Using Statistical Process Control to Investigate Reproductive Failure in Swine

CHAPTER 111 Using Statistical Process Control to Investigate Reproductive Failure in Swine



ROBERT B. MORRISON, GARY D. DIAL, PETER B. BAHNSON, WILL E. MARSH, JAMES E. COLLINS, DALE D. POLSON


You are called by a producer who is alarmed that the farrowing crates are half empty this month. What should he do? Another producer calls to report yet another abortion. Should he vaccinate? A third producer calls: She has noticed an increase in the number of returns to service. What should she do?


The diagnostician investigating a case of reproductive failure usually is presented with a primary clinical sign. There are relatively few manifestations of reproductive failure but many potential causes of each. Reproductive failure may be clinically manifested by one or more of the following signs:


• Failure to show estrus

• Return to estrus after being bred

• Abortion or premature parturition

• Failure to farrow despite having conceived

• Abnormal litter characteristics at birth as evidenced by low number of liveborn piglets or an elevated number of stillborn piglets, mummified piglets, or weakborn piglets

For example, one might be presented with a low farrowing rate; that is, a lower than expected percentage of sows that were apparently bred are farrowing. To investigate this case, the diagnostician must first know what the expected farrowing rate is for this herd. Assuming a “real” decrease has occurred, the diagnostician then must determine if the sows were actually bred and whether they conceived, and, assuming they conceived, why they did not farrow (Fig. 111-1). A relatively simple-appearing clinical situation quickly becomes complex. The case is further complicated if the producer lacks detailed reproductive records (arguably, without good records, the case is simple—it is unsolvable).


image

Fig. 111-1 Summary flow chart for investigating reproductive failure.


In investigating a case of reproductive failure, it may help to remember that the breeding herd exists to produce weaned pigs. The number of weaned pigs per unit of time is referred to as the herd’s output and is determined using the following formula:



image



For example:



image



The gilt pool and breeding, gestation, and farrowing facilities will have a certain capacity in terms of pigs weaned per unit of time, and the actual number achieved divided by the potential is referred to as the average utilization rate. This commonly is expressed as a percentage. As might be expected, unused capacity costs money (i.e., the opportunity cost attributed to pigs that could have been produced but were not), and the usual intention is therefore to maximize the utilization rate. The first step in maximizing utilization is to breed the appropriate number of females per unit of time. This having been accomplished, any real decrease in utilization rate may be caused by reproductive failure. The diagnostician’s job is to identify the source of the problem and implement solutions.


Returning to the farrowing rate example, suppose a herd’s farrowing rate has decreased from 85% to 70%. The diagnostician must now determine (1) whether the clinical manifestation represents a “real” change or random variation and (2) if the problem is “real,” what the cause of that problem is. The purpose of this chapter is to guide the diagnostician through these decisions.



APPROPRIATE DECISION MAKING: WHEN IS A SYSTEM OUT OF CONTROL?


Reproductive failure may present itself as an epidemic within a herd (sudden onset—an “outbreak”) or as a chronic, lower-than-expected level of performance. Diagnostically, the method of investigation and the likelihood of identifying the cause differ markedly between these two presentations.



Epidemic Reproductive Failure


In our experience, epidemics of reproductive failure are more likely to be caused by a single infectious agent such as PRRSV or porcine parvovirus. A sudden, but less dramatic, decline in reproductive performance may be associated with epidemics of infection due to pseudorabies virus, transmissible gastroenteritis virus, or swine influenza virus. These latter cases of mild but significant reproductive failure also may be caused by a variety of bacteria, including Leptospira spp. Diagnostically, submission of appropriate samples from a herd epidemic is likely to be very revealing if an infectious agent is involved.


Noninfectious causes of reproductive failure also can present as epidemics. This might include such factors as the use of infertile boars, a change in breeding techniques, poor feed quality, seasonal infertility, and others.


How does the diagnostician know when a herd is experiencing an epidemic of reproductive failure? An epidemic may be defined as the number of failures being “one more than expected.” Unfortunately, managers and consultants usually are not sure of what is expected. Although an intuitive impression may be correct, veterinarians need to be cautious as they evaluate herd performance and make recommendations. Effective decision making entails (1) knowing when a real change has taken place and not random variation, (2) evaluating the cost-benefit ratio of options, and (3) deciding when the problem should be reevaluated.


Two types of errors are possible in making such decisions. A problem might be diagnosed and a treatment or prevention instigated when the apparent problem was really random variation (type I error). It is this type of error that is responsible for many needless vaccination or medication programs. Our bias is to err on the conservative side and intervene when an intervention is not called for. A second type of error (type II) occurs when a situation is examined and intervention is not elected when, in reality, corrective measures were indicated. That is, the situation was “out of control” and action really was necessary. The first step is to chart the data. Columns of data are extremely unrevealing, and graphing techniques have been developed that help present and interpret the trends.


A run chart is a type of line graph on which time is on the x-axis and the variable of interest is on the y-axis (Fig. 111-2). Run charts are very easy to compose and allow quick interpretation of trends. However, it is difficult to differentiate between random fluctuations and early, mild trends in the data. Therefore, another type of chart called a control chart was developed.


image

Fig. 111-2 Run chart of percent of sows returning to service in a 1400-sow herd.


The control chart, like the run chart, plots the actual values over time but also considers the recent variation while estimating control limits (Fig. 111-3). Control limits are calculated by adding and subtracting three standard deviations to the average value. A critical step in using control charts is to start with a process that is “in control” to determine the expected standard deviation. Using these expected standard deviations, if the number of returns to service per unit of time is beyond the upper or lower control limit, one can be extremely confident that the process is “out of control.” At this point, the consultant must intervene and initiate a diagnostic investigation.


image

Fig. 111-3 Control chart of percent of sows returning to service in a 1400-sow herd.


Control charts can be further modified to create an “early warning system.” That is, the area between the average and the control limit can be subdivided into three zones based on standard deviations (see Fig. 111-3). Zones C, B, and A correspond to the average ±1, ±2, and ±3 standard deviations, respectively. If the measurement of interest is normally distributed around the average, then data points should appear approximately normally distributed within the limits. Also, there should be no evidence of trends or recurring cycles. Approximately 66% of the measurements should be in zone C. Approximately 95% of the measurements should be in zone B or C, and 99% of them should be in zone A, B, or C. Therefore, only 1 out of 100 times will a measurement be expected to be outside the control limits by chance alone. In other words, if a measurement occurs outside the control limits, one can be very confident that something has disrupted the process. The zones also can be used in conjunction with statistical probability to establish some guidelines that will detect a process out of control sooner than waiting for the control limit to be surpassed.1 The detection criteria are as follows:


• Presence of 2 out of the last 3 points in zone A

Related posts:

  1. Control and Synchronization of the Estrous Cycle and Ovulation
  2. Bacterial Diseases of the Fetus and Placenta Associated with Fetal Loss in the Mare
  3. Diseases of the Reproductive System of the Bull
  4. Surgical Correction of Abnormalities of Genital Organs of Cows

Stay updated, free articles. Join our Telegram channel
Tags:
Sep 3, 2016 | Posted by in SUGERY, ORTHOPEDICS & ANESTHESIA | Comments Off on Using Statistical Process Control to Investigate Reproductive Failure in Swine

Full access? Get Clinical Tree
Get Clinical Tree app for offline access