CHAPTER 61 Reproductive Health Programs for Dairy Herds: Analysis of Records for Assessment of Reproductive Performance

JOHN. FETROW, STEVEN. STEWART, STEVE. EICKER, PAUL. RAPNICKI

RECORDS AND MONITORING

The effective use of records is a cornerstone of modern dairy production medicine. Records provide access to the performance results of a dairy’s management and serve as a major source of diagnostic information when problems arise. As veterinary service to dairy farms has matured, practitioners have become more involved in herd-level analysis, management consulting, and problem solving. Monitoring is an essential component of any system that must respond to external influences (Fig. 61-1).¹^–³ A parameter of the system is measured and compared with standards, goals, or past performance. If the parameter does not meet the goal, then plans are made and actions are taken (usually including collection of more diagnostic information). Because of both the action taken and the external influences on the system, a result is achieved. The result becomes the new status and the cycle begins again. Even though this activity is routine in most veterinary reproductive programs, in many cases it is not as fully developed or deliberately documented as might be most useful to the client.

Fig. 61-1 Role of monitoring in farm management feedback cycle.

(Adapted from Radostits O, Leslie K, Fetrow J: Herd health: food animal production medicine, 2nd ed. Philadelphia: WB Saunders, 1994.)

Although the general scheme described is consistent across all types of monitoring, it is useful to distinguish between the following four general classes of monitoring.

1. Surveillance. Surveillance is a monitoring system designed to generate action at the first detection of a parameter. Brucella testing for reproductive disease control and public health is an example. The monitoring system (milk ring test or herd serology) is set such that a predetermined course of action will be taken if any indication of brucellosis is detected. These sort of monitoring systems typically play a minor role in day-to-day dairy farm management.

2. Status monitoring. This form of monitoring involves the measurement of a parameter and the comparison of its absolute value to a goal figure. Services per conception might be set at 3.0, with the intent that performance worse than that number will initiate further diagnostic effort or management changes. In outbreak investigation, this is the typical starting point for the veterinarian. Status monitoring simply asks, “Are things as they should be?”

Although attempts have been made to arrive at a single “index” that will assess the performance of a dairy’s reproductive program, none are adequate alone.² Evaluating reproduction involves a complex set of issues that are to some degree unique to the particular dairy. Attempts have been made to develop expert computerized systems to evaluate reproduction,⁴ but none are in widespread use at the current time.

3. Trend monitoring. Typically more robust than status monitoring, tracking the trend in a parameter over time provides an added dimension of understanding to the analysis. Services per conception of 3.0 may be a problem in a herd that had been at 2.5, whereas it may be cause for celebration in a hot climate herd whose historical average has been nearer to 3.5. Trends allow for graphic display of the data and improve both client understanding and motivation.

4. Exception monitoring. Here, the trend of the herd as a whole is not the emphasis; instead attention is focused on displaying the individuals whose performance is substandard. In a sense, this is status monitoring within the herd, applied to individuals. Generally, this sort of detailed monitoring can be accomplished conveniently only with computerized record systems. Exception monitoring results are generally in the form of counts, graphic distributions, or action lists.

Viewed from another perspective, monitoring can be applied to two classes of information:

1. Outcomes: Traditionally, monitoring focused on measuring and evaluating the results of some system in operation on the dairy. Most of the measures used to monitor reproduction on dairies still fall into this class (e.g., days open, conceptions, pregnancies). These outcomes can be seen as the “objectives” of the system in operation, that is, its output. If the system is inadequate (in design or in operation or because of unanticipated external influences), the objectives are not reached.

2. Processes: Monitoring can also be applied to systems in operation to discover whether the prescribed activities of the system are being done and being done properly. To be effective, this presupposes that a specific operating system is defined and the person(s) responsible are aware and properly trained. For example, in a dairy reproduction program based on estrus detection, are personnel actually taking the time to observe the breeding population of cows? Do they know what to watch for? Do they report observations to the right person in a timely manner? Are they accurate in their observations? Are enough cows bred within a specified period? The distinction between these two classes of monitoring sometimes is indistinct (e.g., are enough cows bred?), but the conceptual difference is valuable. Real beneficial changes on the dairy happen at the process level, and problems can often be most quickly detected as undesirable changes in processes. Economic impact largely happens at the outcome level. Process monitoring (oversight) is a major part of a manager’s responsibilities and is key to assuring that the right things are done in the right way. Veterinary involvement in reproductive programs may include such process monitoring, either as a trouble-shooting activity or as part of the dairy’s operating management. To be effective in the latter role, the veterinarian must be on the dairy on a consistent basis and sufficient time has to be committed to the task. Typically, such process monitoring by the veterinarian includes a training component.

