basis of daily animal-to-animal contact duration for this study. Thus, the algorithm detected the community structure for each study day. A random walk may be thought of as a model for an individual walking on a straight line: at each point in time that individual takes 1 step to the right with a probability of p or 1 step to the left with a probability of 1 – p.21 For the purpose of this study, random walks were weighted on the basis of the duration of contact between calves; therefore, we assumed that, for any given calf, it would be more likely (have a higher probability) to walk toward other calves with which it has longer contact times. In short, the algorithm was used to detect groups of calves that tended to spend a substantial amount of time together and formed distinct communities within the pen on each study day.

For each community on each day, the BRD case count was determined on the basis of the BRD status for each calf in the community, and that information was added to the daily record for each calf within that community. Variables generated included the BRD case count in the community for the current day (BRDcommunity), 1 (BRDcommunity1) and 2 (BRDcommunity2) days before the current day, and BRDcommunity1 and BRDcommunity2 combined (BRDcommunityTotal). For example, if on a given day, a community contained 1 calf with BRD, the BRDcommunity would be 1 for all calves in that community.

Correlation analysis

To determine whether there was a correlation between daily calf (animal-to-animal) contact patterns and initial BRD diagnosis on a given day, Pearson correlation coefficients were calculated to assess the respective correlations between the probability of BRD for an individual calf and BRDcommunity, BRDcommunityTotal, ALL, SHED, and PI. All analyses were performed with open-source statistical computing software,l and values of P < 0.05 were considered significant.

Random forest prediction of BRD status

The BRD status was a binary outcome variable for each calf on each study day. A random forest classification algorithm was used to predict BRD on the basis of animal-to-animal and community social patterns. On any given study day, the prevalence of calves with BRD was low; therefore, the prediction outcomes resulted in an imbalanced binary data set. When a data set is imbalanced, the algorithm may not obtain the necessary information about the minority class (ie, calves with BRD) to make an accurate prediction without alternate sampling strategies. For this study, we assessed oversampling, undersampling, both over- and undersampling, and random ROSE methods to compensate for the imbalanced binary data set. The ROSE method generates artificial data on the basis of the study sampling methods and a smoothed bootstrap approach. An ROC curve was created for each alternate sampling strategy to evaluate the accuracy of each method.

To determine whether animal-to-animal contact and community social structure were predictive of BRD, random forest machine learning methods were used to predict the BRD outcome of a calf for each day from days 5 through 27. Within each forest, individual trees were grown with a random sample (with replacement) of the full dataset. Each dataset consisted of the same number of rows (n) and columns (p) as the original dataset. At each node within an individual tree, the number of variables tried at each split were defined by the square root of p (rounded down). The predictive models were assessed for accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. Known calf-contact pattern data obtained prior to the day being evaluated (current day) were used to train the model, and those for the current day were used to evaluate the model. For example, for the random forest algorithm model created for day 5, calf-contact pattern data for days 0 through 4 were used to train the model, whereas the calf-contact pattern data for day 5 was used to evaluate the diagnostic performance of the model. The models for each subsequent day were created in the same manner. Predictors used in the models included ALL, SHED, PI, BRDcommunity, BRDcommunity1, BRDcommunity2, and BRDcommunityTotal. The importance of individual variables in the random forest model was evaluated by means of the mean decrease in Gini, which is an index that reflects the importance (ability) of the variable to partition the data in the model; the higher the mean decrease in Gini, the greater the importance of that variable for data partitioning. All random forest machine classification algorithms were performed with open-source statistical computing software.l

Results

Calves

Bovine respiratory disease was diagnosed once in 28 of the 70 calves, resulting in a cumulative BRD incidence rate of 40% for the study period (Figure 1). Within a given day, the BRD case count ranged from 0 to 5, with the peak BRD case count occurring on day 8. None of the calves treated for BRD required a second treatment; thus, the initial treatment (tulathromycin [2.5 mg/kg, SC]) had a 100% success rate. None of the calves were assigned a CIS of 4 during the study. One calf was found dead at approximately 7:00 am on study day 12, and moderate lung consolidation was the only gross abnormality observed during necropsy. Infectious bovine keratoconjunctivitis (pinkeye) was diagnosed in 2 calves on days 9 and 17. Those calves were treated with tulathromycin in accordance with label directions (2.5 mg/kg, SC, once). The BRD status for those 2 calves following treatment for pinkeye was excluded from all analyses because, given the characteristics of tulathromycin in cattle,22 it was believed the treatment might alter subsequent BRD risk. Twenty-three RTLS tags were lost from the ears of calves at various times throughout the study. During the morning observation, calves with missing RTLS tags were briefly removed from the pen so that a new RTLS tag could be applied to the ear. The time and date when the RTLS tag was replaced were recorded and used as the starting point during review of the locational data to estimate the time at which the prior tag was lost from the ear.

