Haemorrhagic Scepticemia in Buffaloes
Haemorrhagic Scepticemia in Buffaloes
Haemorrhagic Scepticemia in Buffaloes
com/scientificreports
Haemorrhagic septicaemia (HS) is an economically important disease affecting cattle and buffaloes
and the livelihoods of small-holder farmers that depend upon them. The disease is caused by Gram-
negative bacterium, Pasteurella multocida, and is considered to be endemic in many states of India
with more than 25,000 outbreaks in the past three decades. Currently, there is no national policy
for control of HS in India. In this study, we analysed thirty year (1987–2016) monthly data on HS
outbreaks using different statistical and mathematical methods to identify spatial variability and
temporal patterns (seasonality, periodicity). There was zonal variation in the trend and seasonality of
HS outbreaks. Overall, South zone reported maximum proportion of the outbreaks (70.2%), followed
by East zone (7.2%), Central zone (6.4%), North zone (5.6%), West zone (5.5%) and North-East zone
(4.9%). Annual state level analysis indicated that the reporting of HS outbreaks started at different
years independently and there was no apparent transmission between the states. The results of the
current study are useful for the policy makers to design national control programme on HS in India and
implement state specific strategies. Further, our study and strategies could aid in implementation of
similar approaches in HS endemic tropical countries around the world.
Keywords Neglected tropical disease, Cattle, Buffalo, South Asia, Haemorrhagic septicaemia, Spatial and
temporal patterns, Zonal variation, Control programme
The livestock sector contributes significantly to global food security, and the demand for livestock and their
products is increasing. The Indian livestock sector accounts for 70% (excluding poultry) of the total livestock
population of South Asia, and is known to contribute to India’s Gross Domestic Product (GDP) by around 6%
over the last few decades. It is expected to grow much faster in the f uture1. This sector supports the livelihoods
of over 200 million rural p oor2. Although, the livestock sector has the potential to alleviate poverty2, infectious
diseases of livestock pose an enormous barrier to animal welfare, increased food production and livelihood
improvements. A key infectious disease in this sector is ‘Haemorrhagic Septicaemia (HS)’ in cattle and buffalo.
Haemorrhagic septicaemia is a highly contagious, fatal, and septicaemic disease, caused by Pasteurella mul‑
tocida strains especially serotype B:2 (Asian type) among five prevalent capsular serotypes (A, B, D, E, F) and
sixteen somatic serotypes (1–16), which belong to the family ‘Pasteurellaceae’3–6. Susceptible animals become
infected either by inhaling or ingesting the organism. The incubation period ranges from 1 to 3 days leading to
sudden death without visible clinical signs. In protracted cases, the incubation period can extend up to 5 days
or more. Initially, there is a high fever followed by respiratory distress (rapid and shallow breaths), septicaemia,
muco-purulent nasal discharge, restlessness, oedematous swelling of the throat/brisket region, mild muscular
tremors and recumbency leading to d eath7,8. On post-mortem examination, subcutaneous edema in the man-
dibular/brisket regions and petechial-to-echymotic hemorrhages, congestion and/or consolidation of the lungs,
fibrinous pneumonia, pleurisy and pericarditis are n oted7,8.
In areas where HS is endemic, vaccination either with alum precipitated or oil adjuvanted killed bacteria is
commonly practiced. This provides 6 to 9 months of protective immunity in susceptible a nimals8–10. The vaccine
1
ICAR-National Institute of Veterinary Epidemiology and Disease Informatics (NIVEDI), Ramagondanahalli,
Yelahanka, Bengaluru 560064, Karnataka, India. 2UK Centre for Ecology and Hydrology, Benson Lane, Crowmarsh
Gifford, Oxfordshire OX10 8BB, UK. *email: [email protected]; [email protected]
Vol.:(0123456789)
www.nature.com/scientificreports/
is used in different states of India and is manufactured by either government or private companies. The vaccine
and vaccination service is provided free of cost to the farmers. Affected animals are treated with antibiotics
and antipyretics11. Hygienic practices, and preventing overcrowding of animals during the monsoon season
are advised to prevent the occurrence of HS cases12. India is classified Category A in terms of the global impact
of hemorrhagic s epticaemia10, meaning that the disease is endemic and of utmost economic importance to the
country.
Hemorrhagic septicaemia is an economically important disease of cattle and buffalo. Buffaloes are highly
susceptible to HS with high morbidity and mortality in comparison to other susceptible s pecies13. In another
study, it was found that the morbidity rate was highest in crossbred cattle, followed by buffaloes and indigenous
cattle, but the mortality rate and proportion of cases was highest in b uffaloes14. The economic impact of HS in
Cambodia was found to be $375 per animal, with benefit of $951.58 per household when vaccinated bi-annually15.
