Full text of DOI:10.1079/AHRR200239

© CAB International 2002
ISSN 1466-2523
Animal Health Research Reviews 3(1); 43–55
DOI: 10.1079/AHRR200239
Empirical and theoretical evidence for herd
size as a risk factor for swine diseases
Ian A. Gardner1*, Preben Willeberg2,3 and Jan Mousing3†
1Department of Medicine and Epidemiology, School of Veterinary Medicine, University of
California, Davis, CA 95616, USA, 2Danish Veterinary and Food Administration, Mørkhøj
Bygade 19, DK-2760 Søborg and 3Danish Bacon and Meat Council, Axelborg, Axeltorv 3,
DK-1609 Copenhagen V, Denmark
Received 5 January 2002; Accepted 12 March 2002
Abstract
Herd size is frequently studied as a risk factor for swine diseases, yet the biological rationale
for a reported association with herd size (whether positive or negative) is rarely adequately
discussed in published epidemiological studies. Biologically plausible reasons for a positive
association between herd size and disease include a greater risk of introduction of pathogens
from outside the herd, greater risk of transmission of pathogens within and among herds
when the herd is large, and effects of management and environmental factors that are related
to herd size. However, compared with owners of small herds, owners of large herds might
more frequently adopt management and housing practices that mitigate this theoretically
increased risk. We used studies of pleuritis, pneumonia and pseudorabies to describe the
epidemiological issues involved in evaluations of the relationship between management fac-
tors, herd size and disease. In future studies, we recommend that (i) herd size be measured
in a way that best characterizes the true population at risk; (ii) studies that evaluate manage-
ment-related risk factors should account for herd size wherever possible; (iii)
population-based studies of the interrelationships among management factors and between
management factors, herd size, herd density and pig density be done; (iv) likely biological
reasons for any herd-size effect be postulated; and (v) the distribution of herd sizes in the
source population and the study sample be described.
biosecurity and vaccination practices, and the use of
Introduction
hired employees (United States Department of
Agriculture: Animal and Plant Health Inspection Service:
Veterinary Services, 1995). Although the demonstration
of an association between herd size and disease might
be considered interesting by epidemiologists and pro-
vide direction for future studies, such a finding has little
immediate relevance to swine producers. For example,
producers with large herds often realize economies of
scale and are therefore unlikely to reduce herd size vol-
untarily because such a change would usually result in
increased production costs and a lower gross margin per
pig sold. In addition, large herds have greater capital
investment in facilities that might also preclude them
from decreasing herd size voluntarily. In some countries,
however, animal welfare, environmental and political
considerations may ultimately result in limits on the
maximal size of swine herds. Also, public-health con-
cerns about foodborne disease associated with pig meat
will increase interest in the effects of herd size on
Management and environmental factors are considered
important determinants of the risk of the introduction and
maintenance of an infectious agent in swine herds and
also affect many measures of herd productivity. Among
factors influencing pig health, herd size is considered to
be an important correlate of disease and has been studied
since the early 1970s (Bäckström, 1973; Lindqvist, 1974;
Aalund et al., 1976). However, the biological basis for the
herd-size effect (whether positive or negative) is not fre-
quently known or specified in published studies.
In most epidemiological studies, some measure of
herd size can be determined and this is often used as a
surrogate for management variables that are herd size-
related, such as all-in, all-out and multisite production,
*Corresponding author
E-mail: iagardner@ucdavis.edu
†Present address: FOSS Electric A/S, Slangerupgade 69, DK-3400,
Hillerød, Denmark

44
Ian A. Gardner et al.
pathogen prevalence because larger herds contribute the
greatest number of slaughter pigs to the market.
individually or in addition to other management and
environmental factors. Our literature review was
restricted to manuscripts published in English up to and
including the year 2001. The search was done using the
keywords ‘herd size’, ‘herd size and disease risk’, ‘risk
factor’, ‘pig management’, ‘management factor’, ‘stocking
density’ and ‘herd density’. We read the papers to deter-
mine how herd size was measured and then categorized
studies according to the body system studied and evalu-
ated associations with larger herd size as being
‘positive’, ‘negative’ or ‘none’ (Table 1). These classifica-
tions were often subjective because in some papers
there were no statistical analyses or description of how
herd size was classified, and multivariable analyses were
not used consistently. Because of these limitations, most
of our evaluations were qualitative. We considered the
use of meta-analysis for a more formal comparison and
summary of the results into a single effect measure, but
the lack of comparability of studies precluded this
approach.
Our evaluation of the published papers indicated that
the relationship between herd size and disease varied
with the disease studied (Table 1); however, it is inter-
esting to us that for respiratory diseases, including
pseudorabies, the relationship between large herd size
and disease was typically positive. However, we are
unable to determine whether bias against publication of
negative results might have influenced this conclusion.
In contrast, we found evidence that large herd size was
‘protective’ for some diseases, such as toxoplasmosis
and liver milk spots. Presumably, this protective effect
was in part attributable to more frequent use of confine-
ment systems in large herds, which reduced exposure to
the parasites.
In the reviewed papers, herd size was measured in
many different ways, although we assumed that one of
the underlying goals of the authors was to use the chosen
herd size measure to characterize the true population at
risk for the infection or disease of interest. For diseases
of sows, herd size was usually given as the number of
breeding females (gilts and sows). It is important to note
that the number of females may be based on the current
inventory or the number of sows farrowed, and many
definitions of a sow are possible. Female herd size esti-
mates may be adjusted by excluding those not of
breeding age (though already marked as replacement
gilts) or cull sows that are present on a farm. For dis-
eases in finishing pigs, the choice of a herd size measure
tended to be more variable (and included the number of
pigs slaughtered per annum, the number of pigs in the
herd between weaning and market age, the total number
of pigs in the herd, and the number of pig places per
barn or building compartment). Another herd size meas-
ure used in Danish studies of possible airborne
transmission of pseudorabies virus (PRV) (Christensen et
al., 1990) and porcine respiratory coronavirus pseudo-
rabies virus (PRV) (Henningsen et al., 1988; Flori et al.,
Interest in studying herd size associations in epidemi-
ological studies has probably developed partly because
estimates of size are frequently available in existing
databases or can be obtained readily by questionnaire.
In many studies, however, an unanswered question is
whether the primary factor that produces the herd-size
effect is the size of the population at risk, management
and environmental factors that are related to herd size,
or some other unmeasured factor, such as herd and pig
density in an area.
We believe that one goal of future epidemiological
studies of swine disease is to better characterize the
nature of the effect, if any, of herd size, so that potential
adverse (beneficial) effects of increases in herd size can
be minimized (maximized). Accordingly, it is important to
differentiate real from spurious herd-size effects because a
change in herd size would have no benefit if the effect
was attributable to management or environmental factors
that were not modified. Also, determination of the reason
for any herd-size effect would allow more specific advice
to be given to farmers and decision-makers on appropri-
ate preventive measures.
In this paper, we review epidemiological studies of
herd size and swine diseases, describe measures of herd
size, present biological explanations for possible effects
of herd size on disease occurrence, and give examples
of the relationships between herd size and disease risk
based on our experience. Most of our examples are
studies of pleuritis, pneumonia and pseudorabies
because these are the diseases that have been studied
most frequently with regard to herd-size effects and posi-
tive associations have often been found. In the final
section, we make recommendations for future epidemio-
logical studies of swine diseases that involve the
evaluation of the effects of herd size and management
and environmental factors.
