OBJECTIVES:   The current VLA disability questionnaire (L-VLA), which provides an assessment of disability in a broad range of life activities, queries difficulty in 33 activities, with 4 follow-up questions for each activity regarding the need for accommodations (ACCs).  Our objective was to create a short version of the questionnaire incorporating the use of ACCs, for rheumatoid arthritis (RA).  METHODS:  Data from 2 years' administration of the L-VLA in a longitudinal cohort of individuals with RA were used (n=449,421).  Analyses were first conducted with Year 1 data and then validated with Year 2 data.  Item response analyses were performed.  Items were deleted based on misfit, logit values that duplicated information provided by other items providing information at similar logit values, and substantive considerations.  Psychometric properties of the short version (S-VLA) and L-VLA were compared.  RESULTS:  Partial credit and rating scale models were both tested; the partial credit model had fewer item and step misfits (G2 likelihood ratio = 623.881, p<0.0001).  The scoring range was expanded to 0-5 to account for use of ACCs.  Deletions were progressive, resulting in a 21-item version, a 14-item version, and 4 12-item versions.  A 14-activity version of the VLA questionnaire, with 1 follow-up question for each regarding ACCs, was selected.  The S-VLA had 0 misfitting steps; 2 items had potential misfit but were retained for both substantive reasons and to provide logit spread.  Correlation between L-VLA and S-VLA scores was 0.97; correlations of the L-VLA and S-VLA with other measures of functioning and RA impact were similar.  Analyses were confirmed in Year 2 data. CONCLUSIONS:  The S-VLA appears to be reliable and valid. The S-VLA may be more sensitive to change than other measures of functioning such as the HAQ and may appear more relevant to individuals with RA.