In one study, it was found that economic losses due to morbidity were 23% and losses due to mortality were
77%16. Similarly, the economic losses due to HS were estimated in the state of Karnataka, India, through a primary
survey of HS affected farms17. The loss estimates were shown to be $275, $284, and $415 per head of affected
indigenous cattle, water buffaloes and crossbred cattle, respectively, with a total loss of $23.89 million/year
from 2007–2011, the loss due to HS was estimated to be ₹6816 (~ $80) per head of cattle and ₹10,901 (~ $130)
under high incidence scenario was for the state of K arnataka17. However, in another study using outbreak data
per head of affected buffalo. When scaled up to the national level, the projected loss was ₹5255 crores (~ $ 800
million)16. The difference in the loss estimates may be due to the type of data used. The s tudy17 conducted a
primary survey to estimate the loss of HS, while the other s tudy16 used only outbreak data which may account
for the difference in estimates.
Foot-and-Mouth Disease (FMD) is the number one disease of cattle and buffalo in India, which is known to
cause high economic loss to farming18. India has 56.7% of the worlds buffalo population and buffalo is a major
dairy animal that contributes more than 50% to the country’s milk production19. Buffalo is less susceptible to
FMD and the reported mortality is lower than that of crossbred cattle. However, the economic loss due to HS
is higher than F MD16,17. Therefore, economically, HS is a more important disease of buffalo than FMD and
requires higher resource allocation to control and prevent the disease more effectively. The resource allocation
can be focused on increasing testing rates and improving vaccination coverage in buffalo so that the disease can
be effectively prevented.
Seasonality may or may not be present in the temporal outbreak data. If it is present, it may be due to climatic
factors or temporal autocorrelation in the data as well as nature of the disease being studied. It is expected that
HS outbreaks will exhibit seasonality, either due to the influence of weather or the presence of carrier/recovered
animals that can act as a source of infection to other animals. This is further compounded by the waxing and
waning of immunity, which lasts for 6–9 m onths9,10. The presence of carrier animals makes other animals sus-
ceptible to the disease, leading to seasonality in occurrence of the disease. Additionally, the general practice of
vaccinating for HS before the onset of monsoon contributes to the seasonality in the occurrence of the disease.
In animal disease outbreaks, the data tends to show spatial autocorrelation, meaning that observations that are
closer to each other are more related than those farther apart. Spatial autocorrelation may occur due to the spread
of the disease to neighbouring districts (for example due to the movement of animals) or due to the presence
of covariates with spatial structure. HS is known to be a seasonal disease and rainfall has been associated with
it’s occurrence. However, very few studies13,20–22 have investigated the spatial variability, seasonal pattern and
periodicity in the occurrence of HS outbreaks in India. Epidemiological analysis of HS data is crucial in formulat-
ing a systematic control strategy in endemic c ountries10 by identifying seasonal patterns and spatial variability.
In India, there is no active surveillance for HS and diseases are reported by field veterinarians with the vil-
lage as an “epi-unit” through a passive surveillance mechanism. Field veterinarians are assigned with many
responsibilities such as treatment, vaccination, deworming and other departmental extension activities. There is
always a chance of under-reporting of HS in India. The field veterinarians are under pressure to report a disease
and this will affect the overall passive surveillance data used in the country. Overall, the country has weaker
veterinary infrastructure compared to developed countries. There are only 65,815 veterinary centers catering to
the needs of more than 500 million livestock and 800 million poultry population of the country23. Additionally,
there are limited numbers of district and state-level laboratories to test suspected samples. However, diseases like
HS with very prominent clinical signs and a high case fatality rate will seldom be missed by the farmers and are
invariably brought to the notice of field veterinarians. The HS outbreak data used in the current study is passive
surveillance data collected from different states of India and maintained at ICAR-NIVEDI. ‘Village’ is a spatial
unit and ‘month’ is a temporal unit in the collection of HS data. The outbreaks of HS are reported by the field
veterinarians and they are either clinically confirmed or laboratory confirmed cases.
Present study aimed to address following questions through the analyses of three decades (1987–2016) of
HS outbreaks data from India.
1. Which zone of India is worst and most frequently affected by HS over the study period? (Spatial analysis)
2. Is seasonality of HS consistent over the study period and between regions and could this knowledge be help-
ful for planning control? (Singular spectrum analysis)
3. Can we detect evidence of long term inter-annual cycles in outbreak numbers? (Wavelet analysis)
4. Is there evidence of an increase or decrease in numbers of HS outbreaks over the study period, how does
this trend vary between regions and can it be linked to changes in livestock numbers, control or surveillance
policy? (Piecewise regression analysis)
5. What is the evidence that patterns in HS are linked or synchronous between states, due to the influence of
inter-state risk factors? (Bayesian spatio-temporal model)
Vol:.(1234567890)
www.nature.com/scientificreports/
Overall, the analyses of HS disease data were aimed at understanding the spatial and temporal patterns of the
disease as well as to provide basis for policy makers to design effective control strategy for the disease in India.