Figure 1—
Figure 1—

Bovine respiratory disease case count (black bars) and cumulative BRD incidence rate (gray line) over time for 70 weaned commercial beef steers during the first 28 days in a feedlot. For the purpose of this study, BRD was defined as a CIS > 1 and a rectal temperature > 40°C.

Citation: American Journal of Veterinary Research 79, 12; 10.2460/ajvr.79.12.1277

Contact patterns

Locational data for a full 24 hours were not collected on study days 0 and 27; therefore, ALL, SHED, and PI were not reported for those days. The ALL, SHED, and PI varied substantially among calves within and across study days (Figure 2). However, the median ALL and PI over time did not vary in a meaningful manner or have an obvious pattern. The median SHED increased and decreased over time in accordance with the number of calves presumed to be shedding BRD pathogens.

Figure 2—
Figure 2—

Box-and-whisker plots of the number of seconds each calf described in Figure 1 spent in contact with all other calves (ALL;A), calves assumed to be shedding BRD pathogens (SHED; B), and the 1 BVDV-PI (PI; C) and histogram of the number of calves assumed to be shedding BRD pathogens (D) during each day of the study period. Calves with BRD were assumed to shed BRD pathogens for the 2 days before and 5 days after BRD diagnosis. For each box-and-whisker plot, the lower and upper limits of the box represent the first and third quartiles, the horizontal line within the box represents the median, and the whiskers delimit the minimum and maximum values (excluding outliers). See Figure 1 for remainder of key.

Citation: American Journal of Veterinary Research 79, 12; 10.2460/ajvr.79.12.1277

The data set containing the number of seconds that each pair of calves was in contact during each study day consisted of 139,947 rows of data. After data for invalid and duplicated calf pairs were removed, as well as data for the 2 calves following pinkeye treatment, a total of 136,490 rows of data remained. The final aggregated data set of valid observations for individual calves on a daily basis consisted of 1,914 rows of data; the data set could have contained 1,960 rows of data had all 70 calves had valid data for all study days.

Correlation analysis

The probability of BRD for individual calves was positively correlated with SHED (r = 0.055; P = 0.017), BRDcommunity (r = 0.19; P < 0.001), and BRDcommunityTotal (r = 0.12; P < 0.001) but was not significantly correlated with ALL (r = 0.022; P = 0.337) or PI (r = 0.029; P = 0.204).

Random forest prediction of BRD

The ROC curves for the various sampling strategies used to compensate for the imbalanced binary data set during random forest modeling were plotted (Figure 3). The ROC curve for the ROSE strategy had the highest area under the ROC curve (0.821), which indicated that strategy provided the best diagnostic discrimination among the 4 strategies evaluated. Therefore, the ROSE strategy was used for all subsequent random forest models.

Figure 3—
Figure 3—

Receiver operating characteristic curves for the undersampling (dotted line), oversampling (dashed line), both under- and oversampling (both; dotted and dashed line), and ROSE (solid black line) strategies used to compensate for the imbalanced binary data set during creation of the random forest classification algorithm models used to predict BRD for the calves of Figure 1 on the basis of animal-to-animal and community social patterns. The solid gray line represents the line of identity (x = y), which would be consistent with a sampling strategy that provides no diagnostic discrimination. The ROSE strategy had the largest AUC (0.821), which indicated that strategy maximized the capability of the random forest algorithm to predict BRD, and was selected for all subsequent random forest modeling. AUC = Area under the ROC curve.

