Pascal Geldsetzer believes in open access, in disseminating science as quickly as it happens. Even so, last summer, as he uploaded the surprising results of his latest study to the MedRxiv preprint server, the Stanford University epidemiologist was feeling something other than the usual excitement. “I was scared to put this up because it’s such a different approach from what’s generally done in epidemiology and medicine,” he said.
During his Ph.D. studies at Harvard, Geldsetzer had taken classes in econometrics — a field that has become focused on building sophisticated statistical techniques to decipher cause and effect amidst the messiness of the real world. Rather than design studies that try to control for every potential confounding variable, these methods rely on identifying “natural” experiments, such as when a government policy assigns people to an intervention based on a threshold. People on either side of the cutoff are essentially the same but receive different treatment. “It’s like a randomized controlled trial, but it’s done by nature,” Geldsetzer said.
That’s what he did in his new paper to test the emerging idea that dementia can be caused by the shingles virus. Social scientists have been using these methods for decades, but very few researchers had ever applied it to a clinical setting.
This article is exclusive to STAT+ subscribers
Unlock this article — plus in-depth analysis, newsletters, premium events, and news alerts.
Already have an account? Log in