Results
A total of 26,305 HS outbreaks were reported between January 1987 and March 2016 in India. The proportion
of outbreaks in each year that were recorded in each zone was plotted to understand which zones consistently
contributed to high or low number of HS outbreaks. The proportion of each year’s outbreaks that were reported
in each zone is shown in Fig. 1. The South zone reported proportionately more outbreaks compared to other
zones across all the years except for 2013 wherein the East zone reported more outbreaks. Karnataka and Andhra
Pradesh states consistently reported the highest proportion of annual HS outbreaks across the study period
(Fig. 2), with Karnataka reporting the highest proportion of outbreaks between 1987 and 1991 and again in
Figure 1. Year-wise proportion of HS outbreaks in each zone. Proportion is shown on a scale of 0–1.
Figure 2. Heat map showing the state-wise distribution of HS outbreaks in different years. Proportion is shown
on a scale of 0–1.
Vol.:(0123456789)
www.nature.com/scientificreports/
the second half of the 2000s, and Andhra Pradesh reporting the highest proportion of outbreaks in the inter-
vening period (1992–2005). Andhra Pradesh reported the greatest number of outbreaks (n = 10,285) followed
by Karnataka with reports of 7399 outbreaks during the study period. Madhya Pradesh, Gujarat, and Punjab
also frequently reported a reasonable proportion (~ 10%) of annual HS outbreaks across the study period. The
remaining states reported small numbers of annual HS outbreaks sporadically. Overall, the South zone reported
the maximum proportion of the outbreaks (70.2%), followed by the East zone (7.2%), the Central zone (6.4%),
the North zone (5.6%), the West zone (5.5%) and the North-East zone (4.9%). The Central zone has a higher
proportion of cattle (26.71%) and buffalo (37.89%) compared to other zones. The North zone has 11.84% and
23.78%, the East zone has 25.72% and 9.27%, the South zone has 15.36% and 13.78% and the North-East zone
has 6.96% and 0.53% proportion of cattle and buffalo population respectively.
Proportion of HS outbreaks in different states in each zone is shown in Fig. S1A–F. Proportion of HS out-
breaks in comparison to buffalo and cattle population in each zone is given in Fig. S2. State-wise proportion
plot of cattle and buffalo population in each state (Fig. S3) and state wise population maps of cattle and buffalo
are shown in Fig. S4.
Figure 3. Decomposition of the monthly HS time series using singular spectrum analysis. The original time
series is overlaid with the seasonality, trend and residual components.
Vol:.(1234567890)
www.nature.com/scientificreports/
Figure 4. Wavelet transformation to detect dominant frequencies in monthly HS outbreaks. The white dotted
line is the cone of influence indicating the region of time and frequency where the results are not influenced by
the edges in the data and are reliable. The solid black line corresponds to the 95% confidence interval and the
areas within this black solid line indicate significant variability at the corresponding periods and times. X-axis:
time in months from January 1987 (= month 1), Y-axis: localised periodicity in months.
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 5. (A) Plots of piecewise regression analysis with year as dependent variable. Grey filled areas are 95%
confidence interval. Blue is the break point identified. (B) Plots of piecewise regression analysis with month as
dependent variable. Grey filled areas are 95% confidence interval. Blue is the break point identified. (a) North
zone, (b) North East zone, (c) East zone, (d) Central zone, (e) West zone and (f) South zone.
HS had been reported from Madhya Pradesh and other states like Uttaranchal, Uttar Pradesh, while Chhattisgarh
had reported sporadic outbreaks. In the East zone, HS outbreaks had been reported from all the states with the
Vol:.(1234567890)
www.nature.com/scientificreports/
Table 1. Change point detection using piece wise regression analysis using year and month as independent
variable. The change point is shown with ± standard error (SE).
maximum reports from districts of West Bengal. The majority of the districts of Gujarat and Maharashtra had
reported HS outbreaks in Western zone. Majority of the districts of Karnataka and Andhra Pradesh had reported
HS outbreaks followed by Kerala. Tamil Nadu reported sporadic HS outbreaks.
Spatial model
The spatial models were fitted to different zones separately, and the results are shown in Table 2. The comparison
of models based on the lowest DIC shows that iid models are best for all the zones indicating the presence of
distinct district-level factors contributing to the district- level variability in the occurrence of HS outbreaks. The
spatial unstructured heterogeneity maps are shown in Fig. S8A–F.
Discussion
Understanding the spatial and temporal patterns of animal diseases is critical in planning systematic surveillance
in at-risk areas during high-risk periods so that effective prevention and control strategies can be formulated. In
the present study, thirty-year HS outbreak data was analysed to identify the spatial and temporal patterns of HS
outbreaks in different zones of India using spatial and temporal methods. We used a spatio-temporal model to
identify the inter-annual variability of the disease in different states of India. There is seasonality in the occur-
rence of HS outbreaks in India with the peak of outbreaks happening in August. No inter-annual periodicity
was detected in all zones except the North-East zone, indicating that long term immunity or climate extremes
may not be playing a role in determining the occurrence of HS outbreaks. There are some district-level factors
that determine the spatial variability of HS outbreaks in different zones of India, and there is no apparent trans-
mission between states. Hence, this study shows that there is a need for state specific or zone specific policies to
prevent and control HS in India.
