The Brazilian Household Food Insecurity Measurement Scale (EBIA) has eight general/adult items applied in all households and six additional items exclusively asked in households with children and/or adolescents (HHCA). Continuing an investigation programme on the adequacy of model-based cut-off points for EBIA, the present study aims to: (i) explore the capacity of properly stratifying HHCA according to food insecurity (FI) severity level by applying only the eight ‘generic’ items; and (ii) compare it against the fourteen-item scale.

Latent class factor analysis (LCFA) models were applied to the answers to the eight general/adult items to identify latent groups corresponding to FI levels and optimal group-separating cut-off points. Analyses involved a thorough classification agreement evaluation and were performed at the national level and by macro-regions.

Data derived from the cross-sectional Brazilian National Household Sample Survey of 2013.

A nationally representative sample of 116 543 households.

In all households and investigated domains, LCFA detected four distinct household food (in)security groups (food security and three levels of severity of FI) and the same set of cut-off points (1/2, 4/5 and 6/7). Misclassification in the aggregate data was 0·66 % in adult-only households and 1·06 % in HHCA. Comparison of the scale reduced to eight items with the ‘original’ fourteen-item scale demonstrated consistency in the classification. In HHCA, the agreement between both classifications was 96·2 %.

Results indicate the eight ‘generic’ items in HHCA can be reliably used when it is not possible to apply the fourteen-item scale.

For the emerging DSM-V, it has been recommended that dimensional and categorical methods be used simultaneously in diagnostic classification; however, little is known about this combined approach for abuse and dependence.

Using data (n=37 708) from the 2007 National Survey on Drug Use and Health (NSDUH), DSM-IV criteria for prescription opioid abuse and dependence among non-prescribed opioid users (n=3037) were examined using factor analysis (FA), latent class analysis (LCA, categorical), item response theory (IRT, dimensional), and factor mixture (hybrid) approaches.

A two-class factor mixture model (FMM) combining features of categorical latent classes and dimensional IRT estimates empirically fitted more parsimoniously to abuse and dependence criteria data than models from FA, LCA and IRT procedures respectively. This mixture model included a severely affected group (7%) with a comparatively moderate to high probability (0.32−0.88) of endorsing all abuse and dependence criteria items, and a less severely affected group (93%) with a low probability (0.003−0.16) of endorsing all criteria. The two empirically defined groups differed significantly in the pattern of non-prescribed opioid use, co-morbid major depression, and substance abuse treatment use.

A factor mixture model integrating categorical and dimensional features of classification fits better to DSM-IV criteria for prescription opioid abuse and dependence in adults than a categorical or dimensional approach. Research is needed to examine the utility of this mixture classification for substance use disorders and treatment response.