2.4.4 Network properties
The network of goat and cattle movements in the three animal health
wards of the MTA were analysed using directed and symmetrized methods
(Borgatti et al., 2013). Nodes were the communities within the wards and
the ties were live goat and cattle movements between communities. A
descriptive statistical analysis was performed and network analysis was
calculated on a directed binary network using UCINet6.66.4 (Analytical
Technologies, USA) (Borgatti et al., 2013). The following indices were
used for the calculation of the network centrality measures:
The degree centrality is the normalized value accounted for by analyzing
the number of ties in each node. The node with a high value reflects a
high number of ties or the channel of node connection. Directed networks
are best described as either out-degree or in-degree centrality, based
on whether the ties are directed away from a node or directed toward a
recipient node.
The between-ness centrality is the Freeman normalized value, which is
considered the shortest path between two nodes. The node with a high
value indicated a high frequency of animal movements through the node.
The closeness centrality is a normalized value that is considered the
geodesic distance from one node to all remaining nodes. A node with a
high value indicated that it is easy to move animals to the linking
node.
The clustering coefficient is calculated from three connected nodes
forming a triangular shape (transitivity) in the network. A network with
a high clustering coefficient means that many node triangles are
present.
The network density is the proportion of actual ties that are present in
the network out of all possible ties.
2.3
Data analysis
Descriptive statistics were presented as frequencies and percentages.
Continuous data were described either using mean ± standard deviation or
medians and interquartile ranges (IQR). The normality assumption for
quantitative variables was assessed by calculating descriptive
statistics, plotting histograms and performing the Anderson-Darling test
for normality within MINITAB Statistical Software, Release 16 (Minitab
Inc., USA). Normally distributed variables were presented as means ± SD
and comparisons performed using one-way ANOVA. Kruskal-Wallis tests were
used to compare centrality measures across the three animal health wards
of the study area. The Wilcoxon signed-rank test was used to compare
centrality measures between goat and cattle movement networks.
Statistical analyses were performed in commercially available software
(IBM SPSS Statistics Version 24, International Business Machines Corp.,
Armonk, New York, USA) and significance was set at P<0.05.
Mapping of the study area displaying the distribution of dip tanks was
performed using ArcGIS 10.2.3 (ESRI, USA), and the sociogram of network
of goats and cattle movements network were performed using the
UCINet6.66.4 programme (Analytical Technologies, USA) (Borgatti et al.,
2013). Nodes were projected using their GPS coordinates estimated using
Google Earth
(https://www.google.com/earth/)
when outside the study area.
3
Results
3.1
Demographic and husbandry findings
A total of 116 smallholder goat farmers were interviewed during
June-July 2018, with 36 respondents from Ward I, 35 respondents from
Ward II, and 45 respondents from Ward III. The median (IQR) age of
respondents for the three wards was 62 (53 – 75), 65 (52 – 79) and 66
(55 – 74) years for Wards I-III, respectively. The major occupation of
respondents was livestock farming with some also involved in private or
government employment (Table 1). In addition to rearing goats, some
respondents also reared cattle, pigs and chickens. The median (IQR)
experience in farming was 27 (14 – 38), 28 (11 – 28) and 19 (10 – 27)
years for Wards I-III, respectively. Eighty-three percent (30/36), 86%
(30/35) and 80% (36/45) of respondents kept cattle in addition to goats
in Wards I-III, respectively. Respondents from Ward I indicated that
their motivation for farming included love for animals (31%),
subsistence (22%), business (19%), draught power (6%), ceremonial and
cultural purpose (31%) and long-term savings and investment (3%). For
the respondents in Ward II, these included subsistence (51%), love for
animals (20%), draught power (11%), business (9%) and long-term
savings and investment (6%). Motivating factors for respondents in Ward
III included subsistence (42%), business (22%), love for animals
(13%), long terms savings and investment (11%), ceremonial and
cultural purposes (7%) and draught power (2%).
Age of respondents, level of education, farming experience and number of
goats owned were not different among the three animal health wards
(Table 2). A total of 134 participants attended the focus group
discussions from across the three animal health wards, with 68%
(91/134) males and 32% (43/134) females.
