Integrative Approaches to Disease

Modelling

Overview and introduction to Dynamic

Drivers of Disease in Africa ConsortiumDelia Grace, Pete Atkinson, Gianni Lo Iacono, Johanna Lindahl and Catherine Grant

Context

Emerging zoonotic disease events 1940-2012

Overall hypothesis

Disease regulation as an ecosystem

service is affected by changes in

biodiversity, climate and land use, with

differential impacts on people’s health and

wellbeing.

Interdisciplinary science

Untangling interactions

Conceptual framework

Country case studies

• Kenya: Rift Valley

fever

• Zambia and

Zimbabwe:

trypanosomiasis

Country case studies

• Ghana: henipavirus

• Sierra Leone: Lassa fever

Why these diseases?

• Commonalities

– Poverty impacts: direct, indirect and potential

– Often under-recognised and under-reported

– Require animal hosts to sustain infection in human populations

• Comparisons

– Different ecosystems: humid, semi-arid, arid

– Different hosts and transmission pathways

– Different political-economic and social drivers

Big drivers, big impacts, big

questions

• Urbanisation

• Irrigation

• Climate change

• Population movement

• Conflict

• Wildlife-livestock interaction

• Commercial farming

Disease

Dynamics

Demography &

Development

Disease management

Integrated policy interventions

Surveillance approaches

Capacity building

Practical and policy impacts

How different modelling approaches

contribute to answers

• Epidemiology and disease burden- Johanna Lindahl

• Process-based modelling for RVF- Gianni Lo Iacono

• Agent based modelling- Pete Atkinson

• Integration of participatory research- Catherine Grant

Consortium partners

• ESRC STEPS Centre, UK

• University of Cambridge, UK

• Institute of Zoology, UK

• University of Edinburgh, UK

• University College London (UCL), UK

• Wildlife Division of the Forestry Commission, Ghana

• University of Ghana, Ghana

• Department of Veterinary Services, Kenya

• International Livestock Research Institute (ILRI), Kenya

• Kenya Medical Research Institute (KEMRI), Kenya

• University of Nairobi, Kenya

• Kenema Government Hospital, Sierra Leone

• Njala University, Sierra Leone

• Ministry of Livestock and Fisheries Development, Zambia

• University of Zambia, Zambia

• Ministry of Agriculture, Mechanisation and Irrigation Development, Zimbabwe

• University of Zimbabwe, Zimbabwe

• Stockholm Resilience Centre, Sweden

• Tulane University, USA

• University of Southampton, UK

Dynamic Drivers of Disease in AfricaIntegrating our understandings of zoonoses, ecosystems and wellbeing

Epidemiology and disease burdenJohanna Lindahl, Alexandra Shaw, Delia Grace

Epidemiology

• The patterns, causes, and effects of health

and disease conditions in populations

• For many diseases we lack knowledge

• Not much research

• Much research but still not understanding

• Lack information on how to prevent

S E I R

New population at risk

Global trade and travelling

Increasedcontact with wildlife

Close contactbetweendifferent species

Transfer or recruitment of new vectors

New habits, new cultures

Migration of people or animals to new areas

New species at risk / host transfer

Decreasedimmunizationand immunity

Markets

Urbanization

Environmentalland degradation

Poverty

Undernutrition, starvation

Governmentalfinances and priorities

Ageingpopulation

Compartmental epidemiological models

Increased risk of exposure

Habitat fragmentation

Decreasedbiodiversity

Increased number of vectors

High density

Lack of knowledge

Less dilution from alternate hosts

Reducedfood safety

Water scarcity

Disrupted social systems

Poverty

Urbanization

Markets

Industrialization of animal productionLittering

Irrigation

Fertilisers

Deforestation

Agricultural intensification and development

Climatechanges

S E I R

Anthropogenic action:

Increased irrigation

Effect on ecosystem:

Creates more larval habitats

Vector consequence:

More infected vectors

Epidemiologicconsequence:

More individualsexposed

Increased

disease

Most drivers are desired- and not

constantly leading to disease

A framework to help understand costs

and to model costs

• Aim is to collect the data necessary to make

an assessment of the multiple burden of

disease

• Using the same framework for multiple

diseases helps comparison

• Economic modelling important for policy

makers

• Money matters

• Priorities

Framework for assessing the

economic costs and burdens of

zoonotic diseaseAlexandra Shaw, Ian Scoones, Melissa Leach, Francis

Wanyoike and Delia Grace

• Zoonoses sicken 2.4 billion people,

kill 2.2 million people and affect

more than 1 in 7 livestock each year

• Cost $9 billion in lost productivity;