There will be either short-term (seasonal) or long-term (inter-annual) dependencies in the temporal data
and hence appropriate methods should be used to quantify their importance. Seasonality was detected across all
zones except, for the North-East zone where seasonality was not detected. There was no significant periodicity
detected in wavelet analysis for all the zones, except for the North-East zone. In the North-East zone, dominant
and consistent periodicity was detected around 5–6 years and around 8–10 years (Fig. 4B). The dominant perio-
dicity around 5–6 years, indicates that HS in this region may occur with a periodicity of 5–6 years.
Piecewise regression analysis detected specific years and months in which there was a trend (increasing or
decreasing). There was non-linearity in the occurrence of the outbreaks at the yearly and monthly levels. The
varying trend in different years indicated specific changes in policy in certain years, such as vaccination cover-
age, changes in livestock population (especially cattle and buffalo), or the influence of rainfall which need to be
further investigated. The varying monthly and yearly trend detection using piecewise regression is advantageous
over other linear models which fail to capture the non-linearity in the occurrence of HS outbreaks in India. The
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 6. Cumulative (1987–2016) spatial distribution of HS outbreaks in the zone ( log10 transformed) in
different zones of India.
varying trend indicates that the disease can be prevented in the future by adopting a specific vaccination schedule
for each zone or state to protect the animals against HS.
Spatial analysis of diseases can help in the eradication of infectious diseases by implementing cost-effective
disease management schemes in newly predicted areas with spatial structure. In this study, we employed dif-
ferent spatial models for each zone to detect spatial patterns in the disease. The iid models outperformed the
Besag or BYM models for all the zones (Table 2A). This indicates that there were certain district level factors
(hosts, weather, land cover etc.,) that were different in each district, and those factors were more important
Vol:.(1234567890)
www.nature.com/scientificreports/
A. Spatial models
Zone BYM Besag iid
South zone 657.24 663.98 657.22
Central zone – – 339.74
Eastern zone 673.24 678.88 673.23
North-East zone 298.45 308.90 298.42
North zone 497.53 507.42 497.23
West zone 428.61 440.48 428.72
B. Space–time models
Model DIC
Model without spatial and temporal autocorrelation 235,817.60
Model with spatial and temporal autocorrelation 22,803.39
AR (1) + besag 22,804.68
AR (1) + iid 22,803.58
Model with spatial and temporal autocorrelation and Type-1 interaction 2743.68
Table 2. Spatial and spatio-temporal models. (A) DIC of different spatial models (BYM, Besag and iid) fitted
for different zones. It shows that the iid models are the best models for all zones based on the DIC values.
Note: For Central zone, DIC could not be computed for Besag and BYM models and it was computed only for
iid model. (B) Comparison of different spatio-temporal models with DIC. Model with spatial and temporal
autocorrelation and Type-1 interaction was the best model with lowest DIC.
rather than factors that are spatially autocorrelated (Fig. S8). In addition, certain states contributed to more
outbreaks (Fig. S1A–F) to that zone compared to other states, and hence, this could influence the spatial pat-
tern in outbreaks. Within the endemic zone or state, there were some districts that were affected more severely
than others. Recent studies on whole genome analysis and genetic typing revealed that there is genetic diversity
of Indian isolates of P. multocida within24 and between the geographical regions25. The factors responsible for
variability in HS outbreaks at the district level can be further investigated by incorporating covariates. There
are no reports on spatial analysis of HS data from India using statistical methods except a recent study that
showed seroprevalence of HS was significantly higher in indigenous cattle compared to crossbred c attle26, but
these methods are commonly employed to analyse livestock diseases in other countries. The spatial pattern of
Foot-and-Mouth disease (FMD) in China was analysed to detect spatial autocorrelation and hotspots for the
disease27. A continental risk map for the occurrence of Rift Valley fever was prepared using a Bayesian spatial
model by including environmental c ovariates28. A Bayesian model was used to identify the risk factors for the
occurrence of Classical Swine Fever (CSF) in B ulgaria29. Exploratory spatial data analysis and the Empirical Bayes
method were used to identify high-risk areas for human brucellosis which can be useful in allocating resources
to reduce the burden of the d isease30.
