Rasch analysis of the long-term conditions questionnaire (LTCQ) and development of a short-form (LTCQ-8).
Batchelder L., Fox D., Potter CM., Peters M., Jones K., Forder JE., Fitzpatrick R.
BACKGROUND: The aim of the current study was to evaluate the structural validity of the 20-item long-term conditions questionnaire (LTCQ) and to explore a potential short-form version of the scale using Rasch analysis. METHODS: Data were collected through postal surveys (February 2016-January 2017) from a sample of 1,211 participants diagnosed with at least one long-term condition (LTC). Identified participants were invited through either local authorities for a social care cohort (n = 294) or primary care practices for a health care cohort (n = 917). Participants were mailed a survey, including the LTCQ, demographic questions, a comorbidities measure, and other validated outcome measures. Respondents were invited to complete a follow-up survey including the LTCQ for assessment of reproducibility. RESULTS: The main assumptions of the Rasch model from the LTCQ were fulfilled, although infit and outfit indices indicated some items showed misfit. Misfitted items, items that did not have a preceding set or showed some local dependence were removed one at a time, with the remaining candidate items to form an 8-item short version, the LTCQ-8. The Rasch model for the LTCQ-8 explained 64% variance and had a reliability estimate greater than 0.80. Several items in the LTCQ showed uniform differential item function (DIF) in relation to the number of reported LTCs, age, cohort and type of LTCs, but fewer items exhibited DIF in the LTCQ-8. Spearman's rho correlations between the LTCQ and the LTCQ-8 were strong across the total sample and various subgroups. Correlations between the LTCQ-8 and all reference measures were moderate to strong, and comparable to correlations found between the LTCQ and these measures. CONCLUSIONS: The LTCQ measures a unidimensional construct, and it is therefore acceptable to use a summed total score. The LTCQ-8 also met the assumption of unidimensionality and had comparable construct validity with the LTCQ. Additional validation is required in an independent sample.