3.2
Animal movements
There were less reported movements into holdings based on questionnaire
responses relative to movement out of the holdings during the previous
12 months (Table 3). Reported movement into holdings were 5 goats in
Ward I, 17 goats in Ward II and 11 goats in Ward III. The corresponding
number of movements for cattle was 12 heads of cattle in Ward 1, 6 heads
of cattle in Ward II, and 5 heads of cattle in Ward III. Respondents
from Ward I indicated the most recent time of animal movement into the
holdings to be median 6 (1 – 30) months for goats and median 2 (1 – 7)
months for cattle. The most recent time for animal movement into the
holdings for respondents in Ward II were median 6 (4 – 12) months for
goats and median 8 (3 – 12) months for cattle. Respondents in Ward III
indicated a median time of 5 (3 – 12) months for goats and only 2
months for cattle. Livestock movement out of the flock for the previous
12 months preceding the study was 45 goats and 30 heads of cattle in
Ward I, 38 goats and 25 heads of cattle in Ward II and 36 goats and 35
heads of cattle in Ward III. The most recent reported time of animal
movement out of the holdings were median 6 (2 – 11) months for goats
and median 2 (1 – 7) months for cattle in Ward I, median 2 (1 – 8)
months for goats and median 3 (2 – 9) months for cattle in Ward II, and
median 4 (1 – 6) months for goats and median 3 (2 – 8) months for
cattle in Ward III. Most of the respondents in all the study locations
also indicated their ignorance on the need to obtain official veterinary
movement permit to move goats and pigs from their holdings within this
control zone (Table 3).
3.3
Network analysis of goat movements
Data from 116 questionnaires and 13 focused group discussions reported
37 nodes and 78 ties with an overall network density of 0.059 (SD =
0.235) across the study area. Most of the nodes had connections with
each other, with extension to nodes outside the study area (Figure 2).
Village A in Ward I and Village F in Ward II had the largest in-degree
centrality values of 9 and 7 respectively. Moreover, the first five
nodes with the highest values for out-degree centrality were Village F
(12), Village G (11), Village E (7), Village H (7), and Village I (7).
Ten nodes had links with communities outside the study area and animals
were routinely moved to these outside communities either for consumption
or husbandry purposes. On average, the actors (respective nodes of the
study network) had a degree of 2.11 for both in-degree and out-degree,
which was quite low given that there were 37 actors in the network.
Overall, the network had 16 nodes within the study area and 21 nodes
outside the study area. Four locations within the FMD fee zone of the
country (Nelspruit, Tzaneen, Barberton and Leboeng) had links with the
movement of goats from the study area. The range of out-degree was
slightly higher (minimum – maximum: 0 – 12) than that of the in-degree
(0 – 9) with more variability across actors in the out-degree than the
in-degree (standard deviation and variance). The network had an
out-degree coefficient of variation of 154 and in-degree coefficient of
variation of 89. In this network, the out-degree graph centralization
was 28% and the in-degree graph centralization was 20% of the
theoretical maximums.
Closeness centrality measures indicated Villages F and G, both in Ward
II to be the nodes with the highest out-degree closeness values followed
by Village M in Ward III and Village E in Ward I. There was more
variation in the out-closeness value relative to the in-closeness
(minimum – maximum: 0 – 20.33).
Villages F and H both in Ward II, Village E (Ward I) and Village O (Ward
III) had the largest between-ness measures. There was a lot of variation
in actor between-ness (range 0 – 175; standard deviation = 37.34
relative to a mean between-ness value of 17.18). Despite this, the
overall network centralization was relatively low (12.91%) with an
overall graph clustering coefficient of 0.248.
3.4
Network analysis of cattle movements
Overall, data from 116 questionnaires and 13 focused group discussions
represented 42 nodes and 90 ties with an overall network density of
0.052 (SD = 0.223). Most of the nodes had connections with each other,
with extension to nodes outside the study area (Figure 3). Villages J
and M both in Ward III, had the highest out-degree centrality values of
10 and 9, respectively. Giyani, a node in Limpopo Province and
Thulamahashe, a node in Mpumalanga Province both had higher in-degree
centrality measures of 12 and 8 respectively. Thirteen of the nodes
(87%) had links with communities outside of the study area. On average,
the actors had a degree of 2.14 for both in-degree and out-degree, which
was quite low given that there were 42 actors in the network. Overall,
the network had 16 nodes within the study area and 26 nodes outside the
study area. The range of in-degree was slightly higher (minimum –
maximum: 0 – 12) than that of the out-degree (0 – 10) and there was
little variability across actors in-degree and out-degree (in-degree
mean = 2.14, out-degree mean = 2.14, in-degree SD = 2.69, out-degree SD
= 2.72, and in-degree variance = 7.26, out-degree variance =7.41). The
network had an out-degree coefficient of variation of 127 and an
in-degree coefficient of variation of 125. The out-degree graph
centralization was 20% and the in-degree graph centralization was 25%
of the theoretical maximums.
Villages J and M (both in Ward III), and Villages E and A (both in Ward
I) were the nodes with the highest out-degree closeness values. There
was little difference between the in-closeness and out-closeness values
for the network.
Village A (Ward I), and Villages I and F (both in Ward II) had larger
between-ness measures compared to other nodes in the network. However,
there was a large variation in actor between-ness (from 0 – 217)
relative to a mean between-ness value of 31.12. Despite this, the
overall network centralization index was relatively low (12%) and the
network had an overall graph clustering coefficient of 0.129.
