Web surveys are becoming increasingly popular among researchers due to many benefits such as reduced costs, efficiency of data entry, and automated navigation of the survey for the respondent.Researchers need to consider the design of the questions and web page to reduce respondent burden, which has been shown to increase breakoff rates and
decrease response quality overall. This study aims to understand how respondent burden varies across different types of questions by using paradata from the World Trade Center Health Registry’s (WTCHR) Wave 4 survey. In particular, we compare breakoff rates, time latencies, and number of navigational backups to identify questions that are particularly burdensome for some respondents. Preliminary findings indicate that 20% of breakoffs occurred on pages displaying more than one question, even though the vast majority of the survey pages displayed only one question. The majority of those screens with multiple questions contained grid questions. We further examine the relationships between these paradata measures and other characteristics of the survey, such as question content, question type, and survey length. Additionally, we assess differences in these paradata measures across demographic groups and device used to take the survey to help differentiate between issues due to the question and respondent characteristics. The results from this study will be used to (1) reduce respondent burden, decrease item nonresponse and therefore improve accuracy of the data for future WTCHR surveys, and (2) inform other survey researchers of ways to reduce respondent burden and improve response quality in their own surveys.