J.M. Huzzey

BOARD-INVITED REVIEW: Using behavior to predict and identify ill health in animals1

We review recent research in one of the oldest and most important applications of ethology: evaluating animal health. Traditionally, such evaluations have been based on subjective assessments of debilitative signs; animals are judged ill when they appear depressed or off feed. Such assessments are prone to error but can be dramatically improved with training using well-defined clinical criteria. The availability of new technology to automatically record behaviors allows for increased use of objective measures; automated measures of feeding behavior and intake are increasingly available in commercial agriculture, and recent work has shown these to be valuable indicators of illness. Research has also identified behaviors indicative of risk of disease or injury. For example, the time spent standing on wet, concrete surfaces can be used to predict susceptibility to hoof injuries in dairy cattle, and time spent nuzzling the udder of the sow can predict the risk of crushing in piglets. One conceptual advance has been to view decreased exploration, feeding, social, sexual, and other behaviors as a coordinated response that helps afflicted individuals recover from illness. We argue that the sickness behaviors most likely to decline are those that provide longer-term fitness benefits (such as play), as animals divert resources to those functions of critical short-term value such as maintaining body temperature. We urge future research assessing the strength of motivation to express sickness behaviors, allowing for quantitative estimates of how sick an animal feels. Finally, we call for new theoretical and empirical work on behaviors that may act to signal health status, including behaviors that have evolved as honest (i.e., reliable) signals of condition for offspring-parent, inter- and intra-sexual, and predator-prey communication.

Short communication: Haptoglobin as an early indicator of metritis

The purpose of this study was to determine whether haptoglobin (Hp) could be used as a predictive measure for metritis. Cattle were grouped into 3 health categories based on the condition of vaginal discharge and body temperature after calving: severe metritis (n = 12), mild metritis (n = 32), and healthy (n = 23). Blood was collected and analyzed for Hp concentration on d -20 +/- 5, -6 +/- 2, -2 +/- 1, and d 0 relative to calving, and then every 3 d after calving until d +21. Cows with mild and severe metritis had greater Hp concentrations than healthy cows between d 0 and d +12. Mean (+/-SE) Hp concentrations peaked on d +3 in the cows with mild metritis (1.06 +/- 0.15 g/L) and on d +6 in cows with severe metritis (1.62 +/- 0.47 g/L). Mean concentrations for the healthy group were 0.58 +/- 0.12 g/L and 0.31 +/- 0.08 g/L on d +3 and d +6, respectively. Clinical signs of pathological discharge for the mildly and severely metritic cows did not occur until, on average, 8.6 +/- 3.9 d and 5.3 +/- 1.9 d after calving, respectively. Cows with Hp concentrations >or=1 g/L on d +3 were 6.7 times more likely to develop severe or mild metritis; this predictive threshold has a sensitivity of 50% and specificity of 87%. These results indicate that an acute phase inflammatory response precedes clinical metritis and that Hp screening may assist in the early detection of metritis, providing increased opportunities for early treatment and prevention.

The effect of dystocia on the dry matter intake and behavior of Holstein cows

Dairy cows that have a difficult calf delivery (dystocia) are more likely to develop health complications after calving, reducing productivity and welfare. Understanding the behavioral cues of dystocia may facilitate prompt obstetric assistance and reduce the long-term effect of the challenging delivery. The aim of this study was to describe the effects of dystocia on dairy cow behavior during the period around calving and to assess the use of these behaviors as potential indicators of dystocia. Individual dry matter intake, water intake, feeding and drinking time, meal size, standing time, and number of transitions from standing to lying positions (bouts) were recorded during the 48-h period before and after the time of calf delivery for 22 Holstein cows [11 cows with dystocia and 11 cows with unassisted delivery (eutocia)]. Cows with dystocia consumed 1.9 kg less during the 48 h before calving compared with cows with eutocia (14.3 +/- 1.0 vs. 16.2 +/- 1.0 kg, respectively), and this difference increased to 2.6 kg in the 24 h before calving (8.3 +/- 0.7 vs. 10.9 +/- 0.7 kg/d). There were no differences in drinking time between the groups, but cows with dystocia consumed less water 24 h before calving (22.4 +/- 4.4 vs. 36.2 +/- 4.4 kg/d, respectively) and consumed more water during the 24-h period after calving (56.9 +/- 3.1 vs. 48.7 +/- 3.1 kg/d) compared with cows with eutocia. Cows with dystocia transitioned from standing to lying positions more frequently than cows without dystocia beginning 24 h before calving (10.9 +/- 0.7 vs. 8.3 +/- 0.7 bouts/d). Dry matter intake and standing bouts in the 24 h before calving were the most accurate variables in discriminating between cows with and without dystocia, suggesting that cows with dystocia begin to alter their behavior beginning 24 h before calving.

