Background: Technological advances have made it easier for researchers to collect more frequent longitudinal data from survey respondents via personal computers, smartphones, and other mobile devices. Although technology has led to an increase in data-intensive longitudinal studies, little is known about attrition from such studies or the differences between respondents who complete frequently administered surveys in a timely manner, and respondents who do not.

Objective: We examined respondent characteristics and behaviors associated with continued and on-time participation in a population-based intensive longitudinal study, using weekly web-based survey interviews over an extended period.

Methods: We analyzed data from the Relationship Dynamics and Social Life study, an intensive longitudinal study that collected weekly web-based survey interviews for 2.5 years from 1003 18-and 19-year-olds to investigate factors shaping the dynamics of their sexual behavior, contraceptive use, and pregnancies.

Results: Ordinary least squares and logistic regression analyses showed background respondent characteristics measured at baseline were associated with the number of days respondents remained enrolled in the study, the number of interviews they completed, and the odds that they were late completing interviews. In addition, we found that changes in pregnancy-related behaviors reported in the weekly interviews were associated with late completion of interviews. Specifically, after controlling for sociodemographic, personality, contact information, and prior experience variables, we found that weekly reports such as starting to have sex (odds ratio [OR] 1.17, 95% CI 1.03-1.32, P=.01), getting a new partner (OR 1.76, 95% CI 1.53-2.03, P<.001), stopping the use of contraception (OR 1.28, 95% CI 1.10-1.49, P=.001), and having a new pregnancy (OR 5.57, 95% CI 4.26-7.29, P<.001) were significantly associated with late survey completion. However, young women who reported changes in pregnancy-related behaviors also had lower levels of study attrition, and completed more interviews overall, than did their counterparts.

Conclusions: We found that measures of participation in a longitudinal study with weekly web surveys varied not only by respondent characteristics, but also by behaviors measured across the surveys. Our analyses suggest that respondents who experience the behaviors measured by the study may maintain higher participation levels than respondents who do not experience those behaviors.