Spatial and temporal patterns of animal diseases are important not only to contain the diseases within a
country but can also help to prevent spread to neighbouring c ountries31. Space–time clusters were detected for
bluetongue outbreaks in the Balkan Peninsula using scan s tatistics32. Bayesian survival models were used to
understand the spatio-temporal dynamics of pneumonia in bighorn s heep33. Spatio-temporal analysis of anthrax
outbreaks in Ukraine during the past century (1913–2012) identified disease hotspots and found that outbreaks
have reduced during the latter part of the c entury34. In the present study, space–time multi panel plots showed
that the HS outbreaks had inter-annual variability in each state (Fig. 7). Inter-annual variability may be either
due to variability in weather or because of an effective vaccination strategy in certain years compared to other
years. Space–time data can be analysed by different methods35. In this study, space–time data on HS outbreaks in
each state was analysed using a Bayesian Poisson Generalised linear model accounting for spatial and temporal
autocorrelation. The model with spatial and temporal autocorrelation random effects outperformed the mod-
els without random effects indicating the importance of these terms in explaining the state-wise inter-annual
variability of HS outbreaks (Table 2B). Interestingly, iid model for spatial terms and AR (1) for the temporal
component was better than the model with Besag component and AR (1). Additionally, the inclusion of type-1
interaction (interaction between iid of spatial component and iid of temporal component) further reduced the
model DIC compared to the model without the interaction term (Table 2B). This indicates that there were some
state-specific policy changes in certain years (e.g. effective vaccination) or change in state-specific covariates
(weather, host, or land cover) that led to the inter-annual variabilities within and between the states. The model
presented in this study can be extended to include covariates either at the district or state level for developing
an early warning system for HS in India.
There is a need to practice vaccination at least 3 months before10 to achieve maximum protection. Each zone
was different in seasonality of the disease and hence, specific months should be selected to achieve maximum
protection. There was also spatial variability in the occurrence of HS outbreaks between and within zones. Hence,
resources (vaccines, personnel etc.) could be allocated accordingly to reduce the economic burden of the disease
in India. The comprehensive epidemiological data analyses of HS outbreaks presented in this study are likely to
assist policy-makers in systematic planning of control strategy for the disease in India.
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 7. Spatio-temporal patterns of HS outbreaks in India (1987–2016). The HS outbreak data were log10
transformed. Number denotes names of different states and they are: 1, Andhra Pradesh; 2, Arunachal Pradesh;
3, Assam; 4, Bihar; 5, Chhattisgarh; 6,Goa; 7, Gujarat; 8, Haryana; 9, Himachal Pradesh ; 10, Jammu and
Kashmir; 11, Jharkhand; 12, Karnataka; 13, Kerala; 14, Madhya Pradesh; 15, Maharashtra; 16, Manipur; 17,
Meghalaya; 18, Mizoram; 19, Nagaland; 20, Odisha; 21, Punjab; 22, Rajasthan; 23, Sikkim; 24, Tamil Nadu; 25,
Tripura; 26, Uttar Pradesh; 27, Uttaranchal; 28, West Bengal; 29, Andaman and Nicobar; 30, Chandigarh; 31,
Dadra and Nagar Haveli; 32, Daman and Diu; 33, Delhi; 34, Lakshadweep; 35, Pondicherry.
Vol:.(1234567890)
www.nature.com/scientificreports/
On the basis of analyses presented in this study, it is evident that despite control strategies adopted by respec-
tive state animal husbandry departments for annual HS vaccination (either alum precipitated or oil adjuvant-
based killed HS vaccine) in selected endemic regions over several years, HS outbreaks were recorded which might
be due to improper vaccination strategy (month of vaccination, areas for vaccination, etc.,). Although the average
loss per animal due to mortality following HS was estimated to be around $275 (indigenous cattle), $284 (water
buffaloes) and $415 (crossbred cattle) in different animals, Benefit-Cost Analysis (BCA) of the HS vaccination
was estimated at different incidence scenarios such as 5.97:1 (high), 4.48:1 (medium) and 2.98:1 (low)17, but,
the countrywide actual impact of morbidity/mortality due to HS alone or in association with concurrent infec-
tions/varying risk factors at different agro-climatic regions needs to be estimated. Haemorrhagic septicaemia
continues to haunt the health status (although at a decreasing trend) of the Indian livestock population, a prag-
tentatively scaled up to cost approximately ₹5255 crore per annum (~ $ 800 million)16. Furthermore, a serious
matic National Control Programme on HS is imminent10,36, in order to reduce the disease burden, which was
revisit to the existing strategy involving widely used vaccine strain, vaccination schedule, adjuvant and delivery
systems should be a priority for the sustainable development of the livestock sector under the National Livestock
Mission (NLM) launched by DADF (Department of Animal Husbandry, Dairy and Fisheries), Government of
India36. There is a need to develop a prevention and control strategy for the disease by utilising the resources in
high-risk areas as well as high-risk periods, and targeted surveillance in order to reduce the burden of HS as it
could be curtailed by implementing a pragmatic HS National control program in India.
Conclusion
Overall we found that the South zone is the worst and most frequently affected zone among all the zones.
Therefore, resource allocation for surveillance and vaccination programs should be more concentrated in the
South zone. In the other zones, there is a need to improve the surveillance for the early detection of HS cases.
Seasonality was found in all the zones, with outbreaks being more common in the month of August. Therefore,
vaccination can be carried out in the months of May–June for better protection to the susceptible population.
Except for the North- East zone, no long-term periodicity was detected in any of the other zones. The absence
of long term periodicity indicates that there is no impact of long term changes in climate on the occurrence of
HS outbreaks that needs to be further investigated. Non-linearity was detected in the monthly and yearly occur-
rence of HS outbreaks and some break-points were identified. These breakpoints, whether occurring monthly or
yearly, may be due to changes in vaccination policy or changes in livestock numbers, mainly cattle and buffalo.