Many definitions of ‘herd’ are possible, reflecting
physical contact, physical proximity or shared inputs
(e.g. common feed source, genetics, and the source of
pigs in a production pyramid). Some definitions are
more appropriate to diseases transmitted by direct con-
tact (e.g. mange) than diseases transmitted by indirect
contact (e.g. Actinobacillus pleuropneumoniae). For this
paper, we broadly define a herd to be a cluster or aggre-
gate of pigs in a single building or multiple physically
separated buildings or outdoor lots located on a single
site or on multiple sites. Pigs in the same building or site
can be of different age and production classes. This is
similar to the definition used in a prior study by one of
us (Christensen and Gardner, 2000).
Published studies and measures of herd size
We reviewed published studies of swine diseases in
which herd size was evaluated as a risk factor either

Herd size as risk factor for pig diseases
45

46
Ian A. Gardner et al.
1995) was the number of ‘heat-producing units’ (HPU).
The HPU (1 HPU = 1000 watts at 20°C) is an estimate of
the heat loss (and hence the ventilation requirement) at
different temperatures for different age classes of pigs
(Strom, 1978): 1 sow and her litter = 0.4 HPU; 1 non-
lactating sow, boar or gilt = 0.2 HPU; 1 grower/finisher =
0.17 HPU. This measure accounts for differences in size
and weight of different age classes of pigs in farrow-to-
finish herds and facilitates the comparison of sizes of
breeding and finishing herds. The relative advantage of
this method for accounting for pigs of different types in a
single measure has not been evaluated, but such an
approach seems intuitively appealing for studies of respi-
ratory diseases transmitted by aerosols in intensively
housed pigs where risk is constant across all ages. An
assumption of age-independent risk might be realistic for
first epidemics of many viral and bacterial respiratory
agents before the risk starts to be influenced by passive
and active immunity. In contrast, where transmission
occurs primarily by nose-to-nose contact it would be
more appropriate to give greater weight to finishing pigs
in pens rather than breeding animals in crates. In the lat-
ter case, the use of HPU as a herd size measure would
be inappropriate.
A discussion of herd size measures would be incom-
plete without reference either to herd (and pig) density in
an area, which might also be positively correlated with
herd size, or to pig stocking density within herds. Greater
herd (and pig) density may be confounded with larger
herd size at least in some countries, such as Denmark
(Flori et al., 1995), and therefore might in part explain the
reported herd-size effects. The importance of considering
the density of infected herds or pig density on an area
basis in addition to herd size as risk factors for pseudo-
rabies was recognized in at least five studies (Austin et
al., 1993; Leontides et al., 1994a, 1995; Boelaert et al.,
1999; Maes et al., 2000). Indeed, in one study (Leontides
et al., 1994b), the effect of larger herd size was no longer
evident after related management variables and density of
pigs had been included in the statistical models.
Pen and air-space stocking density have been investi-
gated infrequently as risk factors for swine diseases, and
there a number of possible reasons for this. First, pig den-
sity is difficult and tedious to measure accurately, and the
value changes with the stage of production in farrow-to-
finish herds. Secondly, even if a good measure exists,
measurement is usually done at the pen level and herd-
level analysis is complicated by the need to correct for
clustering or over-dispersion. Thirdly, there is collinearity
between herd size and pig density. Of the 14 unique
reviewed studies of pleuritis or pneumonia lesions (Table
1; seven authors studied both lesion types), only seven
(50%) evaluated a stocking density measure of herd size,
yet standard recommendations for the control of pneumo-
nia in problem herds (Done, 1991) emphasize alteration
of up to 20 environmental variables, including density, air
space, and pigs per building. The importance of air-space
stocking density and floor-space stocking density as
important predictors of respiratory health in finisher pigs
has been reaffirmed in recent studies (Buddle et al., 1997;
Stark et al., 1998; Maes et al., 2001).
Another consideration relevant to the measurement of
floor and air-space density and measures of herd size war-
rants comment. The possible collinearity of these
measures in multivariable models may have dissuaded
researchers from recording more than one or two meas-
ures of herd size. In practice, the most easily acquired
measure of herd size (e.g. the number of breeding females
or the number of finishers slaughtered per year) is gener-
ally used and, for whatever measure is selected, authors
probably assume that the variables are sufficiently highly
correlated to be interchangeable. The latter assumption
may not necessarily be correct. For example, we estimated
pairwise correlation coefficients between three frequently
used measures of herd size: the number of breeding
females, the number of pigs sold per year and the total
inventory of pigs for 1306 US farrow-to-finish farms stud-
ied as part of the National Swine Survey (United States
Department of Agriculture: Animal and Plant Health
Inspection Service: Veterinary Services, 1992). The esti-
mates ranged from 0.20 to 0.61, which indicated only low
to moderate linear relationships. Hence, multiple measures
of herd size might be included in multivariable models.
Biological reasons for an effect of herd size on
disease occurrence
In this section, we describe plausible biological reasons
for increased risk of introduction and spread of infec-
tious agents with increasing herd size. We note that
large herds might be able to mitigate many of these
increased risks by the use of management procedures
(e.g. biosecurity practices, all-in, all-out production and
age-segregated production) which decrease the risk of
introduction of pathogens into herds or their within-herd
transmission. In addition, owners of smaller herds in the
USA frequently use riskier management practices,
including swapping boars with neighbors, increased
likelihood of attending fairs and livestock shows, and
more frequent use of trucks for a single purpose, which
counteract the benefits of small herd size.
Greater risk of introducing infectious agents from
outside the herd
The number of direct and indirect contacts that a swine
herd has with potential outside sources of infection may
increase as the herds expand in size. These contacts
include introduction of breeding and feeder pigs, trans-
portation of feed and of slaughter animals, and visitors.
Consequently, unless protective management practices
are used as outlined above, the risk of introduction of

Herd size as risk factor for pig diseases
47
pathogens will also usually increase. We describe the
interrelationships between management factors and
herd size in more detail below (see Influence of man-
agement-related factors that are associated with herd
size).
pigs (equation 1) could be applied to the airborne spread
of viral diseases with modification of the meanings of p
and n (Willeberg et al., 1994). In the context of airborne
spread, P remains the probability of a herd becoming
infected, p is the probability of an individual pig in the
herd becoming infected if exposed to droplet nuclei con-
taining virus of constant concentration (virus/m3) and n is
the herd size at the time of exposure to the virus
(Willeberg et al., 1994). One way to conceptualize the
meaning of the herd-size effect is to consider that larger
herds occupy a greater area of land. For example, assum-
ing a fixed density of pigs per m2 of floor space, a herd
of 500 fattening pigs would occupy an area five times
greater than a herd of 100 fattening pigs.
Introduction of infectious agents with carrier pigs
To demonstrate the effects of herd size, consider herds
that replace a fixed proportion of the female breeding
herd with gilts purchased from outside herds. Assuming a
binomial distribution, Marchevsky et al. (1989) showed
that the risk (P) of introducing at least 1 infected female is
P = 1 – (1 – p)n
(1)
where n is the number of gilts introduced and p is the
prevalence of the infectious agent in the source popula-
tion. For example, assuming that p = 0.01 and no testing
for the agent was done prior to the introduction, then a
herd of 500 sows replacing 20% of females per annum
would have a risk (P) of 0.64 of introducing an infec-
tious agent compared with a risk of 0.18 for a herd of
100 sows with the same replacement rate. A similar herd
size-related phenomenon occurs when herds are initially
established from the same source without mitigations.