Citation: American Journal of Veterinary Research 79, 12; 10.2460/ajvr.79.12.1277

Beginning on day 5 and for each subsequent study day, the BRD prevalence in the study population and diagnostic parameters of the random forest model for predicting the probability that BRD will be diagnosed in a calf were summarized (Table 1). The negative predictive value of the random forest algorithm exceeded 90%, whereas the positive predictive value remained < 10%, for the duration of the study period. The accuracy was quite variable, ranging from 17.9% to 100%.

Table 1—

Daily prevalence of BRD for 70 weaned beef steers from days 5 through 27 after arrival at a feedlot and associated diagnostic sensitivity, specificity, accuracy, and positive and negative predictive values for prediction of BRD in individual calves as determined by means of random forest classification algorithms that were based on animal-to-animal and community social pattern data collected by an RTLS.

Study dayBRD prevalenceSensitivitySpecificityAccuracy (95% confidence interval)Positive predictive valueNegative predictive value
52.9010097.1 (89.9–99.7)97.1
61.410063.263.7 (51.3–75.0)3.8100
72.95061.160.9 (48.4–72.4)3.797.6
87.2802529.0 (18.7–41.2)7.794.1
94.566.715.617.9 (9.6–30.0)3.690.9
101.5041.540.9 (29.0–53.7)096.4
114.6055.653.0 (40.3–65.5)092.1
12095.595.5 (87.3–99.1)
131.510090.991.0 (81.5–96.6)14.3100
144.510054.756.7 (44.0–68.8)9.4100
15063.663.6 (50.9–75.1)
16381.580.6 (69.1–89.2)7.798.2
170100100 (94.6–100)
18095.595.5 (87.3–99.1)
190100100 (94.6–100)
200100100 (94.6–100)
210100100 (94.6–100)
220100100 (94.5–100)
231.5010098.5 (91.7–100)98.5
240100100 (94.5–100)
250100100 (94.5–100)
260100100 (94.5–100)
271.8010098.2 (90.3–100)98.2

All values are percentages unless otherwise indicated. On any given day, the BRD prevalence was low; therefore, the prediction outcomes resulted in an imbalanced binary data set. A ROSE method was used to compensate for the imbalanced data set during random forest modeling. Known calf-contact pattern data obtained for all days before the day being evaluated (current day) were used to train the model, and that for the current day were used to evaluate the model. Predictors used in the models included the number of seconds that an individual calf spent in contact with all other calves (ALL), calves presumed to be shedding BRD pathogens (SHED), and the 1 BVDV-PI (PI), as well as the number of calves with BRD within the community on the current day (BRDcommunity), 1 (BRDcommunity1) and 2 (BRDcommunity2) days before the current day, and the sum of BRDcommunity1 and BRDcommunity2 (BRDcommunityTotal). — = Not calculated.

The 3 most important predictors varied with each daily iteration of the random forest model (Table 2). Nevertheless, BRDcommunity, BRDcommunityTotal, and PI were the predictors most frequently identified as important for prediction of BRD across all study days, which suggested that the social network patterns of calves within a pen might contribute to the transmission of BRD.

Table 2—

The top 3 predictors from the random forest model for each study day ranked on the basis of importance as determined by the mean decrease in Gini.

