Ideas & Insights / All Perspectives / Ideas & Insights

Anticipating Humanitarian Crises in the Age of Covid-19

Evan Tachovsky — Former Director & Lead Data Scientist, Innovation, The Rockefeller Foundation
Megan Linquiti — Former Summer Fellow, Innovation, The Rockefeller Foundation

Before Covid-19 struck, nearly 168 million people worldwide needed humanitarian assistance to survive. The pandemic has increased the number of vulnerable people worldwide and has compounded risk for countries already experiencing food shortages, conflicts, and economic hardship. Predictive models can help us anticipate the trajectory of these compounding crises and speed response.

To accelerate predictive modeling for anticipatory response in the age of Covid-19, we are pleased to support the United Nations Office for the Coordination of Humanitarian Affairs (OCHA) on a new grant. This project aims to use predictive analytics to improve response to both the Covid-19 pandemic and co-occurring crises around the world. Using data science to inform and improve humanitarian responses can ensure that resources to battle the pandemic are used effectively and will help to anticipate and mitigate a cascade of human suffering.

OCHA’s Centre for Humanitarian Data (the Centre) has been at the forefront of using data to improve humanitarian response. Through their work on the Humanitarian Data Exchange and predictive analytics, the Centre is charting a new course for the humanitarian sector. They are focused on using the best available data to make predictions about future needs in order to mitigate the impact of a predicted shock rather than reacting to a fully evolved crisis. As the gap between needed and available funding for humanitarian work grows, getting ahead of crises is especially important.

Using data science to inform and improve humanitarian responses can ensure that resources to battle the pandemic are used effectively and will help to anticipate and mitigate a cascade of human suffering.

With our grant support, the Centre will build on its strong foundation to complete three activities: build new predictive models to improve crisis response, implement a rapid peer review framework to vet predictive models under development, and identify and fill data gaps to improve modeling efforts long-term.

  1. Support for predictive modeling. OCHA will provide analytical support to country offices that do not have the capacity to build predictive models for pandemic and humanitarian response. By pairing data scientists with OCHA country representatives, they will help specify models for geographies with co-occurring humanitarian crises. They will also support new efforts to model the impact of Covid-19 on humanitarian operations including in Afghanistan, the Democratic Republic of the Congo, Sudan and South Sudan, among other locations.
  2. Creating a process for evaluating predictive models. Across the humanitarian sector, there are many models used to predict crises and estimate humanitarian impact. In order to identify the most reliable models, OCHA will implement a rapid peer review process pioneered by the Centre. This process leverages best practices in model evaluation to identify and document model strengths and weaknesses. The rapid peer review framework will improve promising models and protect populations and decision makers from subpar models.
  3. Filling gaps in available data. To build high-quality models data scientists need high-quality data. Under this grant, OCHA will leverage its existing work on the Humanitarian Data Exchange to identify gaps in existing data for humanitarian response and target outreach to fill those gaps. They will also publish an annual report – The State of Open Humanitarian Data – to track the progress of these efforts which can be used to bring more data contributors into the project.

Accurate predictive models can help improve and direct humanitarian assistance in response to a variety of different crises. The Covid-19 pandemic highlights one clear case where anticipatory action is necessary. There is more work to be done to address both the pandemic and co-occurring humanitarian crises around the world. By supporting the development of predictive models, implementing a system to evaluate new models, and working to fill data gaps, OCHA will improve the way we prepare for and mitigate crises.

While this work represents a strong start, scaling predictive modeling to improve humanitarian response will require additional partnership. If you’re a data provider, modeling team, or funder interesting in contributing to the long-term success of this effort, please reach out. Together we can build the technical community we need to anticipate crises and act to save lives and prevent human suffering.

Photo credit: Linda Tom/OCHA Afghanistan. The Islam Qala border crossing point between Afghanistan and Hirat province in March 2020.