Multigroup IRT modeling clearly illustrates differences in poly-substance use between ages and sexes that are well conceptualized as affecting a single, continuous dimension. Though diversity of substances used does not efficiently estimate specific latent trait levels across the broad dimension of poly-substance use involvement (e.g. the inclusion of frequency items could potentially better capture the high-end of the use severity dimension), it would provide a reasonable estimation of latent trait levels for those who fall at a mid-level of poly-substance use involvement, starting in mid-adolescence.
