Observed trends in mental health: A strategy to adjust for nonresponse bias and demographic changes in survey data

Abstract

To examine whether observed changes in self-reported psychological distress symptoms reflect true changes in psychological distress in the population, or more specifically, to present statistical methods to check for nonresponse bias and demographic changes. Lack of representativity, nonresponse bias and demographic changes were controlled for by weighing of data (sample-balancing or raking). Controlling for age, gender and education, or holding the demographic structure constant over time, slightly weakened the decrease in psychological distress from 1998 to 2002, and left the changes in psychological distress from 2002 to 2008 almost unchanged. When the level of psychological distress was held constant across years in every subgroup defined by age, gender and education, in order to isolate the effects of changes in demography, the distress scores showed only minor variation across data collection occasions from 1998 to 2008. The observed modest decrease in psychological distress from 1998 to 2002 may partly be explained by selection and demographic changes, while this is not the case with the more pronounced changes from 2002 to 2008.