Associations of prepartum plasma cortisol, haptoglobin, fecal cortisol metabolites, and nonesterified fatty acids with postpartum health status in Holstein dairy cows

The association between negative energy balance and health has led to the testing of blood analytes such as nonesterified fatty acids (NEFA) to identify opportunities for improving the management of transition dairy cows. The objective of this study was to evaluate whether prepartum analytes associated with stress (cortisol) or inflammation (haptoglobin) could also identify dairy cattle at increased risk for health complications after calving. Prepartum blood and fecal samples were collected once weekly from 412 Holstein dairy cows on 2 commercial dairy farms (at wk -3, -2, and -1 relative to calving) and analyzed for concentrations of NEFA, haptoglobin (Hp), and cortisol in plasma and cortisol metabolites in feces. Retained placenta (RP), displaced abomasum (DA), subclinical ketosis (SCK), high Hp concentration (HiHp), and death were recorded up to 30 d in milk (DIM), and animals were subsequently categorized into 3 health categories: (1) no disorder of interest (NDI); (2) one disorder (RP, DA, SCK, or HiHp); or (3) more than one disorder (RP, DA, SCK, HiHp) or death. With the exception of prepartum NEFA, no associations were detected between prepartum concentrations of our analytes of interest and the occurrence of one disorder (RP, DA, SCK, or HiHP) by 30 DIM. Haptoglobin concentration tended to be greater during wk -2 and -1 in cows that developed more than one disorder or that died by 30 DIM; however, when calving assistance was included as a covariate in the analysis prepartum, Hp was no longer a significant risk factor for this postpartum health outcome. Primiparous cows with plasma cortisol concentrations >22.2 nmol/L during wk -2 had reduced odds [odds ratio (OR) 0.41; 95% confidence interval (CI) 0.17-0.98] of developing more than one disorder or death by 30 DIM, whereas multiparous cows with plasma cortisol >34.1 nmol/L during wk -2 tended to have greater odds (OR 2.53; 95% CI 0.87-7.37) of developing more than one disorder or death by 30 DIM. Individual variation in daily cortisol secretion patterns and stress responses to the restraint and handling associated with sample collection make prepartum plasma cortisol data and its relationship to postpartum health difficult to interpret. Among multiparous cows, for every 500-unit (ng/g of fecal dry matter) increase in fecal cortisol metabolite concentration during wk -2, cows had increased odds (OR 1.41; 95% CI 1.12-1.79) of developing more than one disorder or dying after calving. For multiparous cows, every 0.15 mmol/L increase in plasma NEFA concentration during any of the 3 wk before calving was associated with an approximately 2-fold increase in the odds of developing more than one disorder or dying by 30 DIM. Fecal cortisol metabolite concentration during the prepartum period did not predict which cows would go on to develop more than one disorder or die within 30 DIM as accurately as prepartum NEFA concentration; therefore, this analyte is not a suitable substitute for NEFA for assessing opportunities to improve herd health.

Associations of subclinical hypocalcemia at calving with milk yield, and feeding, drinking, and standing behaviors around parturition in Holstein cows.