The spatial variation of HS outbreaks at the district level may be attributed to some distinct district level factors
and there is no evidence of transmission between districts. The analyses also revealed that there is no apparent
transmission of HS between states, suggesting that state-level HS policies can be formulated for effective preven-
tion and control of HS in India.
Vol.:(0123456789)
www.nature.com/scientificreports/
Methods
Zonal level time series analysis
Decomposition of HS outbreaks using Singular Spectrum Analysis (SSA) at zonal level. SSA was performed
on the zonal-level HS outbreaks data for all the zones. The SSA algorithm decomposes the time series into
different components and each component can be identified as either a trend, periodic component, or noise.
Subsequently, the original time series was reconstructed by removing the stochastic noise37. Singular spectrum
analysis can also detect non-linearity in the time series. SSA was implemented using the R package R SSA37.
Log(µi ) = β0 + υi + νi
συ2
Si2 =
�= N(i)
where β0 is the intercept.υi is a structured spatial component assuming Besag-York-Mollie (BYM) specification42,
modelled using an intrinsic conditional autoregressive structure (iCAR).
= N(i) is the number of districts that share boundaries with the i-th district, and νi is the unstructured
spatial effect in each district, modelled using an exchangeable prior νi ~ Normal (0, συ2). iCAR is based on a set
of districts that share boundaries for which an adjacency matrix was defined, listing for each district, all other
districts with which it shares a boundary or adjacency. Weights are defined for those adjacencies, and have a
value of one when two districts share a boundary and a value of zero when they do not.
District level spatial models were fitted using Bayesian Poisson generalised linear model for each zone imple-
mented in INLA-R43. Three different models were implemented namely BYM, Besag and iid model for each zone
separately. Deviance Information Criteria (DIC)44 was used to compare different models. All the maps were
prepared using the free and open source QGIS45.
Spatio‑temporal model
The annual number of outbreaks across each state was fitted using space–time Bayesian Poisson Generalised lin-
ear model accounting for spatial and temporal autocorrelation to understand the spatial and temporal dynamics
of HS outbreaks across states. The models were implemented in INLA-R43. In addition, a model was fitted with
IC44 was used to compare different models.
space–time type-1 interaction. In type-1 interaction the D
Vol:.(1234567890)
www.nature.com/scientificreports/
Log(µit ) = β0 + εi + Si + γt + St
jǫN(i)εj
mi =
�= N
σε2
Si2 =
�= N(i)
p
γt = ρj γt−1 + Zt
j=1
Zt ∼ N(0, σ 2 γ )
τt
= 1 − Normal(mt , st2 )
τj t
jǫN(t) τj
mt =
�= N
στ2
st2 =
�= N(t)
where β0 is the intercept.εi is the structured spatial component assuming Besag-York-Mollie (BYM)
specification42, modelled using an intrinsic autoregressive structure (iCAR). = N(i) is the number of states that
share boundaries with the i-th one (i.e. its neighbouring state). si is the unstructured spatial effect in each state
modelled using an exchangeable prior, si ~ Normal (0, σε2). γ t is the structured temporal component assum-
ing (i) an AR (1) structure y y47; (ii) stationarity of the data over time and (iii) all the observations are regularly
spaced in time. τ t is the unstructured temporal heterogeneity in each district modelled using st ∼ N(0, σ 2 γ ).
yit denotes the number of HS outbreaks occurring in year t t(t = 1,….T) in each state i (i = 1,….I). It is assumed
that the number of outbreaks y it for each state i, in year t, has a Poisson distribution with parameters µit and
probability πit with a log link, where linear predictor µit decomposes additively into time and space dependent
effects. Space–time interaction term was added to the model to fit type-1 interaction model.
Data availability
The Haemorrhagic septicaemia outbreaks data are provided in materials and methods in the form of time series
plots, proportion plots and spatial distribution maps. Prior permission from Director, ICAR-NIVEDI is required
for utilising the data for any purpose.
References
1. Ali, J. Livestock sector development and implications for rural poverty alleviation in India. Livest. Res. Rural. Dev. 19, 1–15 (2007).
2. Ahuja, V., Rajasekhar, M. & Raju, R. Animal health for poverty alleviation: A review of key issues for India. Background Paper
Prepared for “Livestock Sector Review” of the World Bank (2008).
3. Carter, G. Studies on Pasteurella multocida. I. A hemagglutination test for the identification of serological types. Am. J. Vet. Res.
16, 481–484 (1955).
4. Carter, G. Pasteurellosis: Pasteurella multocida and Pasteurella hemolytica. Adv. Vet. Sci. 11, 321–379 (1967).
5. Heddleston, K., Gallagher, J. & Rebers, P. Fowl cholera: Gel diffusion precipitin test for serotyping Pasteurella multocida from avian
species. Avian Dis. 16, 925–936 (1972).
