Measuring what matters: evaluating the psychometric properties of goal attainment scaling in a psychosocial dementia trial.

BACKGROUND

It can be challenging to identify clinically meaningful, responsive primary outcomes for psychosocial intervention trials in people with dementia. Goal Attainment Scaling (GAS), a personalised, patient-generated outcome measure, enables individuals to define and evaluate progress toward meaningful goals. Evidence regarding its psychometric properties in dementia trials, however, remains limited.

OBJECTIVE

To evaluate the feasibility, reliability, and validity of GAS within the randomised controlled trial of NIDUS-Family (New Interventions in Dementia Study), a goal-based psychosocial intervention.

METHODS

302 people with dementia and family carer dyads collaboratively set baseline goals toward supporting the person with dementia to live well at home. In a psychometric evaluation guided by the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) framework, we examined follow-up completion (measuring feasibility); rater agreement; intra-rater reliability; construct validity by assessing correlations between change in GAS and quality-of-life (DEMQOL-proxy and CarerQoL); and predictive validity by testing whether GAS change predicted living at home at 12 months.

RESULTS

GAS completion rates were 86% at 6 months and 82% at 12 months. GAS demonstrated excellent inter-rater (Intraclass Correlation Coefficient = 0.99) and intra-rater reliability (Cohen's kappa = 0.91). There were significant but weak positive correlations between GAS change and carer and person with dementia quality-of-life (r = 0.14-0.22). GAS change was associated with remaining at home (odds ratio 1.08 per point increase; P <.01).

CONCLUSION

GAS is feasible, valid and reliable, so may be a sensitive and clinically meaningful candidate primary outcome for psychosocial dementia trials. Correlations with global quality of life were significant but modest, consistent with GAS capturing personalised rather than generic outcomes.