The objectives of this study were to describe the associations of subclinical hypocalcemia with milk yield, and feeding, drinking, and resting behavior during the period around calving. Blood was sampled within 24h of calving and analyzed for serum total calcium. Fifteen Holstein dairy cows were classified as having subclinical hypocalcemia (serum calcium concentration ≤ 1.8 mmol/L, without clinical milk fever) and were matched with 15 control cows (serum calcium concentration >1.8 mmol/L) based on parity and presence of other diseases. Daily feeding and drinking behavior were monitored using an electronic feeding system (Insentec, BV, Marknesse, the Netherlands) and summarized by week relative to calving (wk -3, -2, -1, +1, +2, and +3). Standing behavior was monitored from 7 d before until 7 d after calving using dataloggers. Daily milk yields were obtained for all cows up to 280 d in milk (DIM). These data were summarized by week for the first 4 wk of lactation to assess short-term differences in milk yield, and were summarized into 4-wk periods to assess long-term (280 DIM) differences in milk yield between groups. Cows with subclinical hypocalcemia produced, on average, 5.7 kg/d more milk during wk 2, 3, and 4 compared with control cows; however, only subclinically hypocalcemic cows in their third lactation sustained greater milk yields throughout 280 DIM. Despite greater milk yield during the weeks following calving, cows with subclinical hypocalcemia did not consume more water after calving and tended to have greater dry matter intake only during wk 2. However, these animals made fewer visits to the water bins during the first 2 wk after calving and tended to make fewer visits to the feed bins during wk 1 and 3, suggesting that they used these resources more efficiently. Dry matter intake was, on average, 1.7 kg/d greater during wk -2 and -1 among cows subsequently diagnosed with subclinical hypocalcemia compared with control cows but neither group was lactating during this period. Cows with subclinical hypocalcemia stood for 2.6h longer during the 24-h period before parturition, which may suggest these animals experience increased discomfort at calving; these cows spent 2.7h less time standing during d +1. Although milk yield was greater among cows with subclinical hypocalcemia, this study controlled for the confounding effects of disease incidence; these results do not refute previous research that associates subclinical hypocalcemia with an increased risk for health disorders. The mechanisms by which subclinical hypocalcemia is associated with behavior and production require further investigation.

The effects of overstocking Holstein dairy cattle during the dry period on cortisol secretion and energy metabolism

The objective was to determine whether overstocking during the dry period could alter physiological parameters in dairy cattle associated with cortisol secretion and energy metabolism. Four groups of 10 late-gestation, nonlactating Holstein cows (6 multiparous cows and 4 heifers per group) were exposed to both a control [1 lying stall/cow and 0.67 m of linear feed bunk (FB) space/cow] and an overstocked (1 stall/2 cows and 0.34 m of FB space/cow) stocking density treatment in a replicated crossover design with 14-d treatment periods. On d 1, 3, 5, 7, 9, and 11 of each 14-d treatment period, blood and fecal samples were collected from each cow for the determination of plasma nonesterified fatty acids (NEFA), glucose, insulin, and fecal cortisol metabolite (11,17-dioxoandrostane; 11,17-DOA) concentrations. Glucose and ACTH challenges were conducted on d 13 and 14, respectively, of each treatment period. Dry matter intake per cow was greater during the overstocked period than during the control period (15.9 vs. 14.9 ± 0.5 kg/d). Plasma NEFA and glucose concentrations were greater (0.11 vs. 0.09 ± 0.006 mEq/L and 65.3 vs. 64.2 ± 1.1mg/dL, respectively) and 11,17-DOA concentration tended to be greater (891 vs. 792 ± 86 ng/g of fecal dry matter) during the overstocked period than during the control period. Insulin concentration was the same during the overstocked (29.0 ± 2.1 μIU/mL) and control (31.2 ± 2.1 μIU/mL) periods. Overstocking was associated with slightly slower glucose clearance from circulation as evidenced by a greater area under the curve estimate for the glucose response curves (2,882 vs. 2,657 ± 165 mg/dL × 180 min) but a more attenuated insulin response (insulin area under the curve = 5,258 vs. 6,692 ± 1,104 μIU/mL × 180 min for the overstocked and control periods, respectively). Changes in tissue glucose uptake may be mediated by changes in pancreatic insulin secretion or peripheral tissue responses to insulin. The role of glucocorticoids in mediating these changes in energy metabolism is still unclear because stocking density treatment was not associated with changes in adrenal secretion of cortisol following ACTH stimulation.