As dairies become more rationalized and as science and process development advances, the connection between processes and outcomes becomes tighter and more predictable. As that development continues, more and more time and attention will be committed to process monitoring, with less focus on outcomes because they will follow more reliably (but not necessarily perfectly) from proper processes.

PITFALLS OF MONITORING PARAMETERS

The practitioner should be aware of several pitfalls in status and trend monitoring that can lead to inappropriate inaction (real problems escape notice) or inappropriate action (situations are misdiagnosed as problems and the wrong action is taken). Of the two, inappropriate inaction is probably the more common in reproductive production medicine and the more costly. The major pitfalls are as follows.

1. Variation in methods of calculation: There are many sources of reproductive indices and each has arrived at its own (often undocumented) approaches to calculating reproductive indices.² Recommendations regarding calculation have been put forth, but remain inconsistently implemented.⁵ Furthermore, new approaches have become more appropriate since the last time a committee met to consider standardization. Care is necessary in interpreting numbers as presented, particularly when comparing results between various recording and summarizing programs.^2,^5,⁶

2. The dangers of averages. Reproductive herd records, and therefore reproductive herd monitoring, are rife with averages (means). Average days open, average services per conception, average annual culling rate as a percent of the herd (which is neither an average nor a rate)—these and others are used daily by veterinarians as they try to track the reproductive performance of their client herds. Averages measure one type of central tendency of a distribution of individual observations. Alone, they do not describe the spread of the distribution, nor do they call attention to failures of opportunities. Average days open, for example, can be the same in different herds, but with very different distributions. One herd can have a well-managed, tightly clustered distribution around the mean, while in another herd some cows might become pregnant very early while others are extremely delayed. Both herds might have the same average number of days open.

Herd size can have a dramatic effect on the variation and computation of averages. In a 50-cow herd with 25 confirmed pregnant animals, a single cow with 350 days open increases the average days open of the currently pregnant cows by 10 days. If this cow is then sold, the average of the remaining 24 will drop by 10 days (example assumes average days open of 100 days). If the dairy farmer is unaware of this, false credit for a positive result may be given to an irrelevant intervention. Conversely, two animals conceiving at 37 days would drop the average of the 27 pregnant animals by 5 days. At the other extreme of herd size, in very large herds with many cows contributing to the parameter, the average in any time period will tend to regress toward the long-term mean, obscuring problem cows that as individuals can be quite costly. Herd size effects cannot be avoided when averages are calculated; the practitioner must be aware of the possible pitfalls and use judgment when analyzing reproductive records.

3. Percentages. Percentages, like averages, can be misleading. In small herds or with small subsets of large herds (e.g., first-lactation animals bred by Joe Smith in June), one must be very careful not to overinterpret deviations from proportions used as standards. As an example, the 95% confidence interval (i.e., the true mean likely falls within this range) for services per conception when 13 pregnancies result from 40 breedings (a 33% conception rate) extends from 17% to 48%. One end of that range is a conception rate disaster, the other would be considered outstanding on modern commercial dairies. For many smaller Midwestern dairies, 40 breedings would take 6 months to accumulate, making tracking conception rates a problematic process. The practitioner must use judgment about when to intervene. Dairy managers are rarely willing to wait until something is statistically proved before they intervene. The burden of “proof” necessary to motivate management action can be quite different than the proof needed to establish scientific “truth.” Not doing something when it is needed can be just as costly (or more so) than doing something unnecessary.

4. Momentum. Momentum in a parameter occurs whenever the events from the distant past are included in the current calculation of a parameter. For the parameter to change appreciably, given the weight added by events that are long past, then either the current status must change radically or significant time must pass. Either way, momentum tends to dampen change in a parameter, obscuring the actual status and recent trend. Both average days open and calving interval include a great deal of momentum.