 Predictor importance ranking (mean decrease in Gini)
Study day123
5BRDcommunity (68.6)SHED (37.9)BRDCommunityTotal (19.3)
6BRDcommunity (59.8)BRDcommunity2 (56.5)BRDCommunityTotal (53.6)
7BRDcommunity (67.4)BRDCommunityTotal (64.4)BRDcommunity2 (60.3)
8BRDcommunity (118.1)BRDCommunityTotal (59.3)BRDcommunity2 (26.3)
9BRDcommunity (86.9)BRDCommunityTotal (71.5)BRDcommunity2 (39.3)
10BRDcommunity (96.5)BRDCommunityTotal (52.4)BRDcommunity2 (49.3)
11BRDcommunity (112.4)BRDCommunityTotal (51.3)PI (50.6)
12BRDcommunity (72.3)BRDCommunityTotal (67.5)PI (60.6)
13BRDcommunity (84.9)BRDCommunityTotal (75.2)PI (64.9)
14BRDcommunityTotal (95.6)BRDCommunity (89.7)PI (67.1)
15BRDcommunityTotal (102.5)BRDCommunity (90.7)PI (78.4)
16BRDCommunity (109.2)BRDcommunityTotal (90.32)PI (80.44)
17BRDCommunity (128.50)BRDcommunityTotal (94.81)PI (59.19)
18BRDCommunity (134.75)BRDcommunityTotal (101.89)BRDcommunity1 (65.64)
19BRDCommunity (157.83)BRDcommunityTotal (104.22)BRDcommunity1 (59.03)
20BRDCommunity (163.79)BRDcommunityTotal (99.85)BRDcommunity1 (71.48)
21BRDCommunity (165.01)BRDcommunityTotal (109.17)PI (64.74)
22BRDCommunity (176.57)BRDcommunityTotal (112.99)BRDcommunity1 (72.36)
23BRDCommunity (190.80)BRDcommunityTotal (126.85)BRDcommunity2 (64.56)
24BRDCommunity (189.89)BRDcommunityTotal (139.88)BRDcommunity1 (67.39)
25BRDCommunity (209.35)BRDcommunityTotal (136.86)BRDcommunity1 (74.59)
26BRDCommunity (197.03)BRDcommunityTotal (142.26)BRDcommunity1 (78.31)
27BRDCommunity (218.14)BRDcommunityTotal (143.23)BRDcommunity1 (79.69)

The mean decrease in Gini is an index that reflects the ability of a variable to partition the data in the model; the higher the mean decrease in Gini, the greater the importance (ability) of the variable for data partitioning.

See Table 1 for remainder of key.

Discussion

The role of contact patterns among cattle on the spread of BRD within populations has yet to be determined. In the present study, a group of weaned beef calves was instrumented with RTLS tags and commingled in a single pen at a research facility that mimicked a commercial feedlot setting. The movement of those calves was continuously monitored for 28 days, and the incidence of BRD was recorded. Evaluation of the locational data for the calves of the present study revealed that there was substantial heterogeneity in cattle contact structures among individual calves as well as among study days. That finding was consistent with results of another study13 in which the temporal and spatial movements of calves within their environment were monitored and suggested that individual animals may have a critical role in a disease epidemic.

Results of the present study indicated that BRD was positively correlated with SHED, BRDCommunity, and BRDCommunityTotal, all of which reflected duration of contact with sick animals. That finding supported the hypothesis that direct transmission of BRD pathogens between calves occurs and that, for an unaffected animal, an increase in contact with cattle shedding BRD pathogens increases the risk of that animal developing the disease. The magnitude of the correlation between those variables and BRD likely depends on several factors, such as host susceptibility, extent of pathogen shedding by affected animals, and history of prior pathogen exposure.

With this study, we demonstrated that calf contact structure and community contact structure data could be used in random forest models to predict the BRD outcome for individual calves within a population. The daily random forest models created from the locational data obtained from the calves of the present study had high, albeit variable, accuracy for predicting the BRD outcome for individual calves. The negative predictive values for the models were likewise quite high (> 90%), whereas the positive predictive values were low (< 10%). This was, at least in part, a reflection of the fact that the BRD prevalence within the study population was low on any given day, despite the fact that the cumulative incidence of BRD for the study period was quite high (40%). The final data set consisted of 1,914 calf days, of which only 28 (1.5%) were positive for BRD (ie, BRD was diagnosed in 28 calves throughout the study period, and BRD was not diagnosed in any calf more than once), which resulted in an imbalanced data set in terms of BRD status. For any diagnostic modality, the predictive values are extremely sensitive to the prevalence of the target condition within the population; the negative predictive value increases and the positive predictive value decreases as the prevalence of the target condition decreases and vice versa. Thus, it is often difficult to achieve high positive predictive values with a single machine learning algorithm, especially when the data set is limited in size and the target condition (BRD) has a low prevalence, as in this study.

To our knowledge, the present study was the first to describe use of a random forest algorithm to predict BRD on the basis of animal-to-animal social contact patterns. The accuracy of that algorithm was not perfect, likely owing to the limited sample size and study period and low BRD prevalence within the population, and might be improved if other predictors of BRD (behavioral outcomes) were incorporated in addition to the animal contact and community data. Despite the limitations of the algorithm in terms of accuracy and positive predictive value, results suggested that the contact predictors used in this study can be used to identify animals at high risk for BRD. Use of cattle social patterns within a pen to predict BRD is a novel and practical approach that, with further research and refinement, might be clinically useful in the future.