6. Shivachandra, S.B., Viswas, K. & Kumar, A. A. A review of hemorrhagic septicemia in cattle and buffalo. Anim. Health Res. Rev.
12, 67–82 (2011).
7. De Alwis, M. C. Haemorrhagic septicaemia (Australian Centre for International Agricultural Research Canberra, ACT, 1999).
8. WOAH. Haemorrhagic septicaemia (Pasteurella multocida serotypes 6:B and 6:E). Chapter 3.4.10. Terrestrial Manual of diagnostics
tests and vaccines for Terrestrial animals, 12th Eiditon. 739–750 (World Organization for Animal Health (WOAH), 2023).
9. Verma, R. & Jaiswal, T. Haemorrhagic septicaemia vaccines. Vaccine 16, 1184–1192 (1998).
10. Benkirane, A. & De Alwis, M. Haemorrhagic septicaemia, its significance, prevention and control in Asia. Veterinarni Medicina-
Praha 47, 234–240 (2002).
11. Singh, M. K. & Manoj, J. Therapeutic management of Haemorrhagic septicaemia in buffalo heifers. Intas Polivet 19, 45–46 (2018).
12. Mushtaq, A. et al. Haemorrhagic septicemia: A persistent nuisance in Indian livestock review. Pharma Innov. J. 11(7S), 4382–4394
(2022).
Vol.:(0123456789)
www.nature.com/scientificreports/
13. Jindal, N. et al. Some epidemiological observations on haemorrhagic septicaemia in buffaloes and cattle in Haryana state of India.
Buffalo J. 18, 273–280 (2002).
14. Bardhan, D. et al. Economic losses due to hemorrhagic septicaemia in India. Indian J. Anim. Sci. 90(3), 341–346 (2020).
15. Kawasaki, M., Young, J., Suon, S., Bush, R. & Windsor, P. The socioeconomic impacts of clinically diagnosed haemorrhagic sep-
ticaemia on smallholder large ruminant farmers in Cambodia. Transbound. Emerg. Dis. 62, 535–548 (2015).
16. Singh, B., Prasad, S., Verma, M. R. & Sinha, D. Estimation of economic losses due to Haemorrhagic septicaemia in cattle and buf-
faloes in India. Agric. Econ. Res. Rev. 27, 271 (2014).
17. Govindaraj, G., Krishnamoorthy, P., Nethrayini, K., Shalini, R. & Rahman, H. Epidemiological features and financial loss due to
clinically diagnosed Haemorrhagic Septicemia in bovines in Karnataka, India. Prev. Vet. Med. 144, 123–133 (2017).
18. Govindaraj, G. et al. Farm community impacts of foot-and-mouth disease outbreaks in cattle and buffaloes in Karnataka state,
India. Transbound. Emerg. Dis. 64(3), 849–860 (2017).
19. Singh, G., Chilambarasan, M. & Singh, R. K. Estimation of Economic losses due to Haemorrhagic septicaemia in Haryana. (2017).
20. Sivakumar, T., Thennarasu, A. & Rajkumar, J. Effect of season on the incidence of infectious diseases of bovine in Tamilnadu. Elixir
Meteorol. 47, 8874–8875 (2012).
21. Verma, S. C., Mahajan, N., Malik, G. & Dahiya, J. An epidemiological study on bovine haemorrhagic septicaemia in Haryana.
Prevalence 38, 14–19 (2004).
22. Sudhindra, Y. P. & Rajasekhar, M. Haemorrhagic septicaemia and black quarter forecasting models for India. Epidemic!. sante
anim 1997, 31–32 (1997).
23. Kumar, H. C. et al. Animal disease surveillance: Its importance & present status in India. Indian J. Med. Res. 153(3), 299 (2021).
24. Prajapati, A. et al. Comparative genome analysis of Pasteurella multocida serogroup B: 2 strains causing haemorrhagic septicaemia
(HS) in bovines. Gene 826, 146452 (2022).
25. Prajapati, A. et al. Comparative genetic diversity analysis based on virulence and repetitive genes profiling of circulating Pasteurella
multocida isolates from animal hosts. Infect. Genet. Evol. 85, 104564 (2020).
26. Shome, R., Deka, R. P., Sahay, S., Grace, D. & Lindahl, J. F. Seroprevalence of hemorrhagic septicemia in dairy cows in Assam,
India. Infect. Ecol. Epidemiol. 9, 1604064 (2019).
27. Ma, J. et al. Spatial pattern of foot-and-mouth disease in animals in China, 2010–2016. PeerJ 5, e4193 (2017).
28. Redding, D. W., Tiedt, S., Iacono, G. L., Bett, B. & Jones, K. E. Spatial, seasonal and climatic predictive models of Rift Valley fever
disease across Africa. Philos. Trans. R. Soc. B 372, 20160165 (2017).
