Respondents in an epidemiologic survey had fewer psychotropic prescriptions than nonrespondents: an insight into health-related selection bias using routine health insurance data.

Marie-Noël Vercambre, Fabien Gilbert

MGEN Foundation for Public Health, Paris, France. mvercambre@mgen.fr

Journal of clinical epidemiology 2012 Nov

filter terms:

Using register-based insurance data, we aimed to evaluate health-related differences between respondents and nonrespondents in a mailed epidemiologic survey, with a particular focus on mental health. In 2005, 19,406 adults covered by the national education system health insurance plan were sent a general health questionnaire including a significant part devoted to mental health. Of them, 52% responded. Using multivariable logistic regression models, we compared participants and nonparticipants on various sociodemographic characteristics, morbidity variables, and health care utilization indicators derived from the routine health insurance databases, regardless of response status. Mental health was appraised through the reimbursement of psychiatric services and the volume of psychotropic prescriptions received in 2004-05. In addition to traditional sociodemographic covariates of participation, we observed that respondents used more medical services than nonrespondents (visits to general practitioners, dentists, and specialists other than psychiatrists) but essentially for somatic disorders, as they also were prescribed significantly fewer psychotropic drugs. Response bias may impact estimation quality even in apparently sociodemographically homogeneous populations. Our results confirmed that persons with mental complaints are less likely to respond to a survey focused on their troubles and illustrate the risk of underestimation of psychiatric disease in population-based surveys. Copyright © 2012 Elsevier Inc. All rights reserved.

Citation

Marie-Noël Vercambre, Fabien Gilbert. Respondents in an epidemiologic survey had fewer psychotropic prescriptions than nonrespondents: an insight into health-related selection bias using routine health insurance data. Journal of clinical epidemiology. 2012 Nov;65(11):1181-9