$25 billion in animal mortality;

and$50 billion in human health

Costs of zoonotic disease

Benefits of controlling zoonoses in animals and

along the value chain chain

• Credible economic cost benefit studies

(n=13)

– Average benefit cost ratio 6:1

– Median 4:1

– Range 1.1-19.8

• Implies $85 billion losses could be

averted by $21 billion expenditure21

Ex ante 5Ex post 6.6

Developing countries 3.7Developed countries 7.4

How can we model disease burden?

What do we include in the burden of disease?

• Disability-adjusted life years (DALYs)

• Economic impact

• Society/nation

• Personal

• Environmental impact?

Zoonoses have multiple burdens

• Disease in humans

• Economic consequences of disease in

humans

• For people and society

• Loss of incomes, and costs for treatments

• Disease in animals

• Economic consequences of disease in

animals

• For people and society

• Lost production, trade bans

2 trade offs

1. Between disease control expenditure and

illness in humans and animals

2. Between ecosystem change and disease

incidence

Missed work/school

Less income, lower

education

Reduced living

standards/ reduced nutrition

Increased exposure to pathogen/ reduced immune defense

Disease

The vicious cycle- for people

How can we model disease burden?

• Simplified situation

• Assessing what we can assess

• Direct economic impacts

• Collecting more data on what we don’t know

• Creating a Framework for assessing

economic costs and burdens of zoonotic

disease

2 aspects of costs of disease

1. who pays (public or private sector)?

2. how easy is it to quantify them? (availability

of information and applicability /availability of

market prices).

The multiple burdens of zoonotic

disease: human, animal and

ecosystem health

Actors Cost of Illness Cost of preventionIntangible and

opportunity costs

Private

Individual and household

(1) Treatment costs

(e.g. medication)

(2) Loss of household

production

(1) Risk mitigation

such as boiling

water, buying filters

(1) Disutility of ill

health for individual

(DALY)

(2) Disutility of ill

health for friends,

family, etc.*

Livestock sector

(1) Cost of treatment,

(2) Herd slaughter,

product recall,

mortality,

morbidity, lower

production, loss of

exports

(1) Costs of increased

biosecurity,

(2) vaccination,

practices and

procedures to

control disease

along the value

chain

(1) Cost of future

emerging

diseases*

(2) Loss of animal

genetic resources.

*

Public

Health (human and

animal)

(1) Treatment costs

(hospital provision,

etc.)

(2) outbreak costs,

movement

restrictions, culling,

(3) vaccination

(1) Risk mitigation

such as water

fluoridation,

vaccination

(2) (Disease

surveillance,

research)

(1) Loss of

opportunities

occasioned by

spending on

disease prevention

and care*a

Ecosystem

(1) Spill-over into

wildlife,

(2) loss of ecosystem

services

(1) Bio-security,

avoiding wildlife

and vectors,

(2) disease

surveillance,

research

Included

in

DALYs

The cost of illness and burden of

disease in people- how to measure

Information needed Type of data Possible existing sources Further investigations

Reported cases of disease Record of individuals diagnosed

with disease

Hospital and clinic records, national

and provincial health statistics

May be worth visiting local hospitals

and clinics to collect data if it is not

summarised at national level

Estimate of extent of under-

reporting

Compare recorded cases with

number actually found

Published/grey literature (PGL)

studies or investigations

If field work involves testing people,

or finding people with the disease

then the prevalence or incidence

can be compared to that reported.

Often test high risk groups (people

with fevers not responding to

malaria, people working/living in

close contact with relevant animals)

Burden of disease in affected

individuals

(Valued as Disability-adjusted life

years – DALYs)

Deaths Hospital and clinic records, PGL

data on death rates and DALYs –

the years of life lost (YLL)

component

Visit local hospitals and clinics to

collect data, ask about it in

household interviews

Disability PGL studies and interviews and

DALY estimates, including relevant

disability weights – the years of life

lived with disability (YLD)

component of the DALY

Interview patients and families to

find out about length of illness and

extent of disability.