The variables used in the random forest algorithm might have been influenced by the BRD case definition and the assumed pathogen shedding period for BRD-affected calves. Although the duration of pathogen shedding assumed for the BRD-affected calves of this study (ie, from 2 days before to 5 days after BRD diagnosis) was supported by the scientific literature,20 it did affect the number of cattle assumed to be shedding BRD pathogens on any given day, and altering the shedding window could change the results of the analysis. However, we believe the positive correlation between duration of contact with BRD-affected calves that may be shedding pathogens and probability of BRD diagnosis will likely hold true regardless of the shedding period used. That finding has potential implications for the management of BRD-affected cattle in commercial feedlots. Although some commercial feedlots move cattle to hospital pens following treatment for BRD, most return BRD-affected cattle to their pen of origin within 24 hours after diagnosis and treatment.23 On the basis of the results of the present study, we hypothesize that returning BRD-affected cattle to their pen of origin within 24 hours after diagnosis and treatment will increase the risk of disease for pen mates, and that isolation of BRD-affected cattle until pathogen shedding has likely ceased might decrease BRD incidence in the general population by minimizing contact between diseased and undiseased cattle.

The identification of cattle likely to shed BRD pathogens in the future on the basis of the duration of their contact with BRD-affected cattle might also help decrease BRD incidence in the population if those calves are removed from the pen prior to initiation of pathogen shedding. Of course, that would require real-time monitoring of calf contacts. Various indirect behavioral monitoring technologies for early detection and diagnosis of BRD in cattle have been investigated.24,25 In 1 study,16 use of an RTLS was able to detect cattle suspected of having BRD 0.75 days sooner than trained observers. In the present study, calves with BRD were identified by trained observers rather than the RTLS, and we expect that calves would have been shedding BRD pathogens for only 1.25 days (vs 2 days) had we used the RTLS to identify BRD-affected calves. In the future, RTLS data might be used to determine the social network pattern for groups of cattle and the extent of contact between cattle with and without BRD. That information could then be used in predictive algorithms to help determine the BRD risk for individual calves. Calves at high risk for the disease could then be closely monitored so that, should they develop BRD, they could be removed from the pen and treated earlier in the disease process. However, that will require further research and elucidation of the role of cattle contact structures in the transmission of naturally occurring BRD.

In a study10 of Australian feedlot cattle, seroconversion to respiratory tract pathogens following study enrollment was positively associated with the odds of BRD, which suggests direct transmission of respiratory pathogens among feedlot cattle. Serum antibody titers against specific pathogens reflect the animal's ability to mount a humoral immune response against a challenge by those pathogens, and were significantly associated with the BRD outcomes for the cattle of that study.10 We did not evaluate serum antibody titers against BRD pathogens for the calves of the present study, but had we done so prior to study enrollment, we could have identified seronegative calves, which theoretically would be at greater risk for BRD than seropositive calves. That information could then have been incorporated with the contact data and included in the random forest algorithm, which might have improved the model's ability to estimate the probability of BRD for individual calves. Further research evaluating the effect of serum antibody status on the association between cattle contact patterns and BRD is warranted.

The effect of the presence of a BVDV-PI on the risk for BRD in a cohort of cattle is unclear. In 1 study,26 the presence of a BVDV-PI in a cohort of cattle was positively associated with the risk of BRD in that population. However, results of another study27 indicate that the presence of a BVDV-PI in a cohort of cattle was associated with a decrease in the risk of BRD morbidity and death. There was 1 BVDV-PI among the calves of the present study, and persistent infection with BVDV was identified as one of the most important factors for prediction of BRD. We believe that finding was consistent with the previous literature, although additional research is necessary to specifically determine the role that contact with a BVDV-PI may have on BRD outcome at both the individual animal and group levels.

A weakness of the present study was that pathogen shedding by BRD-affected calves was not confirmed by means of virus isolation and other diagnostic modalities, but we believe the pathogen shedding window assumed for BRD-affected calves was well supported by the scientific literature.20 Collection of biologic samples from calves to monitor viral or bacterial shedding on a daily basis would not be done in a commercial feedlot setting (ie, would not be externally valid), and the extra handling required to collect such samples likely would have affected cattle behavior and performance,28,29 which in turn would alter cattle social behavior and pathogen transmission. Nevertheless, pathogen shedding data might have been beneficial in the present study had it been able to be obtained without altering calf behavior.

