Methods of soliciting self-reported chronic conditions in population surveys: don't ask, don't report

The purpose of this cross-sectional study was to compare the self-reported prevalence of chronic conditions when respondents are simply asked to list all chronic conditions (recall method) versus when respondents are asked explicitly about the presence of specific conditions (recognition method). Using data from two Canadian population surveys, 17 separate logistic regression models were used to estimate the effect of method on the odds of reporting any chronic condition, and each of 16 specific conditions. Respondents exposed to the recognition method were nearly four times more likely to report any chronic condition than those exposed to the recall method. The effect of method varied widely across conditions, with those exposed to the recognition method 25 times more likely to report urinary incontinence, but only 1.3 times more likely to report diabetes, compared to those exposed to the recall method. In short, the estimates of chronic conditions obtained using the recall method will be different from those gathered via the recognition method, and the extent of this difference will vary by condition. Both survey designers and survey analysts must make the decision of which method is appropriate, given the goals of the survey or analysis.