Impact on household incomes

while person is ill

Estimated loss of household income

generated by the patient during their

illness

PGL studies Interview patients and families

The cost of illness and burden of

disease in animals- how to measure

Information needed Type of data Possible existing sources Further investigations

Reported cases (incidence) of

the disease over a certain

period or prevalence (number or

percentage with the disease at a

given point in time)

Record of animals thought to have

the disease

Outbreak investigations

Incidence and prevalence

studies

Reported cases from

veterinary clinics

Other PGL studies

Animal sampling in the field (blood

tests)

Estimate of under-reporting Extrapolation to whole animal

population. Difficult because

studies focus on high incidence

events or high prevalence sub-

populations

Published/grey literature (PGL)

studies or investigations.

Local expertise

Compare results from sampling

with other, pre-existing, estimates

Burden of disease in affected

animals

(Monetary values)

Mortality PGL studies looking at individual

diseases. Sometimes records

from vet clinics and national

veterinary statistics. For many

animal diseases, the only impact

that is recorded is deaths.

Focus group discussions.

Livestock keeper surveys.

Morbidity (lowered productivity) PGL field-based studies

comparing healthy and infected

animals. There aren’t many!

Estimate and value disease impact

on fertility, output (milk, wool,

animal traction, etc.), slaughter

rates and weights (meat), etc.

Note that livestock keepers

reactions (cull sick animals) form

part of the impact.

Livestock keeper and dog-owner

surveys. These are time-

consuming and obtaining a

suitable control group to estimate

impact is difficult. Studying wildlife

and companion animals is even

trickier.

The cost of treatment and control in

people- how to measureInformation needed Type of data Possible existing sources Further investigations

Private costs for treatment and

hospitalisation

Health care seeking costs

(often very high for these

uncommon conditions)

Time spent by family

looking after patient at

home and when looking

for care of being treated

Patient expenditure on

correct and incorrect

medication and

diagnostics

Local clinics and medical

practitioners, hospitals

PGL studies

Patient and patient family

interviews.

Public costs for treatment and

hospitalisation

Cost of hospitalisation,

operations, drugs,

diagnostic

Ministry of Health, hospital

and clinic data

Interviews with care staff in

specialist units

Private costs for disease control Patient and other

members of the public -

costs for vaccination,

quarantine, any other

disease prevention or

mitigation measures

Local clinics and medical

practitioners, hospitals

PGL studies

Patient and patient family

interviews. Interviews with

target populations (e.g. of

vaccination campaigns)

Public costs for disease control Cost of surveillance

Costs of vaccination

Ministry of Health, hospital

and clinic data

Interviews with staff involved

in this work

Costs of prevention-

Humans and animals

Mosquito

nets

Vaccines & routine clinic

visits for kids

Boiling or other water

treatment

Insurance

(annual fee)

Other health

prevention

Mean 762 254 6.8 0.9 586

Range 0-3150 0-5000 4 households paid

between 150-600

220 households

paid nothing, one

household paid

200

0-6000

How much did you spend last year on the following health protection (Kenyan shilling)?

Deworming Vaccinations (to

prevent not to

treat)

Tick and fly

treatments

Insurance

(annual fee)

Mean 928 437 599 0

Range 0-11000 0-5000 0-5000 Not existing

How much did you spend last year on the following health prevention for animals?

221 Kenyan households interviewed

Sharing resources for health delivery

• Efficiency & effectiveness gains

– Shared infrastructure; training, services• Joined up services for zoonoses: Across a range of studies 5-15% reduction

in costs +/or improvement cover

• World Bank (2012) estimates 25% savings across a range of joint services

for AI and 7% additional costs = net savings of 18%

• Developing country health sector expenditure: 250 billion

• Developing country veterinary expenditure: 2 billion

– Amenable to joined up services: $4 billion

33

Increase in

peopleIncrease in

livestock

Yes

No Probably low risk

of increased

disease incidence

No

Yes

Sufficient medical

care and

infrastructure?

Are circulating

diseases

known?Vector control

programs

used?

Appropriate

sanitation?

Increase in

vectors

Can they be

prevented

or cured?