The loss of RTLS transmitter tags (retention failure) from some calves was a minor problem in the present study. Reasons for retention failure included calf interactions with facilities and poor tag placement. For calves that lost an RTLS tag, the data collected between the time when the tag was lost and its replacement with another was coded as missing in the raw data set. Evaluation of the raw data set failed to reveal any pattern to the missing data, and it was assumed that the missing data were random.30 Moreover, there did not appear to be any association between retention failure and BRD status or calf contact time.

We were unable to document the extent of contact among the study calves prior to arrival at the research facility. Knowledge of the extent of commingling among and the number of morbid animals to which the study calves were exposed prior to study initiation would have enhanced our understanding of the BRD risk associated with calf contact at study onset. Also, no diagnostic testing (eg, virus isolation or determination of serum antibody titers) aside from clinical observation was performed on the study calves at arrival to the research facility. Definitive information regarding the disease status of calves at study initiation might have helped refine the role of animal contact on BRD risk, and future studies should account for that information as it relates to disease dynamics and social behavior within cohorts of cattle. Collecting samples from calves at various times during the study or at study conclusion might have provided some indication of which animals seroconverted or were exposed to specific pathogens during the observation period. The calves of the present study were of similar age and weight, and were purchased, transported, and commingled at the research facility just prior to study initiation; therefore, they were considered at similar risk for BRD. Future studies of the effect of cattle contact on disease risk should be conducted with cattle of varying risk for BRD (ie, cattle of different types [beef vs dairy or feedlot cattle vs herd replacements], ages, or serologic status or in various housing [open-air drylot vs enclosed or semienclosed barn] and weather conditions) because that will help enhance the understanding of BRD risk and cattle management.

Results of the present study indicated that direct transmission of BRD pathogens likely occurs among cattle housed in commercial feedlots. The results also suggested that the duration an individual animal spends in contact with a BRD-affected animal is positively correlated with the probability it will subsequently develop BRD, although the magnitude of that correlation was low and the importance of animal-to-animal contact on BRD transmission remains poorly understood. The RTLS used in the present study provided continuous locational data that were used to determine the extent of animal-to-animal contact and community social structure for a cohort of calves. The ability to quantify cattle contact structures will enhance the study of transmissible diseases and facilitate the development of disease mitigation strategies.

Acknowledgments

Supported by the Kansas State University Department of Clinical Sciences and the Beef Cattle Institute.

Dr. White is an owner-partner in Precision Animal Solutions, the company that set up, installed, and managed the RTLS used in the study.

The authors thank Dr. Sarah Capik, Hannah Seger, Alison Hoehn, and Tara Fountain for technical assistance.

ABBREVIATIONS

BRD

Bovine respiratory disease

BVDV

Bovine viral diarrhea virus

BVDV-PI

Calf persistently infected with bovine viral diarrhea virus

CIS

Clinical illness score

ROC

Receiver operating characteristic

ROSE

Random over sampling examples

RTLS

Real-time location system

Footnotes

a.

Smartbow GmbH, Weibern, Austria.

b.

Bovi-Shield Gold 5, Zoetis, Florham Park, NJ.

c.

Ultrabac 7, Zoetis, Florham Park, NJ.

d.

Agrimectin, AgriLabs, St Joseph, Mo.

e.

Draxxin, Zoetis, Florham Park, NJ.

f.

Nuflor, Merck Animal Health, Madison, NJ.

g.

Liquamycin LA-200, Zoetis, Florham Park, NJ.

h.

Precision Animal Solutions, Manhattan, Kan.

i.

KNIME Analytics Platform, KNIME AG, Zurich, Switzerland.

j.

WalkTrap, version 0.2. Available at: www-complexnetworks.lip6.fr/~latapy/PP/walktrap.html. Accessed Jun 26, 2017.

k.

Csardi G, Nepusz T. The igraph software package for complex network research, InterJournal, Complex Systems 1695. 2006.

l.

R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at: www.R-project.org. Accessed Aug 22, 2017.

