|
|
| The term "COBRA" is an acronym for Cohort-Based Record Analysis developed by Drs. Mark Kinsel and Marguerita Cattell. |
|
|
 |
 |
|
| COBRA: A New Type of Record Analysis |
by Mark L. Kinsel, DVM, PhD
As dairy operations have become more and more sophisticated, so has the information systems required to manage them. In the last 25 years, dairy records systems have evolved into critical components of the dairy's ability to make wise management decisions. During this same period, research has focused on developing new analytical tools to identify suboptimal performance. One recent advance in record analysis is the concept of "COBRA".
WHAT IS COBRA?
The term "COBRA" is an acronym for Cohort-Based Record Analysis developed by Drs. Mark Kinsel and Marguerita Cattell. The term "cohort" comes from the Roman word for “group” and a cohort of animals is defined as a group of animals that share a common trait or characteristic. For example, we could define all the cows that calved in May as a cohort, or all the first lactation heifers in a herd as a cohort, or all the cows that were treated for mastitis over the last 30 days as a cohort. Thus, COBRA is the analysis of records based on evaluation of the performance of specific cohorts.
HOW IS COBRA BETTER THAN TRADITIONAL ANALYSIS?
There are several advantages of COBRA over conventional analysis techniques. First, since we already manage cows in groups (cohorts), it makes sense that we should analyze their performance based on similar groups. For example, it would be logical to run a nutritional analysis on cohorts defined by string (pen) as all cows in a given string (pen) share the same feed. Second, COBRA allows us to make comparisons across groups since we are now tracking the performance of subgroups within the herd rather than the entire herd,. This is useful when we want to evaluate the effects of a management change. Using our nutrition example, we could compare the effect of using one ration fed a group of cows with a different ration fed to a second group. Third, for cohorts defined by stage of lactation (time since calving), COBRA provides the ability to follow changes in performance over time (trends). This has significant advantages over many traditional measures of dairy performance, especially reproductive measures. It has been difficult to rapidly detect whether reproductive performance was getting worse, staying the same, or getting better. With COBRA, for each new time period (i.e. weekly cohort, monthly cohort), we can see if we're winning, losing, or staying the same compared to past performance. Finally, COBRA is the basis of a concept called “statistical process control” that provides more confidence that observed change is “real” or normal variation.
EXAMPLES OF COBRA
The priciples and concepts used in COBRA are not new. For a long time, nutritionists have used cohort analysis of the performance of a string or pen in evaluating nutrition programs; however, only recently have COBRA concepts been applied to disease monitoring and reproductive performance. Figure 1 shows COBRA results from a research trial conducted in Idaho to evaluate a new breeding program where all cows were bred within 21 days of the voluntary waiting period on a 2000 cow dairy. |
| |
 |
| |
In this graph, the Y (vertical) axis represents the percentage of cows that became pregnant within 21 days of the voluntary waiting period for each week of calvings (indicated on the X-axis). The numbers on the X-axis (i.e. 9909, 9910) refer to the week of calving. The first two digits indicated the year and the second two digits indicating the week of the year. For example, week 9909 represents the group of cows that calved during the 9th week of 1999 ( between March 2, 1999 and March 8, 1999). Therefore, the first bar on the graph indicates that for the cows that calved during Week 9909, 45% became pregnant within 21 days of the voluntary waiting period. Week 9910 would represent the group of cows that calved during the 10th week of 1999 with about 30% of these cows become pregnant in the first cycle. The dark horizontal line on this graph represents the average first cycle pregnancy rate for all weeks (9909 through 9926). Notice the two cohorts (9912 and 9916) indicated by the black arrows. We can quickly see that the performance in these cohorts of cows was worse than other cohorts. When questioning the dairy manager about possible reasons for this, he indicated that the study protocol was not followed during that week and the AI timing was incorrect. The cows in cohort 9916 were bred with a new brand of semen that the dairy was trying. The advantage of COBRA is that performance can be linked to specific events on the dairy, facilitating rapid evaluation of specific herd management practices. For example, if a new management practice were implemented with the cows in cohort 9927 (the next cohort to show on the graph), immediate evaluation of the value of the practice is possible when the result of their pregnancy check was added.
When using cohorts based on dates (i.e. calving date, conception date, dry off date), an appropriate time period must be selected. For example, if we were considering basing our cohorts on calving date as shown above, we would need to decide if each cohort is a week of calvings, a month of calvings, etc. The decision boils down to finding a balance between selecting a short enough time period to detect changes as soon as possible and having enough animals in the cohort to detect a true change. Cohort size can be estimated as the number of animals expected to enter the cohort in a year divided by the number of time periods in a year. For a 2000 cow dairy, we would expect 2000 cows to calve annually. Using weekly cohorts, we would expect ~40 animals (2000 cows / 52 weeks) per cohort while using monthly cohorts, we would expect ~150 animals per cohort. As a general rule, cohort sizes less than 20 cows should be avoided. For dairy herds less then 200 cows, special techniques must be used to make COBRA appropriate.
COBRA analysis is not limited to just evaluating reproductive performance. Figure 2 shows the percentage of cows that developed a fever (body temperature > 103o F) during the first 10 days following calving for the same Idaho research trial. |
| |
 |
| |
In this example, the arrow marks the point where the dairy changed herd manager. Even though we can’t prove cause and effect, there is good evidence that the percentage of cows that developed a fever in the first 10 days after calving dropped after the arrival of the new herd manager.
An important question that arises from the COBRA results in Figure 2 is how can we tell if the percentage of cows with fevers truly dropped? This has always been the challenge in evaluating dairy performance as there is always some “normal” variation and it has been very difficult to distinguish between normal variation and the start of a true change. An extension of the COBRA concept that can help us make this decision is a technique called statistical process control.
Statistical process control (SPC) is a family of analytical techniques developed for the manufacturing industry in the 1920’s. Although SPC has been used for over 75 years in the manufacturing industry, it is only recently being adapted for use in agriculture. The foundation of SPC is that variation can be divided into two sources: background (biologic) variation and variation associated with changes in a process. Figure 3 below shows the same data as was shown in Figure 2 except that it has been evaluated using statistical process control. Without describing all the technical details of SPC, the important concept to understand is that this technique uses calculations regarding past performance to determine if subsequent performance is outside of expected limits. For the example shown, the performance of the first herd manager was set as the “baseline” performance and SPC tested whether the second herd manager was truly different. The arrow marks where the second herd manager took over and data points with a number are significantly different than the performance prior to the arrow.
|
| |
 |
| |
| WILL MY RECORD SYSTEM DO COBRA? |
| Do current commercial record systems have the ability to do COBRA? This is not an easy question to answer. No current record system has all the features of COBRA built in, but most have some ability to perform COBRA through custom (user-defined) reports. Several record software companies are developing full COBRA capability and Pfizer is including COBRA as part of their 100 Day Contract reproductive program. |
|
|
|
Email to a friend
Resource Summary