One possible strategy for owners of larger herds to
mitigate this theoretically increased risk is to purchase
pigs from non-infected herds or to only introduce test-
negative pigs into their herds.
Thorburn et al. (1991) suggested a modification of
equation 1 to include sampling without replacement,
and the equation could be further modified to include
sampling of variable numbers of pigs from herds with
different prevalences of infection. For simplicity, we do
not present these modifications here. Also, to allow
direct comparison with the model in the following sec-
tion, we assume that, once an infectious agent is
introduced into a herd with an infected pig, it is trans-
mitted to other pigs in the same herd.
The probability of the herd becoming infected (P)
increases non-linearly with n and p, but when p is small
it can be shown that the relationship is approximately
linear, i.e. P = 1 – (1 – p)n ꢀ np (Fig. 1; e.g. when p ꢀ
0.001). For two herds of size n1 and n2 with constant
risk p of infection on a per-pig basis, the odds ratio (OR)
of becoming infected is the ratio of the two herd sizes,
i.e. OR = n1/n2 (OR = 5 for the two fattening herds
referred to above). In a logistic model of the ln(OR) of
infection, logarithmic transformation yields
ln(OR) = ln(n1) – ln(n2)
(2)
Field data are used in logistic regression analysis to
model ln(OR) as a function of the herd-size variable (X)
as follows:
ln(OR) = βX
= β(x1 – x2)
(3)
where x1 and x2 are two levels of the herd-size variable
(X), measured on a continuous scale, and β is the logis-
tic regression coefficient for a 1-unit change in herd size.
We note that equation 2 is a special case of equation 3
in which β = 1 and x1 = ln(n1) and x2 = ln(n2).
Evidence in support of this simple model is provided
by results of the following two analyses. Christensen et
al. (1990) analysed data from outbreaks of pseudorabies
in southern Denmark from 1987/88. Cases were
pseudorabies-positive herds for which no known contact
with other infected herds had occurred. Control herds
were non-infected herds from the same general geo-
graphical area as the cases. In the analysis, herd size was
expressed as HPU to reflect the relative metabolic sizes
of the pigs. Because the distribution of HPU was posi-
tively skewed, a logarithmic transformation of HPU was
used to model the effect of herd size. The final model
(which adjusted for whether the herd was a conventional
or SPF herd and whether the herd was a fattener or sow
herd) included a statistically significant effect of ln(HPU)
with β estimated as 1.26. In a subsequent paper co-
authored by one of us (Mousing et al., 1991), however,
the control sample was restricted to 11 postal districts for
which complete data were available for outbreak herds
in both the 1987/88 and the 1989/90 epidemics. Here
Introduction of infectious agents transmitted by
airborne routes
Even for herds not introducing pigs and with limited
contacts with other herds, there is empirical evidence
that large swine herds are at increased risk of airborne
introduction of viral diseases. For example, for infections
such as PRCV and PRV (for which airborne transmission
in Denmark has been reported), an increased risk of
infection was evident in large herds (Henningsen et al.,
1988; Christensen et al., 1990; Flori et al., 1995). In addi-
tion, the authors of these three studies found that a
natural logarithmic transformation of herd size was the
best formulation of the independent effect of herd size
on the odds of positive PRCV or PRV status.
On the basis of this empirical evidence and the
assumption that the herd-size effect was not confounded
by herd density, we proposed that the model specified
for the risk of disease introduction associated with carrier

48
Ian A. Gardner et al.
Fig. 1. Probability of airborne infection of swine herds (P = at least 1 pig becoming infected) as a function of the number of
pigs in the herd (n) and the probability of an individual pig becoming infected (p). Calculations are based on equation 1 in
the text: P = 1 – (1 – p)n.
also, only fattener/sow management was accounted for,
and the β estimate for ln(HPU) was 0.86. Leontides et al.
(1994b) studied herd risk factors for PRV infection of
breeding sows in northern Germany and estimated β for
ln(herd size) to be 0.89 after controlling for possible intro-
duction of infection with carrier pigs (purchase of
replacement gilts versus use of own inventory). In a study
of risk factors for herd-level PRCV seropositivity, Flori et
al. (1995) analysed data from a subset of 125 swine herds
with a seropositive neighboring herd and found a ln(HPU
of own herd) estimate of 0.97 for PRCV infection.
For these examples of probable among-herd airborne
transmission, we speculate that the agreement between
observed and expected data [i.e. β estimates for ln(HPU)
close to 1] may be explained by the fact that airborne
transmission of the virus closely follows assumptions
inherent in the model: equal, independent and low risk
of initial infection of individuals. In our opinion, if the
model were inappropriate or confounding or other
biases were present in the data, then one would expect
to find β estimates substantially different from 1.
and Longini (1994). Briefly, these factors are the infec-
tiousness of infected individuals, the quantity and
quality of the agent transferred by the different routes of
transmission, the number of contacts per unit time (for
the different routes), the susceptibility of non-infected
individuals, and the number of different individuals for
which there is contact. Analogous factors affect the
transmission of pathogens among herds: infectiousness
of infected herds, the susceptibility of non-infected
herds, the amount of viable pathogen that is transferred
during a contact, the rate at which contacts occur and
the number of herds that make contact (Stegeman et al.,
1999).
To help explain and predict disease patterns, many
models of disease transmission have been proposed. For
brevity we mention only one: the Reed–Frost model
(Fox et al., 1971; Frost, 1976; Yorke et al., 1979;
Anderson and May, 1982). The Reed–Frost model pre-
dicts that the effective contact rate and the number of
susceptible and infected pigs will affect the spread of an
agent in a herd:
The use of equation 1 for a probabilistic assessment
of virus airborne transmission has also been proposed
by Manuel-León and Casal (2001) to explain patterns
of herd infection with pseudorabies in 1981/82 in
Yorkshire, England.
Ct+T = St(1 – qCt)
(4)
where Ct is the number of cases at time t, St is the num-
ber of susceptible pigs at time t, and q is the probability
that a susceptible pig will not become infected during
the next time period of duration T. This equation indi-
cates that the occurrence of an epidemic is largely a
function of the number of susceptible pigs and the con-
tact rate. The number of susceptible pigs is usually
greater in large herds unless their risk of infection is
reduced by vaccination. Moreover, the probability of an
effective contact might also be greater because of hous-
Greater risk of transmission of infectious agents
within and among herds
Factors that affect the transmission of infectious agents
among individuals have been described by Koopman

Herd size as risk factor for pig diseases
49
ing in common areas or air spaces, and the increased
likelihood of direct and indirect contact with susceptible,
non-immune pigs. Direct contacts between pigs in the
same space often increases with increasing stocking
density. In the Reed–Frost model, changes in these fac-
tors will result in different patterns of incident cases
with time (Fig. 2) and in the total number of affected
pigs before a given age or at slaughter (as is evident by
comparing Fig. 2a and b).
Although q in equation 4 could be independent of
population or herd size, a herd-size effect implies that
herd size (denoted N) should be incorporated explicitly
in the model. Accordingly, De Jong (1995) has recom-
mended the following form with a herd size-dependent
transmission rate (SI/N, where S is the number of sus-
ceptible animals, I is the number of infectious animals
and N is the total number of animals; for a more
detailed discussion see De Jong, 1995):
pleuritis lesions is more limited (Christensen and
Mousing, 1992).
