The incidence of acute gastrointestinal illness in Sweden.
Bottom Line: To validate the data on symptoms, we compared the study results with anonymous queries submitted to a Swedish medical website.Comparing the number of reported episodes with web queries indicated that the low number of episodes during the first 6 months was an effect of seasonality rather than recall bias.Further, the result of the recall bias analysis suggested that the survey captured approximately 65% of the true number of episodes among the respondents.
Affiliation: Department of Knowledge Development, Public Health Agency of Sweden, Sweden.Show MeSH
Related in: MedlinePlus
License 1 - License 2 - License 3
Mentions: A vast majority of all reported episodes of stomach disease were reported to have occurred during the last 6 months of the recall period, i.e. between 1 November 2008 and 30 April 2009, which could be an effect of recall bias. From the syndromic surveillance system based on web queries, it has previously been noted that queries on vomit have a strong seasonality, while queries on diarrhoea are evenly distributed over time . In the study, we used web queries on (the Swedish equivalents to) stomach disease, diarrhoea and vomit. The total number of queries per month on these three symptoms varied between 1005 in July 2008 and 7017 in February 2010, with a median of 2140 queries per month. When comparing the results of the reported episodes of stomach disease (gastrointestinal symptoms) with the included web queries, we found that the reported episodes followed the same pattern as the web queries (Figure 2). This finding can explain why the vast majority of the episodes were reported to have occurred between November 2008 and April 2009 and were not necessarily due to an effect of recall bias. However, in exploring possible recall bias, we found that the number of reported episodes of stomach disease was significantly associated with both the web queries (=0.00016, p=0.03) and time (=0.095; p=0.01).Thus, after taking into account the effect of the pattern of the web queries, it still remained that the logarithm of the number of reported episodes increased approximately by 10% per month, which we interpret as a result of better recall closer to the date of the survey. We did not find any statistically significant autocorrelation in the model (DW statistic=1.38, p=0.082).
Affiliation: Department of Knowledge Development, Public Health Agency of Sweden, Sweden.