Two-handed task performance in upper limb prosthesis users: rethinking measurement approaches.

Abstract

The UEFS-P Two-Handed Task scale measures physical function of upper limb prosthesis users. Original scoring assumed a response of not performing a task indicated greater difficulty than the response, 'could not do' a task and collapsed several responses. This study explored alternative scoring approaches. Sample of 423 prosthesis users with major upper limb amputation. Scoring approaches included a partial credit model (PCM) of task difficulty rating, and six generalized linear mixed models accounting for number of tasks performed and task difficulty. Model fit and reliability were compared, and known group validity assessed. After selecting the best model, item fit, threshold and other parameters were evaluated. Concurrent validity with the original score was assessed. The Poisson and PCM model was selected. Measures were significantly different by amputation level, had the largest F statistic, reliability was 0.78; Pearson correlation with original scoring was 0.82. The new scoring strategy addresses face validity and clinical relevance concerns. Including a count of tasks performed with a prosthesis is part of the latent construct of prosthesis task performance, adding information not considered in a PCM model.