The Parental Stress Scale revisited: Rasch-based construct validity for Danish parents of children 2-18 years old with and without behavioral problems

Abstract

Background: Experiencing parental stress is common among parents of children of all ages and is elevated in families characterized by stressors such as poverty, mental health problems, and developmental problems. The Parental Stress Scale (PSS) is a short measure for the assessment of perceived stress resulting from being a parent. Methods: This study examines the construct validity and psychometric properties of the Danish PSS using Rasch and graphical loglinear Rasch models in a sample of parents of 2-18-year-old children with and without known behavior problems. We emphasized analyses of differential item functioning, to ascertain whether the scale yields unbiased scores for subgroups of parents. Results: The 18-item PSS did not fit the Rasch model or a graphical loglinear Rasch model. After dichotomizing item responses and eliminating items 2 and 11, we found the PSS to consist of two distinct subscales measuring parental stress and lack of parental satisfaction. For the total sample, the Parental Stress subscale fit a very complex graphical loglinear Rasch model with differential item functioning relative to parental education and whether children had behavior problems or not. The Lack of Parental Satisfaction subscale fit a simple graphical loglinear Rasch model with differential item functioning only relative to subsample. When dividing into subsamples of parents of children with and without behavior problems, the Parental Stress subscale fit a simple graphical loglinear Rasch model, though still with differential item functioning, while the Lack of Parental Satisfaction subscale fit the Rasch model in each subsample of parents. Both subscales performed best for parents of children with behavior problems. Conclusions: The PSS should be used in a 16-item version and scored as two subscales. The PSS appears better suited for use among parents of children with behavior problems than within a sample without any known difficulties.