Randomized Trial of Questionnaire Length

Abstract

Listen

Translate article

To the Editor: We conducted a pilot study to assess the feasibility of recruiting, enrolling, and following cohort study participants through the internet. The study population comprised Danish women planning to become pregnant. One objective was to determine how length of the baseline questionnaire affected enrollment and completeness of data. Eligible women were asked to complete a baseline enrollment questionnaire and provide their e-mail address and Central Population Registry number, a unique 10-digit number assigned to each resident of Denmark. They were then randomized to receive either a short (up to 233 questions [74%]) or long (up to 317 questions [100%]) version of the baseline questionnaire. Some questions were contingent on responses to others. Women could log out of the questionnaire and return at a later time. Both questionnaires included demographics; menstruation and contraceptive history; height and weight; and lifestyle exposures (eg, tobacco use, alcohol and caffeine intake, and medication use). The short version included an abbreviated reproductive and medical history, including questions about sexually transmitted diseases. The long version asked for additional details about prior pregnancies, including pregnancy outcome, gestational age, and weight gain, other diseases that could affect fertility, and any prior diagnosis of infertility. Other questions included only in the long form were measurement of waist and hip circumference, menstrual cycle characteristics, infertility history of mother, night shift work, and partner and parents’ education. After 6 months, 4807 women completed the screener, of whom 3070 (64%) met the eligibility criteria. Among these, 702 (23%) did not provide their Registry number, leaving 2368 women randomly assigned to the short or long baseline questionnaire. The median times to complete the long and short questionnaires were 23 and 18 minutes, respectively. Characteristics of the 2 randomly assigned groups were similar (eTable 1). Enrollment proportions were virtually identical for the 2 randomized groups, with 1129 of 1167 (97%) of those given the short form completing enrollment versus 1159 of 1201 (97%) of those given the long form. Few data items were missing, allowing for only small differences between the groups. For example, little information on smoking history was missing for both groups (0.3% missing in the short version vs. 0.4% in the long version), and coffee consumption was missing for 1.4% of women who received the short version and 1.2% who received the long version. Our results indicate no important differences in enrollment or extent of missing data between those assigned a short or long version of a web-based questionnaire. Studies of the length of paper-based questionnaires have mostly indicated a modest effect of questionnaire length on response.1–4 Other factors, including monetary incentives, recorded delivery, order of questions, or a teaser on the envelope had stronger effects on response than questionnaire length.1,5 Our study participants may have been more highly motivated than other study populations, possibly obscuring effects of questionnaire length that might be more apparent in less highly motivated populations. Our short questionnaire contained 26% fewer questions than the long version, which translated into only about a 5-minute difference in the median time to complete. Thus, our findings may not apply to questionnaires that are considerably longer. Nevertheless, these results indicate that using the internet for recruitment into a prospective cohort study seems to be an attractive strategy, and that in a motivated study population a questionnaire that takes longer than 20 minutes to complete is not necessarily a deterrent for participation. Kenneth J. Rothman Department of Epidemiology Boston University School of Public Health Boston, Massachusetts RTI Health Solutions Research Triangle Park North Carolina Ellen M. Mikkelsen Anders Riis Department of Clinical Epidemiology Aarhus University Hospital Aarhus, Denmark Henrik T. Sørensen Department of Epidemiology Boston University School of Public Health Boston, Massachusetts Department of Clinical Epidemiology Aarhus University Hospital Aarhus, Denmark Lauren A. Wise Department of Epidemiology Boston University School of Public Health Boston, Massachusetts Slone Epidemiology Center Boston University Boston, Massachusetts Elizabeth E. Hatch Department of Epidemiology Boston University School of Public Health Boston, Massachusetts