5. Lag. Lag or delay occurs when the outcome of an event or intervention cannot be measured for some extended period of time. Lag is inherent, for example, in the calculation of conception rates, because conception (pregnancy) cannot be confirmed until at least several weeks after breeding. Lag becomes more severe when an inappropriate parameter is chosen to monitor a management program’s results. An example might be the use of age at first calving to monitor the results of a prostaglandin synchronization program in heifers. The lag period is nearly a year before the effect of the program can be measured. A parameter such as the number (and identification) of heifers older than 13 months and not bred would have far less lag and thus would be more valuable as a monitoring tool for this prostaglandin program.

6. Bias. Bias can be introduced in many ways into the calculation of parameters that monitor status and trends on dairy farms. Bias exists when a systematic error is made in the selection of animals used in the calculation, the information used is incomplete or inaccurate, or the assumptions made about the biology are wrong. Bias distorts the parameter’s true representation of reality. Some of the causes of such bias in reproductive parameters are as follows.

• Effect of culls and “do not breed” cows. In many record systems, animals that have left the herd are not included in the calculation of some reproductive parameters. For example, the average days in milk at first breeding for the past 6 months should logically include any animal that was inseminated for the first time in the past 6 months, whether she is currently in the herd or is no longer present. Similarly, cows designated as “do not breed” cows may also be excluded. These animals are commonly those with the worst reproductive performance. Their exclusion from the data set would make the reported parameter better than reality in the herd.

• Presumption of reproductive outcome. Reproductive calculations often may make optimistic assumptions about the outcome or status of some animals. These compromise-approaches to calculating parameters are legitimately necessary in some cases, but the practitioner needs to understand that the parameter may represent the best-case scenario. As an example, some record systems may assume that all cows that are bred and not checked for pregnancy are pregnant for the purpose of calculating average days open.

• Exclusion of subpopulations. Some measures report on the performance of individuals with a positive (or otherwise known outcome), but ignore (or do not reflect) the current numbers of animals either pending status confirmation or past a management cutoff with no action. As an example, average days open in pregnant cows excludes cows from the calculation that are not yet confirmed pregnant. These may be the most important cows from a reproductive program perspective.

• Synchronization programs. The routine use of prostaglandin or other hormone interventions for estrus induction, although often cost-effective management strategies, wreaks havoc with the calculation of many reproductive parameters. Apparent estrus detection rates are often calculated based on the assumption of 21-day estrous cycles. Inducing estrous cycles obviously makes this assumption false. Similarly, calculation of the interestrous interval and the expected pattern of estrous cyclicity are severely confounded in herds in which estrus synchronization is used.

• Use of bulls. The calculated reproductive parameters can be severely distorted by the use of bull breeding, especially in herds in which the use of bulls is poorly recorded. For example, pregnancy to the bull may be calculated as resulting from an artificial insemination (AI) breeding, influencing the apparent services per conception rates.

Although the preceding list of pitfalls may seem long, it should not deter practitioners from working with records to analyze a herd’s reproductive performance. Instead, knowing these pitfalls should lead to a healthy skepticism about the number presented, a determination to increase accuracy and completeness of the underlying data, and a fuller understanding of the conditions under which parameters may not truly represent herd status.

Inadequacies of the Calving Interval

The classic parameter for monitoring the status of a reproductive program has been calving interval. Although a short calving interval might be the goal of a reproductive program, it is totally inadequate as a monitoring parameter. Calving interval has severe momentum. It requires two consecutive calving dates, so on a herd calculation basis, events from as much as 2 years before still enter into the current calculation. Similarly, it suffers from severe lag, because the outcome from reproductive management efforts must wait through at least one entire gestation before the results can be calculated. Calving interval introduces bias by excluding populations (culled cows, first-lactation animals, animals pregnant but not yet calved, “do not breed” cows). Last, by being an average, calving interval has all of the computational pitfalls described for averages. Veterinarians, as a profession, must stop using calving interval as a monitoring parameter.

