Measuring locomotor disability: a new scoring mechanism for some familiar items
- Arthritis Research Campaign National Primary Care Centre, Primary Care Sciences, Keele University, Keele, Staffordshire, UK
Background
Locomotor disability is a critical measure for studying functional decline in later life and has featured in several major epidemiological studies. Yet measurement has mostly been confined to single items or ordinal scales rather than truly unidimensional measures with interval properties, thereby limiting the detection and interpretation of change in locomotor disability over time. Our aim was to combine items from the well-known SF-36 Physical Functioning Scale (PF-10) into a single score using Rasch analysis and test the properties of this new measure.
Methods
A health survey was sent to a random sample of 500 adults aged 50 years and over registered with a single general practice in North Staffordshire. The questionnaire contained the PF-10, from which five items (two relating to stair climbing and three to walking) were selected. Response dependency exists between the items, and so they were formed into two “super-items” (stair-climbing and walking). These items were assessed for their fit to the Rasch measurement model in terms of threshold ordering, unidimensionality, item and person fit, invariance along the latent trait and differential item functioning (DIF) by age-group (50–65, 66+ years) and gender. Rasch analysis was carried out in RUMM2020. The score was confirmed in three external datasets and tests carried out for its repeatability and validity.
Results
372 people responded to the health survey and of these, 368 responded to at least one of the five PF-10 items. Both super-items displayed disordered thresholds that were corrected by combining adjacent response categories. There was no evidence against unidimensionality, and item and person fit was good. There was no evidence against invariance along the latent trait and no signs of DIF by either age-group or gender. The items from the external datasets also met Rasch model specifications and the score was shown to be repeatable and valid.
Conclusions
Applications of measurement scales with confirmed interval level properties are advantageous for epidemiological studies using repeated measures over time. We have developed and externally validated a scoring method for deriving a brief self-reported locomotor disability scale with Rasch properties from commonly gathered items. This should allow the interval-level measurement of locomotor disability in general populations and so improve the quality of longitudinal studies in this area. The use of items from the PF-10 means that this improvement in quality is at no extra burden to the respondent or researcher, as in many cases, data can be analysed retrospectively.