For PRV, evaluation of the effects of herd size on
within-herd transmission is easier because a single infec-
tious agent is involved and pigs remain chronically
infected (or at least serologically positive for life). Duffy et
al. (1991a) proposed that more replacement gilts are
introduced in large breeding herds than in small herds;
therefore, large herds will usually have more susceptible
females than small herds—provided that replacement gilts
are PRV-negative at introduction. A deterministic mathe-
matical model of PRV showed also that the most
important determinants of viral persistence were herd size
and the density at which sows were maintained. A thresh-
old herd size of 66 sows was identified, below which
virus would be eliminated from the herd even when no
specific control measures were implemented (Smith and
Grenfell, 1990). In contrast, Bouma et al. (1995), using a
susceptible–infectious–recovered (SIR) model, failed to
show that the PRV transmission in groups of 10 and 40
vaccinated pigs was dependent on population size.
Ct+T = St(1 – e–ꢁCtT/N
)
(5)
For pneumonia and pleuritis, direct comparison of
empirical data with the theoretical predictions of the
Reed–Frost model at an individual animal level is diffi-
cult because of complex interrelationships among
host, agent and environment factors, the involvement
of multiple etiological agents, and because prevalence
rather than incidence is the usual outcome measure in
studies of slaughter pigs. The prevalence of pneu-
monic and pleuritic lesions in slaughter pigs is
determined by the incidence rate of lesions during the
grower/finisher phase, the distribution of age at infec-
tion and the rate of lesion resolution. There is
evidence that many pneumonic lesions heal partly or
completely before slaughter; resolution of chronic
Influence of management-related factors that are
associated with herd size
In swine herds, management and environmental factors
are interrelated and often associated with herd size. Large
herds are different from small herds in many ways other
than merely the number of pigs or stocking density. Thus,
any herd-size effect (positive or negative) is often mixed
with the effects of other factors which may not be
recorded per se but which are associated with herd size.
These factors vary from herd to herd and from country to
(a)
(b)
600
60
50
40
30
20
10
0
500
400
300
200
100
0
0
1
2
3
4
5
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9
10
Time
Time
Fig. 2. Reed–Frost model estimates of the number of cases of an infectious disease for a contact rate = 0.02 and for initial
population sizes of 50 (a) and 500 susceptible pigs (b). The category ‘immune’ includes pigs at risk of being chronically
affected.

50
Ian A. Gardner et al.
country, but many factors, including confinement housing
(Anderson et al., 1990), all-in, all-out and multisite pro-
duction systems, automation and the use of hired
workers, probably occur more frequently in large herds,
while others, such as the purchase of pigs for finishing,
are more frequent in small herds in some countries
(Pointon et al., 1985). Specialized large herds often incor-
porate improved environmental features in new buildings,
including small compartments (Elbers, 1991), but there
are few recent population-based studies which have
described the relative frequency of practices in herds of
different size or in herds with buildings of different ages.
Disease-control practices such as vaccination and
medication policies are also likely to differ by herd size
(Svensmark et al., 1989a, b; Siegel and Weigel, 1999). In
the USA, for example, the use of vaccines to control
reproductive and neonatal diseases increases with
increasing number of breeding females in the herd
(United States Department of Agriculture; Animal and
Plant Inspection Service; Veterinary Services, 1995). If
large herds use vaccine more frequently than small
herds and vaccination is effective, then vaccination
might counteract any increased risk of transmission in
large herds. Although vaccination can alter both the sus-
ceptibility of non-infected pigs and the shedding of
pathogens by infected pigs, the assumption that vac-
cines for all diseases are effective is inappropriate. For
example, vaccines for agents such as Actinobacillus
pleuropneumoniae have often failed to be protective
(Hunneman, 1986).
housing factors, a higher initial prevalence of infection,
or to the use of tests with a sensitivity of less than 100%
(discussed in the following section). For example, Siegel
et al. (1993) studied factors associated with time under
quarantine for swine herds in the voluntary phase of
pseudorabies eradication in Illinois, USA. Larger herds
(>80 versus 30–80 sows) had an increased quarantine
time; however, large herd size, initial seroprevalence,
confinement housing and delay in initiating a herd
clean-up plan were positively interrelated. After inclu-
sion of all factors in a Cox proportional hazards model,
the positive association with larger herd size was
reduced but was still marginally significant.
Herd size effects attributable to imperfect diagnostic
tests and sampling strategies
Time to eradicate infection
When disease-eradication programs are based on tests
of imperfect sensitivity, the expected number of false-
negative test results is directly proportional to herd size
(assuming that the entire population is tested and that
test sensitivity and prevalence are constant from herd to
herd) (Martin et al., 1992). Assuming that a herd with n
pigs with a prevalence of infection p is tested with a test
of sensitivity s, the expected number E of infected pigs
that will test negative is
E = np(1 – s)
(6)
The skill and experience of people interacting with
the pigs can influence neonatal mortality, reproductive
performance and other productivity measures (Wilson et
al., 1986). Whether there is also an influence on infec-
tious disease has not been demonstrated clearly, but the
astute herdsperson or manager may be capable of early
detection and intervention in clinical illness and poten-
tially influence the course of disease. Such an influence
might be reflected in a decreased frequency and severity
of lesions at slaughter. The skill and knowledge of the
herdsperson or manager is especially difficult to quantify
in an objective manner and, hence, this effect usually
will be included with any residual herd effect.
Substantial scope probably exists to identify important
management variables that influence disease occurrence in
large herds. Willeberg (1979) and Martinsson and
Lundeheim (1988) showed that for pleuritis lesions at
slaughter, prevalence increased with the number of pigs
slaughtered per year, but there were some large herds that
had prevalences similar to those in small herds.
Explanations for these differences among large herds were
not determined but were probably related to management.
During test-and-removal programs for diseases, time
to eradicate infection (approximated by time under
quarantine) could be greater for large herds even if no
spread of infection occurs during the eradication
process. This might be attributable to management and
For example, a test of 90% sensitivity used in a herd of
100 pigs with a prevalence of infection of 10% would be
expected to yield one false-negative test result, while in a
herd of 1000 pigs with the same prevalence it would
yield 10 false-negative results. Usually, pigs testing nega-
tive on the first test in an infected herd are retested again
until a predetermined number of negative herd tests are
recorded that provide sufficient evidence that the herd is
free of infection. The number of rounds of testing to elim-
inate infection should therefore depend on herd size, the
correlation of the test–retest results of pigs that initially
tested negative yet were truly infected, and the transmis-
sion rate of the pathogen between infectious and
susceptible pigs. Similar considerations apply when com-
binations of tests or pooled tests are used (Christensen
and Gardner, 2000). In the study by Siegel et al. (1993),
only nine of 164 PRV-infected herds were on a test-and-
removal program, so we are unable to determine whether
this theoretical consideration truly applied to the study.