3.5
Comparison of movement network centrality measures
There were no significant differences of movement centrality measures
among areas (Table 4). However, median out closeness and between-ness
centrality measures for the network were different by species, with goat
network having more out-closeness centrality (P = 0.021) and cattle
network having more between-ness centrality (P = 0.008; Table 5).
3.6
Additional animal health information and FMD history
The number of goats and the inspection efficiency descriptively varied
by dip tank. However, there was no significant difference in goat
numbers among wards (P = 0.277). The number of pigs in the study area
was small and there was 100% inspection efficacy. Animal health Wards I
and III, had fewer pigs relative to Ward II (P<0.001). Cattle
FMD vaccination coverage for the period preceding the study
descriptively varied by node (dip tank; Table 6). The number of cattle
within the three wards was not significantly different (P = 0.306).
Cattle from Wards I and II had a descriptively longer inter-vaccination
interval (>220 days) than the 120-day inter-vaccination
interval prescribed by South African veterinary services. During 2017, a
SAT2 FMD outbreak was reported at Village E, a community within the
animal health Ward I (OIE-WAHID, 2017). However, the outbreak was
contained within the community without spreading to other nodes within
the study area.
4
Discussion
The primary aim of this study was to evaluate the role that movement of
livestock might play in the spread of FMD within a disease control area
and identify high-risk disease locations for improved surveillance and
strategic vaccination programmes. In this paper, the livestock movement
networks were investigated because of our desire to improve FMD control
in the country. It is expected that the results of this study will be
useful for disease control through the implementation of risk-based
surveillance and strategic vaccination in disease endemic countries
prioritizing FMD protection zones and high-risk production sectors
within their regions.
The number of goats kept by respondents in this study were similar to a
smallholder study conducted in another part of the country (Braker et
al., 2002). In addition to keeping goats, communal farmers also kept
cattle and pigs, which are also susceptible hosts for FMD. Most communal
farmers within the study area reported that they do not require
livestock movement permits to move goats to neighboring and distant
locations. This suggests a need to educate farmers concerning the risk
of livestock movement out of disease control areas. Most goats are bred
and consumed locally within the communities with some reported movements
outside the study area.
Livestock movement contributes to the spread of infectious diseases from
endemic to free zones (Nremark et al., 2011). In South Africa, the
majority of the country was previously classified as FMD free (DAFF,
2014). The KNP and adjoining nature reserves were classified as part of
the FMD infected zone due to the existence of wildlife reservoirs
including the African buffalo (Syncerus caffer ) (DAFF, 2014).
Communal farming areas surrounding the KNP were classified as the FMD
protection zone with vaccination, where cattle are routinely inspected
and vaccinated against FMD (Lazarus et al., 2018). Movement of livestock
from the FMD protection zone to any other part of the country requires
inspection and a movement permit (DAFF, 2014). In this study,
respondents reported more movement of animals out of their holdings
relative to movement into their holdings during the previous 12 months.
The most influential nodes for goat movements were communities closer to
urban settlements with an accessible road network. Villages F and G both
in Ward II, had the highest out-degree centrality measures for goat
movements, and thus were the most influential communities for the
possible spread of diseases. Goat movement out of the holdings was
independent of each other to reach everyone in the network as
demonstrated by high out-closeness centrality measures.
Villages J and M, both in Ward III, were the most influential in the
cattle movement network as demonstrated by their high out-degree
centrality measures. The spread of infection is reportedly associated
with out-degree centrality measures (Dubé et al., 2008). Consequently,
the nodes with high out-degree centrality are spreaders of disease and
are likely to increase the size of an epidemic. However, to prevent
these nodes from spreading FMD, the local veterinary authority could
give high priority to FMD control and vaccination within these
communities. The cattle vaccination coverage for Villages F and G for
the period preceding the study was 72.3% and 74.6% respectively.
However, the median (IQR) time since last vaccination for the first five
nodes with the highest out-degree centrality was 232 (220 – 233) days
and this should be interpreted relative to the desired 120-day
inter-vaccination interval. This prolonged interval has been previously
reported from the study area (Lazarus et al., 2017). Cattle within these
nodes have exceeded the expected vaccination interval and thus there
might be a high proportion of susceptible cattle present. The cattle
movement network had high likelihood of being an intermediary to reach
everyone in the network as demonstrated by the high between-ness
centrality measures.