References

  • 1. Griffin D. Economic impact associated with respiratory disease in beef cattle. Vet Clin North Am Food Anim Pract 1997;13:367377.

  • 2. Snowder GD, Van Vleck LD, Cundiff LV, et al. Bovine respiratory disease in feedlot cattle: environmental, genetic, and economic factors. J Anim Sci 2006;84:19992008.

    • Search Google Scholar
    • Export Citation
  • 3. Cernicchiaro N, White BJ, Renter DG, et al. Associations between the distance traveled from sale barns to commercial feedlots in the United States and overall performance, risk of respiratory disease, and cumulative mortality in feeder cattle during 1997 to 2009. J Anim Sci 2012;90:19291939.

    • Search Google Scholar
    • Export Citation
  • 4. Taylor JD, Fulton RW, Lehenbauer TW, et al. The epidemiology of bovine respiratory disease: what is the evidence for predisposing factors? Can Vet J 2010;51:10951102.

    • Search Google Scholar
    • Export Citation
  • 5. Babcock AH, Cernicchiaro N, White BJ, et al. A multivariable assessment quantifying effects of cohort-level factors associated with combined mortality and culling risk in cohorts of US commercial feedlot cattle. Prev Vet Med 2013;108:3846.

    • Search Google Scholar
    • Export Citation
  • 6. Cernicchiaro N, Renter DG, White BJ, et al. Associations between weather conditions during the first 45 days after feedlot arrival and daily respiratory disease risks in autumn-placed feeder cattle in the United States. J Anim Sci 2012;90:13281337.

    • Search Google Scholar
    • Export Citation
  • 7. Hay KE, Barnes TS, Morton JM, et al. Risk factors for bovine respiratory disease in Australian feedlot cattle: use of a causal diagram–informed approach to estimate effects of animal mixing and movements before feedlot entry. Prev Vet Med 2014;117:160169.

    • Search Google Scholar
    • Export Citation
  • 8. Hay KE, Morton JM, Clements AC, et al. Associations between feedlot management practices and bovine respiratory disease in Australian feedlot cattle. Prev Vet Med 2016;128:2332.

    • Search Google Scholar
    • Export Citation
  • 9. Theurer ME, White BJ, Larson RL, et al. Relationship between rectal temperature at first treatment for bovine respiratory disease complex in feedlot calves and the probability of not finishing the production cycle. J Am Vet Med Assoc 2014;245:12791285.

    • Search Google Scholar
    • Export Citation
  • 10. Hay KE, Barnes TS, Morton JM, et al. Associations between exposure to viruses and bovine respiratory disease in Australian feedlot cattle. Prev Vet Med 2016;127:121133.

    • Search Google Scholar
    • Export Citation
  • 11. Martin SW, Bateman KG, Shewen PE, et al. A group level analysis of the associations between antibodies to seven putative pathogens and respiratory disease and weight gain in Ontario feedlot calves. Can J Vet Res 1990;54:337342.

    • Search Google Scholar
    • Export Citation
  • 12. Martin SW, Darlington G, Bateman K, et al. Undifferentiated bovine respiratory disease (shipping fever): is it communicable? Prev Vet Med 1988;6:2735.

    • Search Google Scholar
    • Export Citation
  • 13. Chen S, Sanderson MW, White BJ, et al. Temporal-spatial heterogeneity in animal-environment contact: implications for the exposure and transmission of pathogens. Sci Rep 2013;3:3112.

    • Search Google Scholar
    • Export Citation
  • 14. Chen S, White BJ, Sanderson MW, et al. Highly dynamic animal contact network and implications on disease transmission. Sci Rep 2014;4:4472.

    • Search Google Scholar
    • Export Citation
  • 15. Porto SMC, Arcidiacono C, Giummarra A, et al. Localisation and identification performances of a real-time location system based on ultra wide band technology for monitoring and tracking dairy cow behaviour in a semi-open free-stall barn. Comput Electron Agric 2014;108:221229.

    • Search Google Scholar
    • Export Citation
  • 16. White BJ, Goehl DR, Amrine DE. Comparison of a remote early disease identification (REDI) system to visual observations to identify cattle with bovine respiratory diseases. Int J Appl Res Vet Med 2015;13:2330.