Identification of the herd for intervention
On the other hand, because herd-level sensitivity (the
probability that an infected herd will test positive, usu-
ally meaning that at least one pig in the herd tests

Herd size as risk factor for pig diseases
51
positive) is strongly determined by sample size, a larger
infected herd would be more likely to be identified for
an intervention if a fixed proportion of the herd (rather
than a fixed number per herd) were tested (assuming
other determinants of herd-level sensitivity were held
constant) (Martin et al., 1992).
subsequent analysis. Crude analysis indicated a higher
risk (odds ratio = 2.75) of PRV outbreaks in conventional
herds than in SPF herds. Because herd size was also
strongly positively associated with pseudorabies risk and
conventional herds (mean = 225 pigs) were significantly
(P < 0.0001) smaller than SPF herds (mean = 484 pigs),
adjustment for herd size in a logistic regression model
increased the odds ratio for herd type (conventional ver-
sus SPF) to 10.8. In this analysis, the magnitude of the
conventional herd effect was about four times greater
than that reported by Christensen et al. (1990). We attrib-
ute this difference primarily to the different methods of
selecting control herds.
Examples of the relationship between herd size and
disease risk
Herd size might be non-causally (spuriously) associated
with disease occurrence, because herd size and manage-
ment systems are interrelated and management factors
influence disease occurrence. In this section, we give
some examples, based mainly on our experience in
Danish swine production, to demonstrate how herd size
might influence disease risk independently or in associa-
tion with other management risk factors. We note that
most of the studies are from Denmark, where the swine
population is more homogeneous with respect to man-
agement factors than in many other countries. Hence,
caution must be used in extrapolating findings from
Danish studies of herd size and disease risk to produc-
tion systems in other countries.
Effect modification between herd size and
management variables
Effect modification (also termed ‘interaction’ when used in
a statistical context) occurs when the joint effect of two
variables on disease risk is greater or less than that pre-
dicted from their individual effects. Interaction can be
assessed on an additive or multiplicative scale depending
on how the disease outcome is modeled. One manifesta-
tion of interaction is a difference in stratum-specific odds
ratios, with the crude odds ratio intermediate between the
values for the different categories (strata) of the risk factor.
We present two examples to demonstrate that interac-
tion between herd size and management factors is
possible. Aalund et al. (1976) showed that, in large herds
(>400 fatteners slaughtered per year), a herd history of
diarrhea seemed to protect (odds ratio = 0.6) against
chronic pleuritis, whereas in small herds a history of
diarrhea contributed to a two-fold increase in the odds of
pleuritis. This difference in stratum-specific estimates was
considered evidence of interaction between a herd his-
tory of diarrhea and herd size, and the odds of chronic
pleuritis. Moreover, this interaction was independent of
whether the pigs were purchased or raised in the study
herd. Pointon et al. (1985) found that factors associated
with high prevalences of pneumonia were different in
herds of 20–70 sows (sow culling rate, purchase of pigs
for finishing, pigs per shed section, and pigs per group)
compared with herds of >100 sows (pen stocking rate
and air-space stocking rate). Stratified analyses were not
done by the authors to determine whether confounding
and/or interaction with herd size occurred or whether
the management factors were interrelated, but their find-
ings confirm that the impact of environmental and
management factors often depends on herd size.
Confounding of management variable associations by
herd size
Confounding occurs when the estimate of the effect of a
factor of interest is in part attributable to other risk fac-
tor(s) for the disease. The extraneous risk factor must be
associated with the exposure and also be an independ-
ent risk factor for disease. The net effect of failure to
account for confounding is that the risk estimate (e.g.
odds ratio, risk ratio) for the factor of interest is over- or
underestimated. Statistical methods such as stratified
analyses or multivariable modeling can be used to adjust
risk estimates for confounding variables so that the esti-
mates are unbiased. Detailed descriptions of criteria for
confounding are covered in epidemiological texts.
To demonstrate confounding by herd size, we
reanalysed data from pseudorabies outbreaks in 1987/88
and 1989/90 from 11 districts in southern Jutland adjacent
to the Danish–German border to estimate the risk of
pseudorabies according to herd health status [specific
pathogen-free (SPF) or conventional]. Possible airborne
transmission of PRV was evaluated by restricting analysis
to those herds which neither purchased pigs from nor
had other known contacts with infected herds. Case herds
were those which had positive virus isolation or at least
one seropositive pig, and controls were seronegative
herds located in the same geographical area as the case
herds. Approximately five control herds were sampled for
each case herd. Three hundred and forty herds (58 cases,
282 controls) had complete data and were used in the
Management and infectious agent associations that
are dependent on herd size
In Denmark, complex interrelationships exist between
herd size, health status (SPF or conventional), the occur-

52
Ian A. Gardner et al.
rence of infectious agents, and production system (fat-
tening herd or farrow-to-finish herd). The data used by
Mousing et al. (1990) were reanalysed to demonstrate
these interrelationships and to show that a residual herd
size association can exist after controlling for manage-
ment variables. Although few management factors were
evaluated, the study was unique because it considered
exposure to known infectious causes of pleuritis as risk
factors for slaughter lesions. A subset of 2718 records
[no missing values on any variables and no history of
vaccination against A. pleuropneumoniae serotype 2
(AP2)] were reanalysed to demonstrate how the incorpo-
ration of serological data on exposure to infectious
agents (specifically AP2) may explain at least part of the
herd-size effect. For the analysis, herd sizes were
grouped in quartiles [<620 (reference category),
620–1200, 1201–2000 and >2000 pigs slaughtered per
year]. In bivariable analysis, seropositivity to AP2
increased the odds of pleuritis 9.9 times and the odds
ratio for pleuritis was greater than 1 only for the largest
herd-size quartile (odds ratio = 2.4). After inclusion of
AP2 serostatus in the multiple logistic regression analy-
sis, the odds of pleuritis were substantially reduced in
the largest herds (odds ratio = 1.4), although the confi-
dence interval for the estimate still excluded 1. Estimates
for the two intermediate herd size categories were
essentially unchanged. The residual effect of the largest
herd size category could also not be explained by other
factors examined (herd type and exposure to other
known infectious causes of pleuritis), but since only a
few factors were evaluated it is possible that unmea-
sured management factors were at least part of the
explanation.
age-segregated production systems, and greater use of
compartments for finishing pigs to reduce pathogen
transmission. Many of these changes can be imple-
mented more readily in large herds, and hence the
historic positive relationships between larger herd size
and disease risk may no longer exist.
Although many of the published studies and data sets
that we reanalysed involved Danish swine production,
the underlying issues are widely applicable to other
countries and production systems. Accordingly, we
believe that more structured approaches are necessary
to determine how herd size, herd density, pig stocking
density and other management and environmental fac-
tors affect disease risk. For disease control decisions, it
is especially important to determine whether any herd-
size effect is real or spurious so that appropriate advice
about mitigations can be given.
To ensure the greatest utility of findings from future
studies on the relationship between herd size and dis-
ease, we make the following recommendations.
(i) Herd size should be measured in ways that best
characterize the population at risk for the disease
or productivity measure of interest. For infectious
agents, this would ideally be the number of sus-
ceptible pigs, but in practice only the total number
of pigs and number per building or compartment
will probably be available. In addition, we recom-
mend measurement of pig density (in terms of
floor and air space occupied) especially for dis-
eases of weaned and grower/finisher pigs. Results
of analyses of the herd-size associations should
include evaluation of the joint and individual con-
tributions of each type of herd size measure.
Because standards for stocking density and air
space are well accepted, relatively stable and less
subject to arbitrary categorization than other herd
size measures, we strongly recommend their eval-
uation in all studies of infectious disease. Also,
improved characterization of herd size will help
remove the rather arbitrary classification of herds
as ‘large’ and allow better comparison of results
among studies in different countries.