Clinical ketosis and standing behavior in transition cows

Ketosis is a common disease in dairy cattle, especially in the days after calving, and it is often undiagnosed. The objective of this study was to compare the standing behavior of dairy cows with and without ketosis during the days around calving to determine if changes in this behavior could be useful in the early identification of sick cows. Serum β-hydroxybutyrate (BHBA) was measured in 184 cows on a commercial dairy farm twice weekly from 2 to 21d after calving. Standing behavior was measured from 7d before calving to 21d after calving using data loggers. Retrospectively, 15 cows with clinical ketosis (3 consecutive BHBA samples >1.2mmol/L and at least one sample of BHBA >2.9mmol/L) were matched with 15 nonketotic cows (BHBA <1.2mmol/L). Five periods were defined for the statistical analyses: wk -1 (d -7 to -1), d 0 (day of calving), wk +1 (d 1 to 7), wk +2 (d 8 to 14), and wk +3 (d 15 to 21). The first signs of clinical ketosis occurred 4.5±2.1d after calving. Total daily standing time was longer for clinically ketotic cows compared with nonketotic cows during wk -1 (14.3±0.6 vs. 12.0±0.7h/d) and on d 0 (17.2±0.9 vs. 12.7±0.9h/d) but did not differ during the other periods. Clinically ketotic cows exhibited fewer standing bouts compared with nonketotic cows on d 0 only (14.6±1.9 vs. 20.9±1.8bouts/d). Average standing bout duration was also longer for clinically ketotic cows on d 0 compared with nonketotic cows [71.3min/bout (CI: 59.3 to 85.5) vs. 35.8min/bout (CI: 29.8 to 42.9)] but was not different during the other periods. Differences in standing behavior in the week before and on the day of calving may be useful for the early detection of clinical ketosis in dairy cows.

Short communication: Relationship between competitive success during displacements at an overstocked feed bunk and measures of physiology and behavior in Holstein dairy cattle

The objective of this study was to evaluate how behavioral and physiological parameters are affected based on a cows level of success at displacing others at an overstocked feed bunk. Forty Holstein nonlactating, late-gestation dairy cattle were housed in an overstocked pen [5 stalls/10 cows and 0.34 m of linear feed bunk (FB) space/cow] in groups of 10 (4 heifers and 6 multiparous cows) for 14 d. Plasma nonesterified fatty acids, glucose, and fecal cortisol metabolites (11,17-dioxoandrostanes) were measured in blood and feces sampled every 2d. A glucose tolerance test and an ACTH challenge were conducted on all cows on d 13 and 14, respectively to further explore the effects of competitive success on energy metabolism and stress physiology. Feeding behavior and displacements at the FB were recorded between d 7 to 10 of the observation period. A competition index (CInd) was calculated for each cow by dividing the number of times the cow displaced another at the FB by the total number of displacements the cow was involved in, either as an actor or reactor. Cows were then divided into 3 subgroups based on their CInd: high success (HS: CInd ≥0.6), medium success (0.4 ≤ CInd <0.6), and low success (LS: CInd <0.4). Heifers accounted for 7, 36, and 79% of the total number of animals in the HS (n=15), medium success (n=11), and LS (n=14) groups, respectively. No differences were observed in daily feeding time, total number of displacements, and time to approach the FB following fresh feed delivery between the 3 CInd groups; however, cows in the LS group had greater daily nonesterified fatty acid and 11,17-dioxoandrostane concentrations relative to cows in the HS group. No differences existed in cortisol response to an ACTH stimulation test between CInd categories. During the glucose tolerance test, glucose response curves were the same between all 3 CInd categories; however, the peak insulin response of LS cows was 130 μIU/mL greater than the peak HS response, indicating that LS cows may have decreased tissue responses to insulin or increased pancreatic responses to glucose. In an overstocked environment, dairy cattle physiology is associated with a cows level of success at displacing other individuals at the feed bunk.

