A series of class projects help guide life-saving public health policies in India, Peru and Botswana.
As California’s shelter-in-place order went into effect in March, Stanford professor of management science and engineering Ashish Goel’s thoughts turned both to the teachable moment it presented for his students and to the dire ramifications COVID-19 might have in his native India and other developing nations.
“In many cases these nations are densely populated and have very little testing or medical capacity,” Goel says.
And so, with less than a week remaining before the start of a new quarter, Goel – whose research and teaching usually focus on developing computer algorithms and social network applications – designed a new course to use computer modeling to help developing countries confront the pandemic. The result was MS&E 433: The Socio-Economic Impact of the Novel Coronavirus on the Developing World.
Goel quickly shaped a syllabus, got it approved for academic rigor and attracted 15 last-minute enrollments. The students were divided into teams and asked to create mini-projects in one of three developing countries where Goel and some students had public sector contacts: India, Peru or Botswana.
During videoconferencing sessions, Goel coached students in the intricacies of computer modeling and how to develop analytical frameworks to compare different approaches to solving policy problems. Notable guests made virtual appearances, including Goel’s good friend, Chief Economic Advisor in the Indian government Krishnamurthy Subramanian, and Peru’s Vice Minister of Education Sandro Parodi. These leaders described the challenges their nations faced in the pandemic, and the students developed and improved computer models to help solve them through a progress report and feedback loop just as teams of professional experts might approach the task.
The products of their quarter-long exploration impressed both students and instructor and created a composite picture of a disease whose spread varied by country and was affected by social and economic structures that, in turn, necessitated nuanced decisions on how to deal with COVID-19.
For instance, Goel and his students found that India’s massive population of rural poor, who often travel long distances to work in the cities, drove the spread of the disease. Migratory workers who contracted the disease on the job could potentially carry it back to the villages where they lived – a recipe for a monumental public health challenge that the computer models quantified.
Goel said he was delighted at the ingenious ways in which the students framed the problems they sought to address, and then developed computer algorithms and models to help suggest life-saving solutions.
Following examples are some key insights developed by three teams in Goel’s class.
One India project team tackled the dearth of tests and the expense of testing. Their solution was to “pool” tests by including swabs from many individuals in a single test – the members of an extended family, perhaps. Pooling typically is employed as a statistical tool to gauge infection levels in a broader population. But the students developed models suggesting that India use pooling in a different way, as a basis for isolating infectious persons. The students believe their pooling strategy could increase screening by up to 30 times to stretch the reach of available tests and lower the cost of testing. Through their models, the students found that isolating small groups based on pooled results, even if only some of them were infected, was effective at heading off the disease while lowering testing costs. This approach also promised to help curb the spread of infection at less economic cost than shutting down entire regions of the country.
After hearing Peru’s vice minister of education describe his country’s return-to-school quandary, another student team set about devising the safest possible reopening strategy. “We combed the available scientific literature and came across a paper that proposed a ‘two weeks in school, two weeks at home’ approach,” said team member Nicolas Hochschild, explaining the team’s rationale targeting the 14-day window in which most patients develop symptoms. Developing a computer model of such a staggered schedule, the students produced results promising enough that Parodi requested additional modeling to examine alternating groups for one week at a time to improve results. The student team has promised to produce a report on those findings and present it in a full briefing to Parodi even after the class had concluded.
After team members wrote the leaders of the Botswanan presidential task force, asking them to suggest areas in which they might help, they learned that Botswana is heavily reliant upon trade with South Africa that is conducted largely by trucks that carry goods and produce between the two countries. “There was a noted prevalence of the disease in Botswanan truck drivers,” said graduate student Tumisang Ramarea, who was born and raised Botswana. So, the student team created a model that weighed the status quo – doing nothing – against various mitigation strategies in the driver community, including isolating drivers from the general population and switching drivers at the border. Their goals were to stifle the disease while maintaining trade. Their models found that doing some intervention was better than allowing the status quo to persist. Doctoral candidate Kiran Shiragur, who supported the Botswana and Peru teams, said work has continued after the class ended to refine the suggested interventions.