(ii) Studies that evaluate management-related risk fac-
tors at the level of herd, building or compartment
should consider herd size as a potential con-
founder, unless there is adequate prior evidence to
suggest that herd size is not a causal factor. If herd
size is used as a surrogate for one or more unmea-
sured management risk factors, it is important to
be aware that residual confounding may still exist
after adjustment for herd size, unless herd size is a
perfect correlate for these variables (Greenland
and Neutra, 1980), a situation that we believe is
most unlikely in swine production systems. In
addition, the relationship between herd size and
herd (and pig) density in an area should be clari-
Spurious associations between herd size and disease
For some diseases, an apparent effect of herd size may
be adequately explained when other important risk fac-
tors, such as herd density, pig density and management
factors, are included in analyses. Leontides et al. (1994b)
studied factors associated with PRV circulation in fatten-
ing herds in an intensively vaccinated area of northern
Germany. Herd size increased the risk of PRV (odds
ratio = 1.69 for a 100-pig increase in herd size) in crude
analyses, but when the related demographic and man-
agement variables [farrow-to-finish or feeder-to-finish
(herd type) and pigs/km2] were included in logistic
regression models, the herd size association was no
longer significant (odds ratio = 1.02).
Recommendations for future herd size studies
In the last decade there have been substantial changes
in swine production internationally, including the
increased implementation of all-in, all-out, multisite and

Herd size as risk factor for pig diseases
53
Andersen H (1981). Studies of the relationships between herd-
size, the percentage of pigs culled during the finishing
period and the prevalence rate at slaughter of pigs with
lesions. Nordisk Veterinaermedicin 33: 413–416.
Anderson RM and May RM (1982). Directly transmitted infec-
tious diseases: control by vaccination. Science 215:
1053–1060.
Anderson PL, Morrison RB, Molitor TW and Thawley DG
(1990). Factors associated with circulation of pseudorabies
virus within swine herds. Journal of the American
Veterinary Medical Association 196: 877–880.
Assadi-Rad AM, New JC and Patton S (1995). Risk factors asso-
ciated with the transmission of Toxoplasma gondii to sows
kept in different management systems in Tennessee.
Veterinary Parasitology 57: 289–297.
Austin CC, Weigel RM, Hungerford LL and Biehl LG (1993).
Factors affecting the risk of infection with pseudorabies
virus in Illinois swine herds. Preventive Veterinary
Medicine 17: 161–173.
Bäckström L (1973). Environment and animal health in piglet
production. A field study of incidences and correlations
Acta Veterinaria Scandinavica (Supplementum) 41:
1–240.
Bäckström L and Bremer H (1978). The relationship between
disease incidences of fatteners registered at slaughter and
environmental factors in herds. Nordisk Veterinaermedicin
30: 526–533.
Bäckström L, Morkoc AC, Connor J, Larson R and Price W
(1984). Clinical study of mastitis-metritis-agalactia in sows
in Illinois. Journal of the American Veterinary Medical
Association 185: 70–73.
Benard HJ, Stärk KD, Morris RS, Pfeiffer DJ and Moser H
(1999). The 1997–1998 classical swine fever epidemic in
The Netherlands—a survival analysis. Preventive
Veterinary Medicine 42: 235–248.
Boelaert F, Deluyker H, Maes D, Godfroid J, Raskin A,
Varewijck H, Pensaert M, Nauwynck H, Castryck F, Miry C,
Robijns JM, Hoet B, Segers E, Van Vlaenderen I, Robert A
and Koenen F (1999). Prevalence of herds with young
sows seropositive to pseudorabies (Aujeszky’s disease) in
Northern Belgium. Preventive Veterinary Medicine 41:
239–255.
Bouma A, de Jong MCM and Kimman TG (1995). Transmission
of pseudorabies virus within pig populations is independ-
ent of the size of the population. Preventive Veterinary
Medicine 23: 163–172.
Buddle JR, Mercy AR, Skirrow SZ, Madec F and Nicholls RR
(1997). Epidemiological studies of pig diseases: 1. Use of
French protocols for risk factor assessment to predict the
health status of Australian herds. Australian Veterinary
Journal 75: 274–281.
fied since these two variables are often associated.
(iii) Population-based studies of the interrelationships
among management factors and between manage-
ment factors, herd size, herd density and pig
density are needed. The availability of multivari-
able statistical techniques provides a powerful tool
to better define these relationships, although no
simple relationship exists between forms of the
selected statistical models and the underlying bio-
logical processes (Greenland, 1989). Statistical
techniques to deal with a large number of inde-
pendent variables in epidemiological studies have
been reviewed recently by Dohoo et al. (1997).
(iv) Likely biological reasons for any herd size associa-
tion should be postulated. Moreover, there should
be adequate discussion of the potential biases and
limitations of the study that affect the causal inter-
pretation of a positive or negative association or
no association between herd size and disease risk
(Savitz, 1992). Such candor should also facilitate
improvements in study design and lead to a more
critical assessment of the likelihood of successful
interventions.
(v) Whenever the herd size distribution of study herds
differs from that of the source population, authors
should report at least the median herd size (and
ideally the distribution of herd size) for the coun-
try or region of the study. For risk calculations,
deviations from the median herd size (continuous
variable) or the use of the category with the popu-
lation median herd size as the reference group
should be used as a standard way to report herd
size associations. Such an approach will account
for shifts in herd size distribution with time and
the likely disappearance of herds smaller than a
certain minimum herd size.
Acknowledgments
This research was supported by funds from the Danish
Ministry of Agriculture (KVL-VEL-4), USDA:APHIS:VS
(Cooperative Agreement No. 12–34–99–0020-CA) and
the Center for Food Animal Health, University of
California, Davis. The authors thank Sten Mortensen and
Claes Enøe for technical assistance, and the reviewers
for their constructive suggestions for improving the manu-
script.
Carstensen B and Christensen J (1998). Herd size and sero-
prevalence of Salmonella enterica in Danish swine herds:
a
random-effects model for register data. Preventive
Veterinary Medicine 34: 191–203.
Christensen G and Mousing J (1992). Respiratory system. In:
Leman et al. (editors). Diseases of Swine. Ames (IA): Iowa
State University Press, pp. 138–162.
Christensen G, Vraa-Andersen L and Mousing J (1995). Causes
of mortality among sows in Danish pig herds. Veterinary
Record 137: 395–399.
References
Christensen J (1996). The influence of selected litter and herd
factors on treatments for lameness in suckling piglets from
35 Danish herds. Preventive Veterinary Medicine 26:
107–118.
Aalund O, Willeberg P, Mandrup M and Riemann HP (1976).
Lung lesions at slaughter: associations to factors in the pig
herd. Nordisk Veterinaermedicin 28: 487–495.
Christensen J and Gardner IA (2000). Herd-level interpretation
of test results for epidemiologic studies of animal diseases.
Preventive Veterinary Medicine 45: 83–106.
Abiver N, Seegers H, Beaudeau F, Laval A and Fourichon C
(1998). Risk factors for high sow mortality in French swine
herds. Preventive Veterinary Medicine 33: 109–119.

54
Ian A. Gardner et al.
Christensen LS, Mousing J, Mortensen S, Sørensen KJ,
Standbygaard SB, Henriksen CA and Andersen JB (1990).
Evidence of long distance airborne transmission of
Aujeszky’s disease (pseudorabies) virus. Veterinary Record
127: 471–474.
