Abstract

Objectives

Cognition is the most commonly used EQ-5D bolt-on, with many different versions varying by descriptors and response levels (3L vs 5L). We aimed to systematically review the psychometric properties of cognition bolt-ons for the EQ-5D-3L and EQ-5D-5L.

Methods

A systematic review was conducted in PubMed, Web of Science, and Google Scholar following PRISMA 2020 guidelines (PROSPERO:CRD42023445567). We assessed the bolt-ons’ performance both as individual items and when added to the EQ-5D. Each bolt-on version was rated as positive (+) or nonpositive (−) within each publication using a checklist, and scores were summed across publications to reflect overall performance.

Results

In total, 101 publications from 72 studies met the inclusion criteria, examining 15 3-level and 13 5-level bolt-ons. The most frequently reported psychometric properties were item-level ceiling (n = 75) and known-groups validity (n = 54). Fewer studies explored convergent or divergent validity (n = 8 for each), responsiveness (n = 3), patient-proxy agreement (n = 2), and test-retest reliability (n = 1). None reported on content validity. Five-level bolt-ons outperformed 3-level bolt-ons in terms of overall performance (3L: 55+/57−; 5L: 45+/28−). Supportive psychometric evidence varied by populations, eg, head/brain injury (3L: 11+/11−; 5L: 1+/3−) and dementia (3L: 9+/8−; 5L: 4+/4−). The most-tested bolt-ons were the Janssen 2013 (5L, cognition: 18+/15−) and Haagsma 2005 (3L, thinking ability: 8+/12−) versions, with fewer than 10 assessments for all other bolt-ons.

Conclusions

Despite several publications, the psychometric evidence remains insufficient to identify a preferred cognition descriptor. Future research should prioritize content validity testing to inform the selection of candidate items, with quantitative psychometric evaluation preferably conducted afterward.