Response to mailed epidemiological surveys has decreased in recent decades. Since subjects with respiratory symptoms are usually early responders to surveys performed in Southern Europe, this trend could bias prevalence estimates. The present study aimed at evaluating the impact of non-response bias on prevalence estimates of respiratory symptoms and smoking habits.
METHODS:
In 9 centres, participating in the Italian Study on Asthma in Young Adults (ISAYA), random samples of people aged 20-45 years were administered a mailed questionnaire between 1998 and 2000. Non-responders were contacted again first by mail and then by phone. Cumulative response percentage was 30.5%, 52.4% and 72.7% (18,873/25,969), respectively, after the 1st, 2nd and 3rd contact.
RESULTS:
The prevalence of self-reported current asthma, asthma-like symptoms, and chronic cough/phlegm was more than halved from the first contact (5.6%, 17.8%, 14.6% respectively) to the third contact (2.7%, 6.4%, 6.9%). This pattern was less pronounced when considering allergic rhinitis and past asthma, whose prevalence decreased, respectively, from 21.5% to 15.6% and from 3.5% to 2.6%. At the same time the proportion of current smokers increased from 29.2% to 38%, while the proportion of ex-smokers decreased from 16.5% to 10.1%. In a multinomial logistic model current asthma, asthma-like symptoms, chronic cough/phlegm and smoking habits, and to a lower extent past asthma and allergic rhinitis, were significant predictors of late response.
CONCLUSIONS:
In Italy when response percentage is low, the prevalence of current asthma, chronic cough/phlegm and ex-smokers is overestimated, while the proportion of current smokers is underestimated.