Cutler RS, Molitor TW, Leman AD and Werdin RE (1983). Farm
studies of porcine parvovirus infection. Journal of the
American Veterinary Medical Association 182: 592–594.
Dangolla A, Willeberg P, Bjørn H and Roepstorff A (1996).
Associations of Ascaris suum and Oesophagostomum spp.
infections of sows with management factors in 83 Danish
sow herds. Preventive Veterinary Medicine 27: 197–209.
De Jong MCM (1995). Mathematical modelling in veterinary epi-
demiology: why model building is important. Preventive
Veterinary Medicine 25: 183–193.
Dewey CE, Wittum TE, Hurd HS, Dargatz DA and Hill GW (1995)
Herd- and litter-level factors associated with the incidence of
diarrhea morbidity and mortality in piglets 4–14 days of age.
Journal of Swine Health and Production 3: 105–112.
Dohoo IR, Ducrot C, Fourichon C, Donald A and Hurnik D
(1997). An overview of techniques for dealing with large
numbers of independent variables in epidemiologic studies.
Preventive Veterinary Medicine 29: 221–239.
Done SH (1991). Environmental factors affecting the severity of
pneumonia in pigs. The Veterinary Record 128: 582–586.
Duffy SJ, Morrison RB and Thawley DG (1991a). Spread of
pseudorabies virus among breeding swine in quarantined
herds. Journal of the American Veterinary Medical
Association 199: 61–65.
Hurnik D, Dohoo IR and Bate LA (1994). Types of farm man-
agement as risk factors for swine respiratory disease.
Preventive Veterinary Medicine 20: 147–157.
Hunneman WA (1986). Incidence, economic effects and control
of Haemophilus pleuropneumoniae infections in pigs.
Veterinary Quarterly 8: 83–87.
Koketsu Y (2000). Retrospective analysis of trends and produc-
tion factors associated with sow mortality on
swine-breeding farms in USA. Preventive Veterinary
Medicine 46: 249–256.
Koopman JS and Longini IM (1994). The ecological effects of
individual exposure and non-linear disease dynamics in
populations. American Journal of Public Health 84:
836–842.
Leontides L, Ewald C, Mortensen S and Willeberg P (1994a).
Factors associated with the circulation of Aujeszky’s dis-
ease virus in fattening herds of an intensively vaccinated
area of Northern Germany. Preventive Veterinary Medicine
20: 63–78.
Leontides L, Ewald C and Willeberg P (1994b). Herd risk fac-
tors for serological evidence of Aujeszky’s disease virus
infection of breeding sows in Northern Germany
(1990–1991). Journal of Veterinary Medicine Series B, 41:
554–560.
Leontides L, Ewald C, Mortensen S and Willeberg P (1995).
Factors associated with the seroprevalence of Aujeszky’s
disease virus in seropositive breeding herds of Northern
Germany during area-wide compulsory vaccination.
Preventive Veterinary Medicine 23: 73–85.
Duffy SJ, Morrison RB and Thawley DG (1991b). Factors associ-
ated with spread of pseudorabies virus among breeding
swine in quarantined herds. Journal of the American
Veterinary Medical Association 199: 66–70.
Elbers AR (1991). The use of slaughterhouse information in moni-
toring systems for herd health control in pigs. PhD thesis,
University of Utrecht, The Netherlands, 152 pp.
Elbers AR, Stegeman JA and de Jong MCM (2001). Factors associ-
ated with the introduction of classical swine fever virus into
pig herds in the central area of the 1997/98 epidemic in the
Netherlands. Veterinary Record 149: 377–382.
Flesjå KI and Solberg I (1981). Pathological lesions in swine at
slaughter. IV. Pathological lesions in relation to rearing sys-
tem and herd-size. Acta Veterinaria Scandinavica 22:
272–282.
Flesjå KI, Forus IB and Solberg I (1982). Pathological lesions in
swine at slaughter. V. Pathological lesions in relation to
some environmental factors in the herds. Acta Veterinaria
Scandinavica 23: 169–183.
Flori J, Mousing J, Gardner I, Willeberg P and Have P (1995). Risk
factors associated with seropositivity to porcine respiratory
coronavirus in Danish swine herds. Preventive Veterinary
Medicine 25: 51–62.
Fox JP, Elveback L, Scott W, Gatewood L and Ackerman E (1971).
Herd immunity: basic concept and relevance to public
health immunization practices. American Journal of
Epidemiology 94: 179–189.
Frost WH (1976). Some conceptions of epidemics in general.
American Journal of Epidemiology 103: 141–151.
Lindqvist J (1974). Animal health and environment in the pro-
duction of fattening pigs. Acta Veterinaria Scandinavica
(Supplementum) 51: 1–78.
Lingaas F (1991). Epidemiological and genetical studies in
Norwegian pig herds. III. Herd effects. Acta Veterinaria
Scandinavica 32: 97–105.
Maes D, Deluyker H, Verdonck M, Castryck F, Miry C, Vrijens B
and de Kruif A (1999). Risk indicators for the seropreva-
lence of Mycoplasma hyopneumoniae, porcine influenza
viruses and Aujeszky’s disease virus in slaughter pigs from
fattening pig herds. Journal of Veterinary Medicine Series
B 46: 341–352.
Maes D, Deluyker H, Verdonck M, Castryck F, Miry C, Vrijens B
and de Kruif A (2000). Herd factors associated with the
seroprevalences of four major respiratory pathogens in
slaughter pigs from farrow-to-finish pig herds. Veterinary
Research 31: 313–327.
Maes DG, Deluyker H, Verdonck M, Castryck F, Miry C, Vrijens
B, Ducatelle R and de Kruif A (2001). Non-infectious fac-
tors associated with macroscopic and microscopic lung
lesions in slaughter pigs from farrow-to-finish herds.
Veterinary Record 148: 41–46.
Manuel-León A and Casal J (2001). Application of a probabilis-
tic approach to the risk assessment of virus airborne
transmission. Veterinary Record 148: 574–575.
Marchevsky N, Held JR and Garcia-Carrillo
C
(1989).
Probability of introducing diseases because of false nega-
tive test results. American Journal of Epidemiology 130:
611–614.
Greenland S (1989). Modeling and variable selection in epidemio-
logic analysis. American Journal of Public Health 79:
340–349.
Martin SW, Shoukri M and Thorburn MA (1992). Evaluating the
health status of herds based on tests applied to individu-
als. Preventive Veterinary Medicine 14: 33–43.
Greenland S and Neutra R (1980). Control of confounding in the
assessment of medical technology. International Journal of
Epidemiology 9: 361–367.
Henningsen D, Mousing J and Aalund O (1988). Porcine corona
virus (PCV) in Denmark: an epidemiological study based on
questionnaire data from screening districts. Dansk
Veterinaertidsskrift 71: 1168–1177.
Martinsson K and Lundeheim N (1988). Inspection findings in
slaughter pigs. The effects of the type of operation and
the size of the herd and the barn. Svensk Veterinartidning
40: 313–319.
Morrison RB, Marsh WE, Anderson PL and Thawley DG (1991).
Factors associated with the seroprevalence of pseudora-
bies virus in breeding swine from quarantined herds.

Herd size as risk factor for pig diseases
55
Journal of the American Veterinary Medical Association 199:
580–583.
Thorburn MA, McDermott JJ and Martin SW (1991). Letter to the
editor: re. Probability of introducing diseases because of
false negative test results. American Journal of
Epidemiology 133: 321–322.
