Simple, but it works well—and that's the whole point
When Boston University reopened its campus in the fall of 2020, it confronted a question that no institution had a reliable answer to: how many beds must be held ready for those who must be separated from the community? A team of mathematicians and public health researchers responded not with elaborate machinery, but with five lines of code — a model that could see ten days into the near future of a pandemic. In doing so, they offered congregate institutions everywhere a way to move from fear-driven reaction to principled foresight.
- BU's early quarantine bed estimates were badly miscalibrated, leaving administrators holding hundreds of empty beds while uncertainty about the real scale of need persisted.
- The pressure to house students safely while managing finite resources created an urgent institutional crisis with no established playbook to follow.
- A statistician, a biostatistician, and a doctoral student built a five-line model drawing on daily case counts, contact tracing, and behavioral patterns like holiday travel to forecast bed needs a full ten days out.
- The model's surprising accuracy shifted BU's posture from reactive scrambling to confident, advance resource allocation.
- Published in The American Journal of Public Health, the methodology is now being offered to prisons, nursing homes, and military facilities — any place where people live close together and the outside world keeps walking in.
When Boston University reopened in fall 2020, administrators faced a problem with no precedent: how many quarantine beds would actually be needed? Their early estimates were wrong — hundreds of beds sat empty while uncertainty reigned. The gap between guessing and knowing was costly.
Eric Kolaczyk, a mathematics and statistics professor directing BU's Hariri Institute, joined forces with biostatistician Laura White and doctoral student Wenrui Li to build something more reliable. Their solution was disarmingly spare: a five-line statistical model that combined daily case counts, contact tracing data, and behavioral patterns — including student travel around holidays — to project quarantine housing needs ten days in advance.
The power of the model lay not in its complexity but in its discipline. Beneath the brevity was rigorous epidemiological reasoning about how infections enter a population and spread locally. Li was struck by how well such a simple approach performed. Peter Smokowski, BU's vice president of Auxiliary Services, acknowledged the early stumbles but credited the refined model with giving the university a concrete foundation for planning rather than fear.
The team published their findings in The American Journal of Public Health with a broader audience in mind. Correctional facilities, nursing homes, and military barracks face the same fundamental challenge — people living in close quarters while remaining connected to the outside world. Kolaczyk framed the publication as a recognition that the need for quarantine space optimization is global and ongoing. The model's promise is simple: institutions need not wait for crisis to force their hand when a tool exists to help them prepare.
When Boston University decided to bring students back to campus in the fall of 2020, the university faced a problem that had no precedent: how many quarantine beds would actually be needed? The initial guesses were wrong. BU had set aside hundreds of beds expecting a surge that never materialized at the scale predicted, leaving administrators scrambling to understand what they were really looking at.
Eric Kolaczyk, a mathematics and statistics professor who directs the Hariri Institute, teamed up with Laura White, a biostatistician at BU's School of Public Health, and Wenrui Li, a doctoral student in mathematics, to build something better. Rather than rely on hunches or worst-case scenarios, they created a statistical model that could forecast quarantine housing needs ten days in advance. The model was elegantly simple: it took daily case counts, contact tracing data, and patterns of human behavior—accounting for when students might travel during long weekends or holidays—and converted those inputs into a reliable projection of how many isolation beds the university would need.
The breakthrough was not in complexity but in restraint. The entire model ran on five lines of code. Yet beneath that simplicity lay rigorous statistical reasoning grounded in standard epidemiological principles about how infected people arrive in a population and how transmission spreads locally. Li found himself surprised by how well such a straightforward approach worked. "Our model is simple, but it works well," he said. Kolaczyk echoed the sentiment: the software's brevity masked a principled method that actually captured what was happening.
BU's experience during the pandemic had taught the team something crucial. Peter Smokowski, the university's vice president of Auxiliary Services, acknowledged that the initial predictions had been off the mark. "We found ourselves in uncharted waters," he said. But the refined model that Kolaczyk's team developed provided a concrete benchmark—a way to move from guessing to knowing. The university could now allocate its housing resources with confidence rather than fear.
The researchers published their findings in The American Journal of Public Health, but they saw the work as pointing beyond BU's campus. Correctional facilities, nursing homes, military barracks, and any other congregate setting where people live in close quarters while maintaining contact with the outside world faced the same challenge. A future variant of COVID-19, or an entirely new disease, could emerge at any moment. Having a tool to predict quarantine space needs ten days ahead meant institutions could shift from reactive scrambling to proactive planning. "The driver for us in transitioning this from a BU project to a publication," Kolaczyk explained, "is the realization of how ubiquitous the need for quarantine space optimization is across the world." The model offered a way to allocate resources from a baseline of relative calm rather than waiting until crisis forced the hand.
Citações Notáveis
We found ourselves in uncharted waters in determining the number of quarantine and isolation beds, but the modeling that Eric's team completed was very helpful in establishing a benchmark number.— Peter Smokowski, VP of Auxiliary Services at Boston University
Our model is simple, but it works well.— Wenrui Li, doctoral student in Mathematics & Statistics
A Conversa do Hearth Outra perspectiva sobre a história
Why did the initial predictions fail so badly? What were they missing?
They were working blind. No one had done this before. The early estimates didn't account for how contact tracing would actually work in practice, or how students would actually behave during holidays. They were essentially guessing at human patterns.
And the five-line code—that's not a limitation, is it? That's actually the point?
Exactly. Complexity doesn't equal accuracy. The team realized that if you understand the underlying mechanics—how infection arrives, how it spreads locally—you don't need a black box. You need something transparent and fast.
Can this really work in a prison or a nursing home? Those are very different from a university.
The core dynamics are the same. You have a confined population with some contact to the outside world. The model adapts to whatever data you feed it. The principle holds.
What happens when a new variant emerges? Does the model break?
Not necessarily. It's built on epidemiological fundamentals that don't change. But you'd need to recalibrate it with new data about transmission rates. The framework survives; the parameters shift.
So this is really about moving from panic to planning.
That's it. Ten days of warning lets you prepare instead of improvise. That's the difference between managing a crisis and being managed by one.