An evaluation and shortening of the Cooking and Food Provisioning Action Scale (CAFPAS) using item response theory

Abstract

The Cooking and Food Provisioning Action Scale (CAFPAS) is a 28-item validated tool for measuring food agency, a latent construct representing an individual’s ability to make and achieve food-preparation and -provisioning goals. Here, key measurement parameters (targeting, threshold ordering, item fit, unidimensionality, differential item functioning, local dependency, and person reliability) of the CAFPAS are evaluated using a specific case of item response theory, Rasch analysis, on data from a development sample (N = 1853; 910 from Sweden; 943 from the US). Winsteps (v.5.1.7) is used for this analysis. The similarity of the Swedish version of the CAFPAS to the original is also assessed. Based on an iterative assessment of the measurement properties with different combinations of items in the development sample, ways to shorten the CAFPAS without jeopardizing construct validity or person reliability are examined. After removing items that do not fit the Rasch model, or that appear redundant in relation to other items, an 11-item version (CAFPAS-short) is suggested and tested using further Rasch analysis on both the development sample and an additional US-based validation sample (N = 1457). Scores of cooking confidence and attitudes are then modelled with measures from the CAFPAS and CAFPAS-short using frequentist and Bayesian analysis. Results suggest that the CAFPAS-short performs similarly to the full-length version, and potential future improvements to the CAFPAS are discussed. This study represents a successful application of item response theory to investigate and shorten a psychometric scale, reducing cognitive load on participants in studies using the CAFPAS whilst minimizing loss of data reliability.