The unique strength of herd investigation is that because the individuals are grouped into a herd, you can compare affected animals, both clinical and subclinical, with unaffected animals, in both a cross-section (at one point in time) and over time to determine the differences between both the animals themselves and the factors affecting them. Basically, herd problem investigation is a thorough search for and interpretation of clues with the goal of arriving at a solution quickly and efficiently.  Note that the flow of this search is often "multithreaded", meaning that a number of these are occurring at once, and "recursive", meaning that new information causes a previous step to be revisited. Beware that common sense is often misleading. Following in approximate order are the typical components of investigations (modified from Lessard and Perry, 1988). See the outbreak investigation flow diagram below.

1. A Problem is Detected

   A farm event triggers the investigation. The event can be dramatic, such as a series of unexpected deaths or a sudden, dramatic production declines, or it can be finally the recognition of a chronic problem that has been occurring for some time. The earlier problems are detected, generally the more successful the interventions. Much of herd production medicine is geared toward establishing systems for the early detection if not the outright prevention of common problems. He who detects the problem is often called on to solve it (Fetrow).

   What is the primary complaint? Be careful of situations where the manager has reasoned from a primary complaint (e.g. low milk production) to what they suspect is the cause (e.g., pneumonia in adult cows) to an almost ubiquitous infectious agent (e.g., Mannheimia (Pasteurella) hemolytica) or to a key determinant known to be associated with this cause (e.g. poor barn ventilation). In this case, the actual cause was a mis-calibrated scale on a grain auger (Hancock).

   The three types of problems are

   1. Acute: The problem was precipitated by a temporally-associated management or husbandry error of sufficient magnitude to be a sole cause of the problem.
   2. Additive or Cyclic: The problem was precipitated by a combination of management or husbandry errors over time and the effects of cyclical factors such as season or production cycle stages such that the combination was sufficient to precipitate the problem. Ex: the summer coliform mastitis outbreak that is associated with the previous winter change to sawdust bedding.
   3. Chronic: The problem was precipitated by the long action of management or husbandry errors that required the passage of time for before the consequences became of sufficient magnitude to be recognized, such as the slow spread of a contagious mastitis agent or of Mycobacterium paratuberculosis. Ex: the recognition of a Staph. aureus mastitis problem associated with the adoption of a less efficacious teat dipping procedure more than a year previously.

2. Establish or Verify the Pathological and Etiological Diagnosis

   If the diagnosis is not definitive and if one is needed, start the process to obtain a definitive pathologic and etiologic diagnosis by collecting samples and submitting these to diagnostic laboratories. If dead or dying animals are involved, have complete field necropsies been done of a sufficient number of representative animals? Often only a few animals have been necropsied, often the necropsies are incomplete (e.g., a wide selection of tissues not submitted from all major organ systems) and often the necropsied animals aren't representative even when fairly large numbers of animals are dying. Be careful of accepting a provisional etiologic diagnosis as definitive without sufficient empirical evidence.  If you do, you are likely to overlook contrary clues and thus seriously mislead yourself, which can have serious consequences both for the producer and for your credibility. On the other hand, recognize that in many cases only establishing a definitive pathological and etiologic diagnosis does not solve the producer's problem. For example, many calf scour agents are ubiquitous and often sampling scouring calves will only confirm this fact. The real question is "Why does this farm have a problem with this agent when many others do not even though the infection is most likely present there as well?".

   Failure to perform complete necropsies and to submit a full set of tissue samples from the major organ systems is a common failure. Don't make the error of submitting only the samples that would confirm your leading differential diagnosis. In a continuing problem in a large dairyherd, the underlying problem was believed to be a severe respiratory condition but laboratory findings on the lung samples from partial necropsies were inconclusive. Complete necropsies subsequently performed on several sacrificed animals revealed a severe uterine condition subsequent to improper postparturient treatment.

3. Establish a Case Definition

Establish a case definition, precise as reasonable, to exclude those cases that are due to endemic background problems. For example, in a group of pregnant cattle approximately 10% of all pregnancies diagnosed prior to 45 days of gestation are lost and approximately 20% of these losses are observed. Including cases that are due to other problems will confuse the analysis and resolution of the problem. Even in general herd problems (e.g. low milk production), some individuals are affected more severely than others. You may wish to establish a case definition with different degrees of certainty (e.g., certainly affected, possibly affected, possibly unaffected, certainly unaffected). Remember the iceberg principle and the spectrum of disease. If you overlook these concepts, in your search for clues you could be comparing clinically affected animals with subclinically affected animals while you believe you are comparing affected to unaffected . This may lead to erroneous conclusions about the factors involved in the problem, which defeats the strength of herd investigation (comparison between groups of animals over time). On the other hand, more severely affected animals may have experienced higher levels of a common risk factor than lesser or unaffected animals.