High risk of increased

incidence of vector-

borne, rodent-borne,

water and food-borne

diseases

NoNo

No

Yes

Yes

YesYes

Surveillance

No

No

No

High risk of increased

disease

Yes

Yes No

Too late, but

good anyway

Political decisions and economy

Wildlife

interface

Yes

In conclusion

• We need to show the multiple burdens of

disease

• We need to show the money savings

• We need to show economical consequences

Because money talks.

Integrative Approaches to Disease

Modelling

Agent Based ModellingPete Atkinson, University of Southampton, UK

Modelling background

• Epidemiological models are traditionally created using dynamic,

compartmentalised approaches…

• Sleeping sickness is represented by the – –

(SIS) model due to the absence of immunity.

38

Population sizes

Infection rate (contact rate)

Space (?!)

The theoretical number of

people in each compartment

at a given time.

Homogeneous mixing.

Differential equations.

Agent-based model

• Models the movements of individual agents:

– Humans

– Animals

– Vectors

• Need to know:

– Landscape

– Agents

– Rules

39

A simple model for Trypanosomiasis

• Only one previous ABM of Sleeping Sickness

• Spatially abstract simulation backdrop represents

– a river (blue), with banks (green) and pasture

• Three agents:

– Human, Cow, Tsetse fly

• Humans are divided into

– cattle farmers and non-farmers

• Black icons represent home settlements

40

1st iteration of the model

Tow

n

HumanCattl

e

Infected

Fly

Uninfected Fly

River and

banks

42

2nd iteration of the model

• Short video simulating one day of real

time

• As before (orange),

(blue), (green)

• Walking speed = 5 km hr-1

• N.B. Frequency of trips to water

increased for demonstration

PRM

Acknowledgements

Neil Anderson

Joanna Kuleszo

Simon Alderton

Kathrin Schaten

Noreen Machilla

Alex Shaw

Thank youpma@soton.ac.uk

Dynamic Drivers of Disease in AfricaIntegrating our understandings of zoonoses, ecosystems and wellbeing

Integration of Participatory ResearchProfessor Peter Atkinson, Dr Gianni Lo Iacono, Catherine Grant, Dr Bernard

Bett, Professor Vupenyu Dzingirai, Tom Winnebah and other members of the

Dynamic Drivers of Disease in Africa Consortium

Our conceptual framework

• Models can provide characterisations and predictions to

advance knowledge and evidence for policy but often they

are constructed by single disciplines representing a

selective view of the world.

• Researchers can be influenced by perspective and the

political and funding arena and, often not considering views

of those actually living with the disease.

• Infectious diseases need to be studied using a

multidisciplinary perspective, including involving local

people to potentially improve model selection and accuracy.

Our rationale for integration of

participatory work

• Explain the benefits to using participatory approaches to

improve model design and facilitating multidisciplinary

research in this area- overcoming disciplinary hurdles

• Proposing practical examples of effective integration

• Models can create tangible information from uncertainty

which leads them to be given an authority which may be

unjustified in a decision-making or policy context.

• This work aims to make models and their predictions more

useful for decision-making and policy formulation and

include information such as predicted behavior change.

Aims of our work

Participatory work in action

The benefits of participatory research

1. Removal of ignorance

2. Confirmation

3. Removal of irrelevance

4. Addition of knowledge

5. Removal of error

Acknowledgement: Pete Atkinson

Participatory research as a tool to:

1. Structure a model: population-based mathematical modelling

2. Structure a model: geographically explicit ABM (previous

presentation)

3. Select the most relevant parameters of the system

4. Identify the most relevant regime of the system

5. Mathematical modelling as a tool to better structure participatory

research

6. Diversity of modelling approaches challenge the conclusions of

other types of modelling

Application of this to our case studies

1. A tool to structure a model: population-based

mathematical modelling

Examples from Sierra Leone

• Provide information on patterns of mobility- increasing

model accuracy

• Provide new data on seasonal activities- allowing the

inclusion of a periodically varying rate of contact with

humans

• Interpreting the reliability of hospital data e.g. seasonal

hospital attendance

Examples from Kenya

resource maps for a village

proportional piling on livestock species

kept

livelihood activities by

gender

RVF Agent Based Model (Bett et al.)

Modelling Exposure

Model Input of

relative

proportion of

hosts

Modelling Risk in

Spatial Models

Acknowledgement: Gianni Lo Iacono

Immigration of infected animals in

RVF free site

Frequency of such movements

Can the site become

endemic?

