Modifying the built environment is increasingly recommended as a means of increasing physical activity, but there is currently little evidence from intervention studies to support this approach. From a discussion of 3 natural experiments in this area (RESIDE, iConnect, and Commuting and Health in Cambridge), several common lessons emerged. First, researchers should anticipate delays in the implementation of interventions that are outside their control, and research funders need to exercise a degree of flexibility to accommodate changing research timetables. Second, new built environments develop and evolve over time, and so do their effects on human behavior. Study designs and exposure measures should take account of this, and long term outcomes should be measured wherever possible to allow for potential sleeper, snowball, or threshold effects emerging over time. Third, it may be difficult to identify suitable control areas for a conventional parallel-group intervention–control design, and it may be necessary to draw on other study designs to provide a counterfactual comparison. Fourth, the effort and cost required to recruit, retain and obtain repeated measurements from participants over a period of years should not be underestimated. Finally, comprehensive process evaluation measures may be required to assess the level and quality of interventions.
Ogilvie and Yang are with the MRC Epidemiology Unit and Centre for Diet and Activity Research (CEDAR), Cambridge, United Kingdom. Giles-Corti, Hooper, and Bull are with the School of Population Health, University of Western Australia.