4. Establish the Presence and Magnitude of the Problem with Objective Data

   Obtain objective data to document and verify the magnitude of the problem; do not rely only on the memories and perception of management and employees. Unless based on objective evidence (e.g., analysis of records), their perception of the problem may be correct but more often than not it is off the mark. Subjective perceptions of employees, managers and practitioners are valuable sources of hypotheses about risk factors; your task is to support or refute these and other hypotheses with objective evidence. Compare the actual number of cases to the expected number to determine whether or not the frequency is excessive. Be very careful of "dangling numerators", that is counting the number of cases without considering the number of animals actually at risk of becoming a case during that time period. A large increase or decrease in the number of animals susceptible to a condition causes a corresponding change in the number of cases of that condition even though the underlying risk remains constant. Because of seasonal effects, few herds maintain a constant number of animals passing through the period of susceptibility year around.

   In a large dairy herd, 1/3 of the retained heifer calves were dying due to salmonellosis. However, the producer did not recognize the magnitude of these losses because the calves were dying one by one and were removed by the rendering service during their regular visits. Only by comparing the current youngstock inventory on the farm with the calving events recorded on a calendar did the producer recognize the magnitude of this loss. In another large dairy, the manager knew that 10% of the cows calving during a two week period were clinically affected by a problem. However, when the records of all of the cows that had calved during this period were reviewed, much to his surprise he found that all had been culled from the herd.

   Be careful of what the producer accepts as "normal" or endemic occurrence. In one high producing herd, the producer believed that 3rd or higher parity Holsteins going down with milk fever was a "normal" occurrence. Thus, he accepted most of his older cows going down with milk fever and did not recognize that as an abnormal situation warranting correction. Neither his veterinarian nor his nutritionist were aware of the high incidence of milk fever on this farm (this was before IV calcium solutions were classified as legend drugs).

   Be careful of the events that the producer is omitting because of assuming that it is not related to the problem of concern. For example, the recent episode of late term abortions that isn't mentioned may well be related to the more prevalent metritis problem. Remember: More mistakes are made from not looking than from not knowing!

   1. Establish the Timing of the problem (the temporal pattern - When?)

      When in calendar time did the problem actually begin? What is the pattern of performance over time? If possible, do not rely on fallible human recollections alone; verify. Clearly establishing the timing of the onset of the problem from objective data (i.e., herd records) is crucial to determining what management changes or other sporadic events may be related to this event, such as a decline in milk production or an increased risk of death. Changes in individual animal performance must be distinguished from changes in total output due to changes in the numbers of producing animals. With count type data, plot epidemic curves. When did the index case occur?

      Be very careful of "pseudoepidemics" caused by the onset of producer awareness of a more chronic problem or caused by a change in problem definition. For example, changing from detection of fetal loss by visual evidence of a conceptus to detection of open cows post early pregnancy diagnosis will cause a pseudoepidemic of fetal loss in virtually any dairy herd. In one case, a dairy producer invested almost $1,000 in lab fees trying to establish the cause of such a pseudoepidemic. They had recently started using a personal computer dairy herd records program, which classified any cow returning to heat after a positive pregnancy exam as an abortion in addition to those with visible signs of late gestation fetal losses.

   2. Establish the Place of the problem (the spatial pattern - Where?)

      Where are the affected vs. unaffected animals located? Because different groups or pens of animals often have different levels of exposures (e.g., different amounts of feed ingredients, different water sources, different housing, different pasture, different origins, different stages of the production cycle) and a dose-response relationship exists for many etiologic agents, this is an important set of clues.
       

   3. Establish the Demographics of affected vs. non-affected animals (Who?)

      The strength of herd investigation rather than individual investigation is the opportunity to compare affected animals to uninfected animals. What are the characteristics of affected vs. unaffected animals in terms of exposure to potential risk factors, age, production level, stage of production cycle and source? As noted above, because of the spectrum of disease be very careful when classifying animals into affected and unaffected groups.
       

5. Assemble and Analyze the Data

   Using a spreadsheet, plots of production data smoothed over time are easy to create. From count data of the numbers of affected and the numbers of susceptible animals, calculate case morbidity and fatality rates by exposure and relative risks.  For an outbreak, establish an epidemic curve. For endemic problems, plot risk over time by cohort group. Examine the effects of other factors that vary over time (e.g., calving pen density, average of weekly high temperature, sources of animals) on risk of occurrence or production. For definitions of terms, see Terminology Specific to Epidemiology. If the herd doesn't have a good production accounting system, don't overlook clues in indirect sources of similar information. Concentrate on the data that will support or refute hypotheses.