    • Search Google Scholar
    • Export Citation
  • 17. White BJ, Anderson DE, Renter DG, et al. Clinical, behavioral, and pulmonary changes in calves following inoculation with Mycoplasma bovis. Am J Vet Res 2012;73:490497.

    • Search Google Scholar
    • Export Citation
  • 18. Theurer ME, Anderson DE, White BJ, et al. Effect of Mannheimia haemolytica pneumonia on behavior and physiologic responses of calves during high ambient environmental temperatures. J Anim Sci 2013;91:39173929.

    • Search Google Scholar
    • Export Citation
  • 19. Shane DD, White BJ, Larson RL, et al. Probabilities of cattle participating in eating and drinking behavior when located at feeding and watering locations by a real time location system. Comput Electron Agric 2016;127:460466.

    • Search Google Scholar
    • Export Citation
  • 20. Grissett GP, White BJ, Larson RL. Structured literature review of responses of cattle to viral and bacterial pathogens causing bovine respiratory disease complex. J Vet Intern Med 2015;29:770780.

    • Search Google Scholar
    • Export Citation
  • 21. Pons P, Latapy M. Computing communities in large networks using random walks. J Graph Algorithms Appl 2006;10:191218.

  • 22. Evans NA. Tulathromycin: an overview of a new triamilide antibiotic for livestock respiratory disease. Vet Ther 2005;6:8395.

  • 23. National Animal Health Monitoring System. Feedlot 2011. Part IV: health and health management on US feedlots with a capacity of 1,000 or more head. Fort Collins, Colo: USDA APHIS Veterinary Services Center for Epidemiology and Animal Health, 2013. Available at: www.aphis.usda.gov/animal_health/nahms/feedlot/downloads/feedlot2011/Feed11_dr_PartIV.pdf. Accessed Jun 8, 2018.

    • Search Google Scholar
    • Export Citation
  • 24. Pillen JL, Pinedo PJ, Ives SE, et al. Alteration of activity variables relative to clinical diagnosis of bovine respiratory disease in newly received feedlot cattle. Bovine Pract 2016;50:18.

    • Search Google Scholar
    • Export Citation
  • 25. Jackson KS, Carstens GE, Tedeschi LO, et al. Changes in feeding behavior patterns and dry matter intake before clinical symptoms associated with bovine respiratory disease in growing bulls. J Anim Sci 2016;94:16441652.

    • Search Google Scholar
    • Export Citation
  • 26. Loneragan GH, Thomson DU, Montgomery DL, et al. Prevalence, outcome, and health consequences associated with persistent infection with bovine viral diarrhea virus in feedlot cattle. J Am Vet Med Assoc 2005;226:595601.

    • Search Google Scholar
    • Export Citation
  • 27. O'Connor AM, Sorden SD, Apley MD. Association between the existence of calves persistently infected with bovine viral diarrhea virus and commingling on pen morbidity in feedlot cattle. Am J Vet Res 2005;66:21302134.

    • Search Google Scholar
    • Export Citation
  • 28. Grandin T. Assessment of stress during handling and transport. J Anim Sci 1997;75:249257.

  • 29. Francisco CL, Cooke RF, Marques RS, et al. Effects of temperament and acclimation to handling on feedlot performance of Bos taurus feeder cattle originated from a rangeland-based cow-calf system. J Anim Sci 2012;90:50675077.

    • Search Google Scholar
    • Export Citation
  • 30. Dohoo I, Marten W, Stryhn H. Veterinary epidemiologic research. 2nd ed. Charlottetown, PEI, Canada: VER Inc, 2009;374375.

Appendix

Description of the CIS system used to evaluate 70 weaned commercial beef steers for signs of BRD during the first 28 days after arrival at a feedlot.

CISDefinition
1Normal (no abnormalities observed)
2Slight illness (mild depression with or without cough, nasal discharge, or ocular discharge)
3Moderate illness (severe depression with or without labored breathing, cough, nasal discharge, or ocular discharge)
4Moribund (little response to stimulus)

Contributor Notes

Dr. Shane's present address is Bayer Corp, 12707 Shawnee Mission Pkwy, Shawnee, KS 66216.

Address correspondence to Dr. White (bwhite@vet.k-state.edu).