Tuovinen VK, Gröhn YT, Straw BE and Boyd RD (1992). Feeder
unit environmental factors associated with partial condem-
nations in market swine. Preventive Veterinary Medicine 12:
175–195.
United States Department of Agriculture: Animal and Plant
Health Inspection Service: Veterinary Services (1992).
National Swine Survey. Data Collection 1990. Technical
Report. Fort Collins (CO): United States Department of
Agriculture.
United States Department of Agriculture: Animal and Plant
Health Inspection Service: Veterinary Services (1995). Swine
‘95. Part 1: Reference of 1995 swine management practices.
Fort Collins (CO), p. 12.
Mousing J, Lybye H, Barfod K, Meyling A, Rønsholt L and
Willeberg P (1990). Chronic pleuritis in pigs for slaughter: an
epidemiological study of infectious and rearing system-
related risk factors. Preventive Veterinary Medicine 9:
107–119.
Mousing J, Mortensen S, Ewald C and Christensen LS (1991).
Epidemiological and meteorological aspects of Aujeszky’s
disease in Denmark and Northern Germany. In: Thrusfield
M (editor). Proceedings, Society for Veterinary Epidemiology
and Preventive Medicine, London, pp. 18–26.
Mousing J, Permin A, Mortensen S, Bøtner A and Willeberg P
(1997). A case–control questionnaire survey of risk factors
for porcine reproductive and respiratory syndrome (PRRS)
seropositivity in Danish swine herds. Veterinary
Microbiology 55: 323–328.
Pointon AM, Heap P and McCloud P (1985). Enzootic pneumonia
of pigs in South Australia—factors relating to incidence of
disease. Australian Veterinary Journal 62: 98–101.
Pritchard GC (1982). Observations on clinical aspects of transmis-
sible gastroenteritis of pigs in Norfolk and Suffolk, 1980–81.
Veterinary Record 110: 465–469.
Pritchard GC (1987). Transmissible gastroenteritis in endemically
infected breeding herds of pigs in East Anglia, 1981–1985.
Veterinary Record 120: 226–230.
Roepstorff A and Jorsal SE (1990). Relationship of the prevalence
of swine helminths to management practices and
anthelmintic treatment in Danish sow herds. Veterinary
Parasitology 36: 245–257.
Savitz DA (1992). Measurements, estimates, and inferences in
reporting epidemiologic study results. American Journal of
Epidemiology 135: 223–224.
Siegel AM and Weigel RM (1999). Herd factors affecting the selec-
tion and success of intervention strategies in the program for
eradication of pseudorabies (Aujeszky’s disease) virus from
Illinois swine farms. Preventive Veterinary Medicine 40:
243–259.
Siegel AM, Weigel RM and Taft AC (1993). Factors associated with
time under quarantine for swine herds in the voluntary
phase of the Illinois pseudorabies eradication program.
Preventive Veterinary Medicine 16: 295–310.
Smith G and Grenfell BT (1990). Population biology of pseudora-
bies in swine. American Journal of Veterinary Research 51:
148–155.
Stärk KDC, Pfeiffer DU and Morris RS (1998). Risk factors for res-
piratory diseases in New Zealand pig herds. New Zealand
Veterinary Journal 46: 3–10.
van der Wolf PJ, Wolbers WB, Elbers AR, van der Heijden HM,
Koppen JM, Hunneman WA, van Schie FW and Tielen MJ
(2001). Herd level husbandry factors associated with the
serological Salmonella prevalence in finishing pig herds in
The Netherlands. Veterinary Microbiology 78: 205–219.
Weigel RM, Austin, CC, Siegel AM, Biehl LG and Taft AC (1992).
Risk factors associated with the seroprevalence of
pseudorabies virus in Illinois swine herds. Preventive
Veterinary Medicine 12: 1–13.
Weigel RM, Dubey JP, Siegel AM, Kitron UD, Mannelli A, Mitchell
MA, Mateus-Pinilla NE, Thulliez P, Shen SK, Kwok OC and
Todd KS (1995). Risk factors for transmission of
Toxoplasma gondii on swine farms in Illinois. Journal of
Parasitology 81: 736–741.
Willeberg P (1979). The Danish Swine Slaughter Inspection Data
Bank and some epidemiologic applications. Proceedings of
International Symposium on Animal Health and Disease
Data Banks. Washington (DC): USDA. Miscellaneous publi-
cation 1381, pp.133–145.
Willeberg P, Gerbola M-A, Kirkegaard Petersen B and Andersen
JB (1984/5). The Danish pig health scheme: nation-wide
computer-based abattoir surveillance and follow-up at the
herd level. Preventive Veterinary Medicine 3: 79–91.
Willeberg P, Gardner IA, Mortensen S and Mousing J (1994).
Models of herd-size effects in swine diseases. Kenya
Veterinarian 18: 189–191.
Wilson MR, Friendship RM, McMillan I, Hacker RR, Pieper R and
Swaminathan S (1986). A survey of productivity and its
component interrelationships in Canadian swine herds.
Journal of Animal Science 62: 576–582.
Wittum TE, Dewey CE, Hurd HS, Dargatz DA and Hill GW
(1995). Herd- and litter-level factors associated with the
incidence of diarrhea morbidity and mortality in piglets 1 to
3 days of age. Journal of Swine Health and Production 3:
99–104.
Yanga ST, Gardner IA, Hurd HS, Eernisse KA and Willeberg P
(1995). Management and demographic factors associated
with seropositivity to transmissible gastroenteritis virus in
US swine herds, 1989–1990. Preventive Veterinary Medicine
24: 213–228.
Yorke JA, Nathanson N, Pianigiani G, and Martin J (1979).
Seasonality and the requirements for perpetuation and
eradication of viruses in populations. American Journal of
Epidemiology 109: 103–123.
Zimmerman JJ, Dreesen DW, Owen WJ and Beran GW (1990).
Prevalence of toxoplasmosis in swine from Iowa. Journal of
the American Veterinary Medical Association 196: 266–270.
Zimmerman JJ, Owen WJ, Hill HT and Beran GW (1991).
Seroprevalence of antibodies against encephalomyocarditis
virus in swine of Iowa. Journal of the American Veterinary
Medical Association 199: 1737–1741.
Stege H, Christensen J, Nielsen JP and Willeberg P (2001).
Data-quality issues and alternative variable-screening
methods in a questionnaire-based study on subclinical
Salmonella enterica
infection in Danish pig herds. Preventive Veterinary Medicine 48:
35–54.
Stegeman A, Elbers ARW, Smak J and de Jong MCM (1999).
Quantification of the transmission of classical swine fever
virus between herds during the 1997–1998 epidemic in The
Netherlands. Preventive Veterinary Medicine 42: 219–234.
Strom JS (1978). Heat loss from swine and poultry as basis for
design of environmental control systems in livestock build-
ings. Statens Byggeforskningsinstitut, pp. 46–47.
Svensmark B, Jorsal SE, Nielsen K and Willeberg P (1989a).
Epidemiological studies of piglet diarrhoea in intensively
managed Danish sow herds. I. Pre-weaning diarrhoea. Acta
Veterinaria Scandinavica 30: 43–53.
Svensmark B, Nielsen K, Willeberg P and Jorsal SE (1989b).
Epidemiological studies of piglet diarrhoea in intensively
managed Danish sow herds. II. Post-weaning diarrhoea.
Acta Veterinaria Scandinavica 30: 55–62.