Deployment of a mixed-mode data collection strategy does not reduce nonresponse bias in a general population health survey.