29. Martínez-López, B., Alexandrov, T., Mur, L., Sánchez-Vizcaíno, F. & Sánchez-Vizcaíno, J. M. Evaluation of the spatial patterns and
risk factors, including backyard pigs, for classical swine fever occurrence in Bulgaria using a Bayesian model. Geospatial Health
8, 489–501 (2014).
30. Zhang, J. et al. Spatial analysis on human brucellosis incidence in mainland China: 2004–2010. BMJ Open 4(4), e004470 (2014).
31. Herholz, C., Jemmi, T., Stärk, K. & Griot, C. Patterns of animal diseases and their control. Vet. Ital. 42, 295–303 (2006).
32. Stojmanovski, Z. & Tabakovski, B. Spatio-temporal characteristics of the bluetongue epizooty in the Balkan peninsula from 2014
to February 2015. Maced. Vet. Rev. 41, 65–72 (2018).
33. Cassirer, E. F. et al. Spatio-temporal dynamics of pneumonia in bighorn sheep. J. Anim. Ecol. 82(3), 518–528 (2013).
34. Bezymennyi, M. et al. Spatio-temporal patterns of livestock anthrax in Ukraine during the past century (1913–2012). Appl. Geogr.
54, 129–138 (2014).
35. Ward, M. P. & Carpenter, T. E. Analysis of time–space clustering in veterinary epidemiology. Prev. Vet. Med. 43, 225–237 (2000).
36. Kumar, A. A. & Shivachandra, S. B. Action plan for control of haemorrhagic septicaemia and Pasteurellosis. In Chapter 18, In
Veterinary and Livestock Sector: A Blueprint for Capacity Building, 383–429 (SS Publishing House, 2012).
37. Golyandina, N. & Korobeynikov, A. Basic singular spectrum analysis and forecasting with R. Comput. Stat. Data Anal. 71, 934–954
(2014).
38. Cazelles, B., Chavez, M., De Magny, G. C., Guégan, J.-F. & Hales, S. Time-dependent spectral analysis of epidemiological time-
series with wavelets. J. R. Soc. Interface 4, 625–636 (2007).
39. Grinsted, A., Moore, J. C. & Jevrejeva, S. Application of the cross wavelet transform and wavelet coherence to geophysical time
series. Nonlinear Process. Geophys. 11, 561–566 (2004).
40. Gouhier, T. C., Grinsted, A. & Simko, V. R package “biwavelet”: Conduct univariate and bivariate wavelet analyses. Version 0.20.
15. (2020).
41. Muggeo, V. M. Segmented: An R package to fit regression models with broken-line relationships. R News 8(1), 20–25 (2008).
42. Besag, J. Rejoinder. Ann. Inst. Stat. Math. 43, 45–59 (1991).
43. Blangiardo, M., Cameletti, M., Baio, G. & Rue, H. Spatial and spatio-temporal models with R-INLA. Spatial Spatio-temporal
Epidemiol. 4, 33–49 (2013).
44. Spiegelhalter, D. J., Best, N. G., Carlin, B. P. & Van Der Linde, A. Bayesian measures of model complexity and fit. J. R. Stat. Soc.
Ser. B (Stat. Methodol.) 64, 583–639 (2002).
45. QGIS Development Team, QGIS Geographic Information System (Open Source Geospatial Foundation, 2022) http://qgis.org.
46. Pebesma, E. spacetime: Spatio-temporal data in R. J. Stat. Softw. 51, 1–30 (2012).
47. Chatfield, C. The Analysis of Time Series: An Introduction (CRC Press, 2016).
Acknowledgements
The authors wish to thank the Indian Council of Agricultural Research (ICAR), New Delhi, India, for constant
support and encouragement to carry out the work. Authors would like to sincerely acknowledge past and present
in charge of the NADRES database for collating data through AICRP-ADMAS (All India Coordinated Research
Project on Animal Disease Monitoring and Surveillance). Authors acknowledge all the previous Directors who
were involved in maintaining the NADRES database from time to time. Authors would like to sincerely thank
Director, ICAR-NIVEDI for providing necessary facilities, infrastructure and support to conduct this study.
MMC, SBS, DH, SSP, YA, AP were supported by ICAR-NIVEDI and to BVP from NERC under the SUNRISE
project (Grant no. NE/R000131/1).
Author contributions
M.M.C. and S.B.S. conceived and planned the study, with input from B.V.P., D.H., S.S., R.Y. and A.P. M.M.C.
collated and processed Haemorrhagic septicaemia outbreaks data. M.M.C. performed all the analyses and pro-
duced output figures. All authors contributed to the interpretation of results. M.M.C. wrote the first draft of the
paper and all authors contributed to subsequent revisions.
Vol:.(1234567890)
www.nature.com/scientificreports/
Funding
The study was supported by institute project on “Epidemiology of Haemorrhagic Septicaemia (HS) in India”
funded by ICAR-NIVEDI, Bengaluru, Karnataka, India.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https://doi.org/
10.1038/s41598-024-56213-z.
Correspondence and requests for materials should be addressed to M.M.C.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Vol.:(0123456789)