PAW presentation on engaging stakeholders and community leaders to impact Ebola outbreaks. It includes some details about modeling that I'd used, as well as sociopolitical contexts prior to the outbreaks, which played a large role in the success of efforts.
2. Who am I?
Social scientist who started
the data science journey as
an MD/PhD student
Data scientist and machine
learning researcher for 10+
years now
Entrepreneur within
healthcare and applied
physics
Author of scientific articles,
lay audience articles, and a
forthcoming machine learning
book
3. What is Ebola?
Caused by a single-stranded RNA filovirus
Transmitted through direct contact with bodily fluids, particularly at later
stages of infection
Produces severe symptoms:
Starts with sudden flu-like symptoms
Progresses to high fever with severe diarrhea and vomiting
Some with symptoms of chest pain or trouble breathing in the later stages
Some with severe bleeding in the later stages
Death from hypovolemia for 50%+ of patients
Ebola is mostly survivable with sufficient medical resources per patient,
though.
4. 2014 West African Outbreak
March 2014 outbreak in Guinea
traced to a single case in December
2013 (death of a one-year-old)
Quickly spread to Sierra Leone and
Liberia
Isolated cases spread to other
countries, prompting international
panic.
Outbreak was officially declared
under control in March 2016.
As of May 2016, the death toll stood
at 11,323.
5. Sociopolitical Context
Guinea had endured periods of unrest internally in the years preceding the
outbreak but been connected to other African nations and outside investors;
Guinea is a mostly Muslim country, following Islamic tradition with respect to
burials.
Liberia, a mostly Christian country, is Africa’s oldest republic, and following
unrest in the early 2000s, Liberia had returned to stability, albeit at high
poverty rates.
Sierra Leone, a mixed Muslim and Christian nation, weathered a lengthy civil
war and path to peace prior to the epidemic; however, women remain
vulnerable in Sierra Leone.
Poverty, religious burial practices, and gender inequality created disparities in
vulnerability and response to the epidemic.
6. Engaging Partners
Within Sierra Leone, Liberia, and Guinea:
Large NGOs to send personnel
Meeting supply provision
Addressing cultural practices
Within countries bordering the most-impacted countries:
Drivers were:
Fear of Ebola spread
Impacts on economy and population health
My involvement started with an entrepreneur friend in Mali, who then engaged the
health ministry and local NGOs.
7. Methodology
Methods
Simple scenario models
Run with an Excel macro
Using a susceptible-infected-
recovered model
At both local and country levels
This gave:
Locations most at-risk for initial case
in country
Locations leading to the biggest
epidemic potential
Transportation risk for initial case
entry
8. Response
Models were able to pinpoint most
vulnerable targets within Mali (including
Kayes, where the first case crossed into
Mali), prompting:
Diversion of resources and international aid
Movement of personnel to vulnerable areas
Phone lines to disseminate information to
illiterate populations
In all, very few people were infected in
Mali after two incidents of sick travelers
occurred.
9. 2018 DRC Outbreak
Current outbreak in the
Democratic Republic of Congo
traced to a cluster of four cases in
August 2018.
Quickly spread across North Kivu
and Ituri provinces
Isolated cases in South Kivu and
Uganda
As of May 2020, the outbreak is
ongoing, with new cases and
deaths reported.
New outbreak declared June 2020
10. Sociopolitical Context
The DRC is a former Belgian colony that has seen a lot of exploitation and war
for the last 150 years.
The Kivu Conflict has been ongoing since 2004 in the affected area.
Funded by exploiting natural resources like cobalt, gold, and diamonds
Includes Congolese rebels and exiled Hutus who had fled to the DRC after the
genocide in Rwanda
Has involved child soldiers, human trafficking, and rape of civilians (estimated to
be 400,000+/year in the impacted region)
Attacks within the past days when the outbreak started
Death toll estimates in the millions in this conflict alone
11. Engaging Partners
International aid workers were already in the region when the outbreak
started.
Vaccines had already been developed in response to the 2014 outbreak and have
been tested in the DRC outbreak.
Most of the international coverage has shown attacks on humanitarian workers
trying to contain the outbreak.
Given the proximity to US interests, military and humanitarian groups have
been mobilized.
One disaster response organization contacted Quantopo regarding predictive
modeling for resource planning purposes (Feb 2019).
Action has been complicated by the ongoing Kivu conflict.
12. Methodology
Methods:
Data from Humanitarian Exchange
Geographic data on travel routes between cities
in Ituri and Kivu
Topology-based methods along with traditional
forecast models
Curvature-based analyses to forecast areas at
risk for major increases in case loads and attacks
against aid workers
Calculated using custom R code
This gave:
Cities at risk for large
increases in numbers of
cases, including Katwa,
Beni, and Goma
Identification of violence-
prone regions (including
Katwa and Beni, where
most attacks have
happened)
13. Response
Within two weeks of our results, attacks picked up in Katwa and Beni, forcing
most international aid organizations to pull out remaining personnel in the
region.
This has complicated the response to the current outbreak.
Numbers of new cases doubled between April and June of 2019, within weeks our
analysis.
Since then, cases have begun appearing in major cities and across borders,
including cases in Goma and Uganda.
The United Nations and World Health Organization have declared this an
international emergency.
So far, cities and bordering countries have been relatively unaffected, and
efforts have focused on limiting the spread of the outbreak.
14. Complicating Issues
The conflict is ongoing, forcing aid
workers to evacuate.
There’s widespread misinformation
about the source of the outbreak
(biological warfare from white aid
workers…).
New cases are popping up with no
known contact with an infected person.
There was a major scare in Tanzania
regarding potential travelers infected
with Ebola. The government has not
been forthcoming with test results.
This outbreak is complex, and
containment strategies that worked
well in West Africa are not working in
the DRC.
15. Current Situation in the USA
Pandemic, deep politics-based mistrust, violent clashes in major cities
16. Lessons: Build Bridges
Need to predict at-risk areas not
impacted yet and areas at risk for
epidemic worsening
Need for ongoing international
response in conjunction with local
governments to meet the needs of
local populations at risk
Need for infrastructure to support
aid workers (armed protection…)
Need to disseminate accurate
information to local populations
with limited technological access
and education