Factors associated with sharing e-mail information and mental health survey participation in large population cohorts

Abstract

Background: People who opt to participate in scientific studies tend to be healthier, wealthier and more educated than the broader population. Although selection bias does not always pose a problem for analysing the relationships between exposures and diseases or other outcomes, it can lead to biased effect size estimates. Biased estimates may weaken the utility of genetic findings because the goal is often to make inferences in a new sample (such as in polygenic risk score analysis).

Methods: We used data from UK Biobank, Generation Scotland and Partners Biobank and conducted phenotypic and genome-wide association analyses on two phenotypes that reflected mental health data availability: (i) whether participants were contactable by e-mail for follow-up; and (ii) whether participants responded to follow-up surveys of mental health.

Results: In UK Biobank, we identified nine genetic loci associated (P <5 × 10-8) with e-mail contact and 25 loci associated with mental health survey completion. Both phenotypes were positively genetically correlated with higher educational attainment and better health and negatively genetically correlated with psychological distress and schizophrenia. One single nucleotide polymorphism association replicated along with the overall direction of effect of all association results.

Conclusions: Re-contact availability and follow-up participation can act as further genetic filters for data on mental health phenotypes.

Keywords: Generation Scotland; Partners Biobank; Selection bias; UK Biobank; cohort studies; follow-up studies; genome-wide association study; mental health.

Figure 1.

Manhattan plot of e-mail contact in UK Biobank.

Figure 2.

Manhattan plot of data available in MHQ follow-up.

Figure 3.

LD score genetic correlations (rg) with e-mail contact (triangle) and MHQ data (circle), with 95% confidence intervals.

Figure 4.

Possible effects of selection bias on polygenic risk score analyses in follow-up studies. (A) Causal model to be tested where PRS causes phenotype Y via phenotype X. (B) Worst-case scenario where PRS influences X but not Y and both phenotypes cause follow-up participation. Analysing only follow-up participants is the same as conditioning on F, which induces a correlation between PRS and Y. (C) More likely scenario, where both X and Y cause follow-up participation. Conditioning on F attenuates estimates of the relationship between PRS and Y. (D) Ideal scenario where X causes follow-up participation, but Y does not. Conditioning on F has no impact on the dependence of Y on PRS. PRS, polygenic risk score; X and Y, phenotypes of interest; F, selection into follow-up; directional solid line, true causal association; dashed line, induced or attenuated statistical dependence.