   For example, the delivery dates and weights on feed invoices can provide approximate information on feed batch disappearance and thus approximate information on consumption patterns. On this basis, expected disappearance of feeds can be compared to actual disappearance. Invoices from rendering services may provide information on dates of animal deaths if they have not been recorded. Often, events such as calvings and breedings are written on calendars or in pocket books. 

   Gathering and analyzing this objective data on a herd problem is an examination process that is analogous to using laboratory tests or imaging procedures in the diagnosis in an individual animal. This objective data supports or refutes clinical impressions of the herd problem much like the testing or imaging supports or refutes clinical impressions of the clinical case.

6. Generate Hypotheses (Differential Diagnoses) about Key Determinants

   Key determinants are those risk factors causing the problem that can be modified on this premises. Based on your knowledge of the natural history of the given disease problem and the objective information that you have collected to this point, generate hypotheses about what key determinants might be involved. If necessary, search the literature for hypotheses about plausible risk factors. Subjective observations by the producer and other professionals are often valuable sources of hypotheses. What risk factor or factors is causing this problem on this premises that can be changed? These hypothesis are important because they provide the basis for further strategic sampling and data analysis rather than scatter-shot sampling and data overload. New hypotheses usually lead to revisiting prior steps in this process. Prioritize your hypotheses by their likelihood and focus your efforts on those with the highest priority until they are either more fully supported or are refuted.

   A serious, but not uncommon, error is to jump to generating hypotheses without first developing the quantitative information (the who, when, where counts) beyond vague clinical impressions (e.g., these animals seem to be affected more than those) to support or refute a specific hypothesis. This is analogous to "scattershot" ordering of laboratory tests in diagnosing individual animal cases, hoping something will pop up rather than using the laboratory tests to rule specific differentials in or out, and will likely be as unrewarding.

   Note that for additive, cyclic or chronic problems, the initial occurrence of the underlying management or husbandry deficiencies are usually not close in time to the recognition of the problem. In many cases, because of the lag between a management change and the recognition of the problem, the manager is reluctant to acknowledge that change precipitated the problem.

7. Test Your Hypotheses

   Establishing and executing good tests of your hypotheses requires a large amount of ingenuity. Based on the above hypotheses, predict what you should find in other animals, such as test results or production effects, and proceed accordingly to test your predictions. Make predictions of the form "if this cause is present, then this finding should be present". Because many causes have multiple effects, finding more of these multiple effects provides stronger support for the presence of the cause than finding only one. Often a single effect can result from several different causes. Finding what you predict supports your hypotheses; not finding what you predict weakens your hypotheses. The key is figuring out what predictions will provide good tests and are "doable". Perform any additional sampling and data gathering needed to confirm or refute your hypotheses. Avoid the error of only taking samples that if test positive will confirm your hypotheses but not taking samples that if test positive will refute it. Avoid scattershot sampling because doing so without an objective in mind is seldom useful and is expensive in money for the client and in credibility and time for you. What is the simplest set of explanations that covers the most findings?

   Example predictions are: "If this infectious agent is being transmitted between animals in this manner, then these other animals are at risk and some should be infected while these others are not and will not be." "If this risk factor (e.g. overcrowding in the fresh pen) is causing the metabolic disease (e.g., displaced abomasums), then I expect see the following pattern in the associated data (e.g., higher proportion of cows experiencing DA's in the cohorts with more crowding in the fresh pen compared to those with less crowding)."

8. Design Interventions and/or Prospective Studies

   Generate action items that are compatible with the specific facilities, economic limitations and management scheme of the premises. Considering the limitations of the facilities and the management and how these factors led to the occurrence of the problem in the first place in the design of these interventions is crucial to their success. If uncertain about the effectiveness of an intervention and the management situation is appropriate, propose a prospective clinical trial or other follow-up studies.
    

9. Report Your Findings

   See Student Guidelines for Written Recommendations. Remember the old truths that "The faintest pen is stronger than the strongest mind" and that "Success has many fathers but failure is an orphan." If your recommendations are successful, because of the passage of time you will not likely get the credit due you unless they were documented.
    

10. Monitor Results of Interventions

    For problems with a significant subclinical component, develop a monitoring scheme to provide early warning of the problem. If the herd doesn't have a good production accounting system to monitor changes in production but one is warranted, propose one. In productivity problems, establish benchmarks of performance if they aren't being used already. Monitoring of the results of your recommendations also provides you important information about the efficacy of your advice. The key is to figure out an economical, easily doable scheme for monitoring a herd. For example, for problems caused by an infectious agent, can pooled samples of effluent or milk be obtained on a regular basis to effectively monitor for the continuing presence of that agent in the herd?

    

[Return to Contents List]

page URL: http://www.vetmed.wsu.edu/courses-jmgay/OutBGuide.htm