The Dutch Activity Card Sort institutional version was reproducible, but biased against women

Abstract

Purpose: To examine the reproducibility of the institutional version of the Dutch Activity Card Sort (ACS-NL) and the possible presence of gender bias. Methods: Older rehabilitation inpatients (N = 52) were included. Intra- and inter-rater agreement for the ACS-NL total and subscale scores was examined by intraclass correlations (ICC), and agreement of individual items by the κ coefficient (k). Gender bias was examined by the proportion of men and women selecting an ACS item. Results: ICC for inter-rater agreement of the ACS total score ranged between 0.78 and 0.87, ICC for intra-rater agreement ranged between 0.79 and 0.89. Median inter-rater κ for ACS-NL items was 0.72 (interquartile scores; 0.62–0.80). The inter-rater agreement (k = 0.43) and intra-rater agreement (k = 0.39) for the five most important activities was lower. Twenty ACS activities favoured men and seven activities favoured women. As a result, men scored systematically higher on the ACS-NL than women. Logistic regression analysis correcting for activity engagement level confirmed our findings. Conclusions: The reproducibility of the ACS-NL was high. The ACS-NL institutional version score may be biased in favour of men.Implications for RehabilitationThe ACS is a valuable addition to currently available measurement tools for occupational therapy. The ACS is unique in its ability to assess the level of activity engagement of older adults at the individual level. The potential of the ACS-NL to reliably measure the activity engagement level is supported in this study.The observed gender bias is a shortcoming. However, since the other versions use other scoring categories, a different scoring procedure (percentage change score) and are targeted to people with other health conditions and in other contexts, it cannot be assumed that the gender bias is present in the other versions to the same extent. Therefore, future studies should focus on the presence of gender bias, since all ACS versions are increasingly used for descriptive purposes and correlation-based analysis.