Use population data instead of recruiting participants as controls.