| Popis: |
Researchers often use data from self-reported questionnaires to study symptoms of mental health, yet due to the sensitivity of these questions, some participants may skip questions or select options such as ‘don’t know’ or ‘prefer not to answer’ (item missingness), while others may not participate in entire questionnaires (non-participation). However, the impact of item missingness and its similarities or differences with non-participation are not well understood in psychiatric genetic research. In this study, we investigate participation in the Stratifying Resilience and Depression Longitudinally (STRADL), a follow up study of Generation Scotland: Scottish Family Health Study, and item missingness in the CIDI-SF depression questionnaire contained within STRADL. We find demographic predictors of participation and complete response, suggesting that data is not missing completely at random. However, we observe discontinuity between participation and complete response: among participants, those with higher propensity scores to participate based on baseline demographic variables were less likely to respond to all depression items in the CIDI-SF, and propensity scores explained 3.18% of the variance in having a missing item response. We also assess the performance of depression polygenic scores in predicting lifetime differences using different approaches to deal with non-participation and missing data. Our findings highlight the impact that non-participation and item missingness may have on downstream analyses in psychiatric genetic research and beyond. |