Short communication: Automatic detection of social competition using an electronic feeding system

The objective of this study was to determine if data derived from a system that electronically monitors feeding behavior could be used to identify competitive interactions of dairy cows at the feed bunk. A short interval between successive feeding events of 2 cows at 1 feed bin was predicted to be associated with a competitive replacement: when one cow displaced a feeding cow and then took her position at the bin. To identify the interval between feeding events that best predicted these replacement events, the feeding activity of 5 Holstein dairy cows was monitored using an electronic feeding system and video recordings. The number of times a cow was replaced at the feed bunk over 3 consecutive 24-h periods was determined using video analysis and these events were paired with the corresponding feeding events recorded by an electronic feeding system (Roughage Intake Control system; Insentec B.V., Marknesse, the Netherlands). A pooled analysis of all 5 cows showed that the optimal interval for predicting replacements at the feed bunk was 26s (sensitivity=86% and specificity=82%); this interval was termed the replacement criterion. This criterion was then applied to feeding data from a sample of 24 independent Holstein dairy cows, each observed for 3d during the week following calving. Video had previously been used to measure the number of times each cow was an actor and reactor of a displacement (when one cow displaced a feeding cow but did not necessarily take her position at the bin). Despite the differences in measures, the number of replacements (as estimated by our algorithm) was positively correlated with the number of displacements [as measured using video; correlation coefficient (r)=0.63 as actor, r=0.69 as reactor]. Estimates of an index of success in competitive interactions (number of times actor/number of times actor = number of times reactor) generated using the 2 methods were highly correlated (r=0.94). These results suggest that competitive behavior at the feed bunk can be automatically quantified using data derived from an electronic feeding system.

Sampling behavior of dairy cattle: Effects of variation in dietary energy density on behavior at the feed bunk

Factors affecting sampling behavior of cattle are poorly understood. The objectives of this study were to measure the effects of variation in feed quality on the feeding behavior of Holstein dairy heifers. Thirty-two heifers were housed in 4 groups of 8. Each group pen had 8 distinct feeding stations. The total mixed ration (TMR) provided was low energy (TMR-L), moderate energy (TMR-M), or high energy (TMR-H). During trial 1 (d 1 to 8), heifers were offered a uniform baseline diet (TMR-M in all 8 feeding stations) interspaced with 2 uniform test diets on d 3 and 6 (TMR-L or TMR-H in all 8 feeding stations). During trial 2 (d 9 to 17) heifers were offered a nonuniform baseline diet (7 feeding stations with TMR-L and 1 feeding station with TMR-H) interspaced with 3 uniform test diets on d 11, 14, and 17 (TMR-L, TMR-M, or TMR-H in all 8 feeding stations). Heifers were observed in pairs (n=16) for 15 min following delivery of fresh feed. Relative to the uniform baseline period of trial 1, 31% fewer switches occurred between feeding stations when offered TMR-H and 51% more switches when offered TMR-L. Relative to the nonuniform baseline of trial 2, 49% fewer, 27% fewer, and 25% more switches occurred during the TMR-H, TMR-M, and TMR-L treatments, respectively. In general, when heifers were offered a diet that was lower in energy density than that previously experienced, they spent less time at each feeding station and when offered a higher energy diet, heifers spent more time at each feeding station. The greater the contrast in energy density between the test and baseline diets, the greater the change in the behavioral response. Competitive interactions at the feed bunk were most frequent when TMR quality varied among the 8 feeding stations; during the nonuniform baseline period of trial 2, the number of competitive interactions was over 3.5 times higher than during all uniform dietary treatments. In summary, dairy heifers sample feed quality by changing feeding locations at the feed bunk and this sampling behavior is affected by variation in diet quality along the feed bunk and across days.