To move forward, sometimes it’s best to look back.
That’s the determination of researchers from the Luddy School of Informatics, Computing, and Engineering in a study recently published in the journal Nature Physics, which shows that backward contact tracing—that is, tracing from whom a disease spreads—is profoundly more effective compared to the forward tracing method of tracking the spread of disease, including COVID-19.
The paper, “The effectiveness of backward contact tracing in networks,” was written by Sadamori Kojaku, a post-doctoral researcher with the Center for Complex Networks and Systems Research at Luddy, and Associate Professor of Informatics and Computing Yong Yeol Ahn, in conjunction with colleagues from the University of Vermont, the Technical University of Denmark. It developed a model that shows the exceptional efficacy of backward contact tracing at identifying super-spreaders and super-spreading events. The researchers hope to apply their method to other areas beyond COVID-19, such as the spreading of misinformation.
“Our paper provides a concrete mathematical explanation of how and why different types of contact tracing work,” Ahn said. “In particular, it emphasizes the importance of backward contact tracing and how crucial it is to have rapid turnaround in testing to leverage the strength of backward contact tracing. With this insight, I hope that testing and tracing strategies across the world can be improved.”
Even if the directionality of infection is unknown, it is possible to perform backward-aiming contact tracing.
“Performing backward contact tracing requires us to identify the pathway through which disease spreads, which is hard in practice,” Kojaku said. “However, by leveraging the nature of spreading and tracing, it is possible to perform backward-aiming tracing, which is nearly as effective as backward tracing, even if we don't know the pathway. Therefore, we believe that contact tracing has much more potential than we have considered so far.”
The project arose from a course Ahn was teaching in network science at the start of the COVID-19 pandemic, and he also was invited to provide a tutorial about network epidemiology at a COVID workshop. Ahn began to think about the interplay of network epidemiology and contact tracing, and he worked with Kojaku to develop the study.
The paper starts from the friendship paradox, which states that a person’s friends tend to have more friends than they do because the more friends someone has, the more often they show up in someone’s friend list. In epidemiological context, it means that those who have more contacts are more likely to be infected or contact-traced. Although this bias has been known, they discovered that there is yet another bias that helps backward contact tracing: The more infections an initial infection source produces—that is, as infections spread from one person—the more frequently that person shows up as a contact of positive cases identified. Unlike forward contact tracing, backward contact tracing is assisted by both biases and thus much more effective.
The study used simulations on both synthetic and high-resolution empirical contact datasets to show that strategically executed contact tracing can prevent a substantial fraction of transmissions with a higher efficacy than case isolation alone.
“Complex networks create so many novel opportunities for unexpected insights, and this study does a great job of showcasing how innovative methods can change the way we approach some of our biggest challenges,” said Kay Connelly, associate dean of research at the Luddy School. “Our faculty do a fantastic job of finding new ways to look at problems, and this study is a perfect example of the positive impact they can have on the real world.”