Published in:
Open Access 01-12-2014 | Research article
Comparison of a French pediatric type 1 diabetes cohort’s responders and non-responders to an environmental questionnaire
Published in: BMC Public Health | Issue 1/2014
Login to get accessAbstract
Background
Type 1 diabetes (T1D) incidence has doubled since the 1980’s for children aged <5 years old, potentially relevant environmental factors having thus to be sought early in the patient’s life. The identification of environmental factors that can explain the changing epidemiology of T1D requires comprehensive environmental inquiries. However, a limitation is the willingness of patients and families to complete these environmental questionnaires. Our objective was to identify patients’ personal and social characteristics predictive of the return, time to the return and completeness of a comprehensive environmental questionnaire.
Methods
The parents of 2832 T1D patients aged <15 years old enrolled in the French Isis cohort were sent a 1379-item environmental questionnaire. A geographic information system was used to collect information on patients’ socioeconomic environment. Multivariate statistical analyses were conducted to identify predictors of questionnaire return, time to its return and its completeness.
Results
Within 6 months, 867 (30.6%) questionnaires were returned. Socioeconomic environment was strongly associated with the probability of response, with fewer responses from cities with high Townsend deprivation index (p =2 × 10−7), high unemployment (p =0.005), blue-collar workers’ rate (p =0.0002) and household overcrowding (p =0.02). Response rates were similar for male and female patients, but were higher for less severely affected patients (p =0.006) and younger patients (p =5 × 10−5). When returned, completeness was high with a mean of 96%.
Conclusion
Identification of personal or socioeconomic characteristics differing between questionnaire responders and non-responders may help target future environmental investigations on those patients who will more likely return the information, and reduce bias using these variables to stratify the analyses.