Web-based questionnaires to capture acute infections in long-term cohorts: Findings of a feasibility study

BACKGROUND: Incidence of acute respiratory infections (ARI) and gastrointestinal infections (GII) are difficult to assess due to high frequency episodes, limited severity and short duration. Retrospective assessments therefore are particularly prone to recall bias, while prospective assessment with conventional questionnaires requires high discipline from participants which is difficult to maintain over longer time periods. Web-based questionnaires (WQ) allow integration of a recall system and thus carry the potential to prospectively capture acute infections. We investigated the feasibility of a weekly WQ assessing symptoms of ARI and GII among participants of the German National Cohort (GNC). MATERIAL AND METHODS: In the study centres Hamburg and Bremen of the GNC participants of the Pretest 1 phase (September to November 2011) were invited to additionally take part in this feasibility study testing the WQ. Every Monday participants received an e-mail, containing a link to the WQ, asking for occurrence of ARI or GII symptoms during the past 7 days. The study took place from the beginning of February until mid-July 2012. We calculated the overall proportion of participation, weekly participation and the number of weekly reports per participant and we estimated incidences of ARI, ILI and GII. RESULTS: Of 200 Pretest 1 participants 171 (86 %) reported having an email address and thus were eligible for the web-based study. A total of 167 (98 %) agreed to participate. Participants of the web-based study were younger and better educated than non-participants. Access to Internet decreased with increasing age. Of the 167 participants in the feasibility study, 144 (86 %) responded at least once during the study period of 23 weeks, 5 persons (3 %) had non-functioning email addresses and 18 (11 %) did not respond at all. The weekly response varied between 62 % and 81 %, the median was 74 % (IQR: 71–77 %). Weekly median reports per person were 20 (IQR: 14–22; range 1–23). More than 90 % of participants responded during the first 3 days. The following mean incidence rates were found: ARI, 12 %; ILI, 0.49 %; and GII, 3 %. CONCLUSION: Use of WQ in prospective studies seems well possible, as Internet access is frequent among study participants and major technical problems did not occur. We observed high participation during the study period of 6 months and low drop out numbers. Participants of the web-based study were slightly younger and better educated than non-participants, so selection bias is possible and must be kept in mind when discussing generalizability of the results.