Using item response theory modeling to address bias in genetic findings for happiness, life satisfaction, and social support

In the current study, we investigated the influence of using skewed sum scores on heritability estimates of life satisfaction and happiness, their genetic overlap, and gene by environment interactions (GxE) with perceived social support. To this end, we analyzed item-level data from a large adult sample (Ns between 3610 and 11,305, depending on the type of analysis) where item response theory (IRT) models were incorporated in all genetic analyses to control for (heterogeneous) measurement error. Adopting item-level IRT models, compared to analyzing sum scores, had small effects on variance component estimates, although these translated into larger effects on genetic correlations. Largest effects were found for the most skewed variable (social support). Finally, in the IRT model for life satisfaction, but not for happiness, heritability estimates decreased at higher social support levels, while this was not observed when analyzing sum scores. Together, our results indicate that IRT can be used to address psychometric issues, especially in the context of GxE, related to the use of sum scores for complex traits like well-being.