Village A had the highest in-degree centrality measure for the goat
movement network followed by Villages F and O, which suggests high-risk
nodes for disease outbreak occurrence. The cattle vaccination coverage
for these nodes varied from 72.3 – 93%, but Villages F and A had
prolonged inter-vaccination intervals. Therefore, more FMD surveillance
should be focused on these nodes as they tend to receive more inward
animal movements relative to all other nodes. Village F had a relatively
high centrality measure for goat movement, and this might be due to its
location and accessible road network compared to the more remote
settings. Farmers in Village F tended to source goats from nearby nodes
and then export them to distant locations. A similar movement pattern
was previously described in the study of traditional cattle trade
network in Tak Province, Thailand (Khengwa et al., 2017). In the current
network, the out-degree graph centralization of the goat movement
network is much greater than the in-degree graph centralization, and
this suggests that there is proportionally more out-degree movement in
this network. These communities require additional education on disease
prevention and control. In terms of the network analysis, such
communities could be described as the disease spreaders. Therefore, in
the event of an FMD outbreak within the country, the relevant
authorities should focus their disease control measures on nodes with
higher out-degree centrality measures and middlemen involved in
livestock movements.
The two nodes with the highest in-degree centrality measures for the
cattle movement network were nodes outside of the study area, Giyani in
Limpopo Province and Thulamahashe in Mpumalanga Province. Both nodes are
associated with urban development and organized abattoirs and
butcheries. Middlemen might therefore be key players in the movement of
animals within this network. This is similar to a social network
analysis of cattle movements in Kampong Cham, Kampong Speu and Takeo,
Cambodia (Poolkhet et al., 2016). Implementing restrictions on trade
activities and livestock movements from the nodes with high out-degree
measures and the middlemen might limit the magnitude of disease spread
to other nodes. In this network, farmers themselves could be effective
disseminators of information to improve communication for the greater
benefit of the network.
The between-ness centrality for the goat movement network was high for
two nodes in Ward II and one node each in Wards I and III, reflecting
the many steps connecting nodes to one another. Villages F and H both in
Ward II, Village E in Ward I and Village O in Ward III, appear to be
more important than the other nodes by this measure. This suggests that
more animals pass through these nodes relative to the other nodes of the
network. Interestingly, in 2017, a SAT2 FMD outbreak was reported in
Village E (Ward I), one of the communities with high between-ness
centrality measure for the goat movement network (OIE-WAHID, 2017).
In 2019, two SAT2 FMD outbreaks (DAFF, 2019a; DAFF, 2019b) were reported
in the FMD free zone of Limpopo Province, which were linked to animal
movements but not associated with the study area. However, on the
3rd of March 2020, a SAT2 FMD outbreak was reported at
Villages J, M and K (Ward III) which subsequently spread to Villages E
and A (Ward I) by the second week of April, 2020 (Mr Solly Mokone,
Animal Health Technician, personal communication). From our study,
Villages J and M are the most influential nodes in cattle movements as
demonstrated by their out-degree centrality measures. The rapid spread
from the three initial villages is therefore consistent with the
expectations based on the current findings.
Goats have been described to be “silent shedders” of FMD without
showing obvious clinical signs and sickness behaviours. These facts make
it very difficult to identify infected goats that might pose a threat
for the spread of disease through animal movements. The results of this
study indicate that goats are moved by communal farmers out of the study
area without official movement permits, although, the absolute number of
movements appears to be low. The study also identified communities at
high risk of disease occurrence and communities that might play
important roles in subsequent disease spread. Four locations in the
(former) FMD free zone of the country (Nelspruit, Tzaneen, Barberton and
Leboeng) were identified as having connections with movement of goats
from the study area and this calls for careful monitoring to mitigate
the potential spread of FMD from the FMD protection zone.
The results of this study should be interpreted considering several
limitations. A goat identification system and an organized database are
not in use within the study area, even though each ward has a small
stock register. Therefore, it was not possible to verify the movement of
individual animals. Animal movement data derived in this study were
solely based on interviews and subject to recall bias and purposeful
misinformation. We were unable to follow up each origin and destination
reported by the respondents outside of the study area due to limited
budgets and resources. Other limitations to the study include the
limited study area and the lack of production records for the
verification of herd additions and subtractions.
The results of this study suggest the need to improve control measures
within FMD protection zones and high-risk production sectors in disease
endemic countries. We recommend the following: control goat movements
via official movement permit systems, establish organized goat auction
points or markets to control trader activities, initiate traceable
livestock identification systems, develop databases for livestock
movements and improve surveillance and inspection of FMD in goats within
FMD protection zones and regions with high-risk production sectors such
as dairy industries.
Cattle entering communities with high in-degree measures should be
properly inspected and ensured that they have originated from herds that
were vaccinated against FMD. Farmers moving cattle from the communities
with high out-degree measures should ensure that their animals are
vaccinated against FMD and that they are inspected and issued official
movement permits in compliance with the local disease control policy.
Presented information could be used to improve FMD control not only
within the study area, adoption in other rural settings of southern
Africa could improve the progressive control of FMD in general.