Conditions for endemicity

Acknowledgment: Gianni Lo Iacono

2. As a tool to structure a model: geographically

explicit ABM

As described in the previous presentation

3. A tool to select the most relevant parameters

of the system

Hunting Bats

Economic factors

Bushmeatculture

4. A tool to identify the most relevant regime of

the system

Participatory modelling can assist in determining whether or not a

system has reached equilibrium, identifying the possible causes

leading to a disruption of the equilibrium, and it can direct the

mathematical approach towards the relevant regime, that is, transient

regime rather than equilibrium.

For example:

• In Ghana, changing farming and hunting patterns and varying

pesiticide use, information gathered from participatory research,

shows that the environment is changing and not in equilibrium.

• In Sierra Leone land use change affects rodent habitats, affecting

population size and where they live.

5. Mathematical modelling as a tool to better

structure participatory research

Using the results from other modelling can help provide new

questions and sources of investigation for participatory research.

For example:

• Mathematical modelling found that human transmission has a

relatively high impact due to the prescence of living virus in

urine. Therefore participatory research could focus on new

areas such as hygiene and potential contact points.

• Focus on movement to inform ABM could lead participatory

research to focus on the politics of who moves where and when.

6. Diversity of modelling approaches challenges the

conclusions of other types of modelling

Reality is too complex to model in full and no model can capture

everything. Different models highlight different issues and are based

on different assumptions, world views and sources of information,

leading to different conclusions about disease risk and the

appropriate actions and policy decisions to take (Leach and Scoones

2013).

Interdisciplinary working can address these issues, embracing

multiple sources of evidence. This can lead to an enriched

interpretation of research findings, integrating perspectives from

those coming from different disciplinary outlooks, and wider-ranging

translation of research. This also means that there is more

opportunity for wider dissemination and that the integrated models

will be more useful in practice and policy.

Conclusion

• This paper shows that reality is too complex to be modelled by one

modelling approach from one discipline.

• The use of the One Health approach, working together to embrace

multiple sources of evidence, can provide more realistic models to

assist with policy decisions that reduce disease and benefit local

people.

• Participatory research, in particular, can help to explain who gets

sick, where and why as well as provide explanations for health

seeking behaviour.

• Participatory research can help illuminate new areas. It is not about

challenging other approaches, but helping provide new ways of

thinking and alternative methods.

• There is lots we don’t know and participatory research can augment

standard modelling and help us move interdisciplinary science

forward, adding nuance and complexity to already useful areas of

enquiry.

• However, there are, of course, challenges to integrating models and

data, due to researchers’ different perspectives on approaches.

Dynamic Drivers of Disease in Africa Consortium

• Introduction to DDDAC (Ecohealth/One Health context)

• Agent Based Modelling (trypanosomiasis in Zambia)

• Process-based modelling (RVF –ecological &

environmental modelling, local knowledge).

• Epidemiology and disease burden

• Interdisciplinarity and Participatory research

Overview

This work is of crucial importance globally

because…..

• More than 60% of emerging infectious diseases over the last few decades

have been zoonotic

• Zoonoses have the potential to result in global disease outbreaks (e.g.

Avian influenza)

• Many zoonoses affect disenfranchised communities, quietly decimating

poor people’s lives and livelihoods

• Diseases of poverty, including zoonoses, are often under-measured and

therefore under-prioritised in national and international health systems

• If unchecked, emerging zoonoses create dangerous future threats

• This is of particular concern for zoonotic diseases with complex

connections to a wider set of ecosystem changes, such as

land use change, habitat loss and climate change.

• One Health involves the environmental, human and animal

health sectors crossing professional, disciplinary and

institutional boundaries to work, challenging as this may be, in

a more integrated fashion.

• Zoonotic diseases provide an archetypal illustration of the utility

of the One Health approach as they are shaped by complex

interactions amongst humans, animals and the environment

and, thus, between epidemiological, ecological, social and

technological processes which affect vulnerabilities to, and

risks of, transmission influenced by wider socioeconomic and

environmental drivers (Leach and Scoones 2013).

Importance of our One Health approach…..

Thank you

For more information on our Consortium:

www.driversofdisease.org

@DDDAC_org

contact@driversofdisease.org