Broadly, I am interested in contextual determinants of health — how do the people we know and the places in which we live affect how long and how well we live?

Context can affect health by encouraging or discouraging physical activity, particularly walking.  Physical activity prevents cardiovascular disease, diabetes, breast and colon cancers, and may improve mood, but most adults don’t get enough.  Still, most of those adults know they aren’t as active as would be best for their health.  Life, in the form of family responsibilities, difficulties finding places to exercise, etc., gets in the way.  Telling people whose lives are constrained that they need to change their lifestyle is not all that effective. Instead, a public health approach identifies changes that broadly encourage activity.

One likely leverage point is the ‘built environment’, or the cities, towns, and suburbs in which most of us live.  One thread of my research has focused on built environment measurement, a necessary precursor to understanding how the environment affects activity. Through an NIH grant Co-PI’d by my primary academic mentor, Andrew Rundle, I built the Computer Assisted Visual Neighborhood Visual Assessment System (CANVAS), a web application to help people gather data from Google Street View imagery.  This system proved reliable for many street audit measures, a subset of which can be combined using measurement modeling techniques to create a valid and reliable measure of neighborhood physical disorder (the ‘broken windows’/urban deterioration factor in a neighborhood).  We  produced a map of this measure in New York City for a scholarly audience and have begun to investigate how (if at all) disorder relates to physical activity. Finally, we have used this same Street View imagery to look at how the pedestrian environment contributes to the risk of being injured or killed by a car, a thread of research I intend to expand on in the future.

But if you’re a skeptic and a geek, you can’t do epidemiology without becoming interested in epidemiologic methods, the other thread of my research interests.  Better methodology was key to understanding  pellagra outbreaks in the early 20th century.  Better methodology helps you deal with imperfect measures, for example by using latent models to identify patterns or by analyzing aggregate measures in ways that minimize the risk of bias.  And, perhaps most importantly, the techniques we’ve already developed will allow us to incorporate ‘Big Data‘ with appropriately skeptical optimism.