Comparing alternative methods of measuring cumulative risk based on multiple risk indicators: Are there differential effects on children's externalizing problems?

Abstract

This study examined several alternative methods to measure cumulative risk (CR) based on multiple risk indicators. Several methods for measuring CR are presented and their conceptual and methodological assumptions are assessed. More specifically, at the individual risk level, we examined the implications of various measurement approaches (i.e., dichotomous, proportion- and z-scores). At the composite level, we measured CR as an observed score, and compared this approach with two variable-centered approaches (consisting of reflective and formative indicators) and two person-centered approaches (consisting of latent class analysis and latent profile analysis). A decision tree was proposed to aid researchers in comparing and choosing the alternative methods. Using a sample of 169 low-income families (children approximately 5 years old, 51% girls; 74% African American, and their primary caregiver), we specified models to represent each of the alternative methods. Across models, the multiple risk composite was based on a set of 12 individual risk indicators including low maternal education, hunger, meal and money unpredictability, maternal psychopathology, maternal substance use, harsh parenting, family stress, and family violence. For each model, we estimated the effect size of the composite CR variable on children's externalizing problems. Results indicated that the variable-centered CR composites had larger effects than the observed summary score CR indices and the person-centered methods.