Terminology Issues

There are many terms used in describing reproductive programs, terms that have survived and become part of the everyday lexicon of dairying. Unfortunately, some of these terms are not consistent with standard meanings of words. Although it is likely that many terms will go on being used even though improper, it would serve the industry and profession if everyone could be a bit more precise in how words are used. Specifically, rate is a term often misapplied to reproductive parameters. A rate is formally a measure of an event or statistic per time period. Thus, miles per hour or deaths per 1000 cows per year are rates. Unfortunately, many parameters associated with reproduction are discussed as rates but in fact are not rates. For example, conception rate is not a rate (no dimension of time). Conception rate (pregnancies per insemination) is a risk (proportion of a population with some particular characteristic ⁷). Older terms with widely accepted meanings are probably best left as is. New terms should be more carefully chosen.

GENERAL GOALS OF A REPRODUCTIVE PROGRAM

It has long been known that there is an important economic advantage to be gained by efficient reproduction in dairy herds.⁸^–¹² The economic effects of a sound reproductive program include increased milk by returning cows sooner to the earlier, more profitable phase of their lactation, increased numbers of replacement heifers and bull calves born, reduced costs of reproductive disease and reduced costs from culling, reduced nonproductive days due to extended dry periods, and increased rate of genetic gain.

On a biologic basis, the goal of a reproductive health program on a commercial dairy can be summarized as follows: Throughout her herd life, a cow should calve without difficulty and deliver a live calf, experience little or no postpartum reproductive disease, begin to cycle soon after calving, be inseminated soon after the voluntary waiting period, conceive to a high genetics bull within an optimal time period (or conceive at the right age as a heifer), and carry each fetus to term. This is not to imply that the goal is elimination of all pathologic events; to do so would be biologically impossible and economically inefficient. Rather, the goal is to have a minimum of pathologic events and a maximum of productivity within the constraints of practical biologic and economic reality. This general goal can be subdivided into sections, as follows.

1. Prompt rebreeding (appropriate voluntary wait period). After calving, uterine involution should occur promptly and cows should be reproductively sound by 45 days in milk (DIM). Once past the farm’s voluntary wait period (VWP; usually 45–60 days), cows should be seen or induced in estrus, be inseminated, and conceive in an efficient manner. If both estrus detection and conception rates are adequate, then only a small percentage of cows should have extended days open.

2. Genetic return. Genetic return on the investment in semen or bulls should be optimized. Some computer packages can calculate optimal bull profiles for selection for artificial insemination, given the farmer’s goals for genetic gain and variability.¹³ It is worth noting that the genetic return on the investment in semen occurs only if the insemination results in a live female calf that subsequently conceives, calves, and has a productive lactation. Genetic improvement on a dairy farm is a long-term and somewhat risky investment. Assuming a 40% conception rate, 44% of females not born co-twin to a bull, a 10% abortion rate after pregnancy diagnosis, and a 20% loss of replacements from parturition until the end of a productive first lactation (includes stillbirths), only 13% of inseminations actually return any appreciable value in genetic gain to the producer. This means it takes at least 7 straws of semen just to produce a first lactation animal. On many farms, they need to purchase well over 10 straws to produce a complete first lactation.

3. Pregnancy wastage. Pregnancy wastage should be at a practical, economic minimum (early embryonic death, abortion, stillbirths). Recent adoption of pregnancy diagnostic tests that can detect conception at an earlier stage is changing how pregnancy wastage is defined and the proportion of lost “conceptions” that are expected.

4. Disease. Peripartum disease morbidity should be minimized, enhancing animal welfare, avoiding impacts on later reproductive performance, lessening loss of cows either by death or culling, and minimizing therapy and labor costs.

MONITORING STATUS AND HERD REPRODUCTIVE TRENDS

Many parameters are used to monitor reproductive status and trends on the dairy farm. Some of these goals are shown in Table 61-1. For the most part, these are the traditional monitoring parameters for dairy reproduction. These herd goal levels must be applied with caution and may not be the appropriate alerting levels for management intervention on an individual cow basis. Several reviews of these parameters and their application have been written, so what follows is only a brief outline of the major parameters.^5,¹⁴^–²⁰

Table 61-1 Breakdown of Cows in a Dairy Herd by Reproductive Status (RPRO)