Psychometric properties of the Rivermead Mobility Index in Italian stroke rehabilitation inpatients

Objective: To examine the internal consistency, validity, responsiveness and test scalability of the Rivermead Mobility Index (RMI) in Italian rehabilitation inpatients recovering from stroke. Design: Seventy-three stroke inpatients undergoing rehabilitation were assessed at admission (T0) and five weeks later (T1), using RMI, the motor (motFIM) and cognitive (cognFIM) subscales of the Functional Independence Measure, the 'leg' section of the Motricity Index and Trunk Control Test. Results: Cronbach's alpha of the RMI was 0.92. The item-to-total correlation coefficients (rrb) ranged from 0.36 to 0.83, all p < 0.003. All correlations between RMI scores and the other instruments, both at T0 and T1, were statistically significant (r ≥ 0.49, all p < 0.0001), except those with cognFIM. The difference in RMI scores over the testing period was statistically significant (sign test: z = 7.1, p < 0.0001) and the effect size was 0.89. The coefficient of reproducibility was 0.95 at T0 and 0.93 at T1, and both coefficients of scalability were 0.67. Conclusions: The study confirms the internal consistency, construct validity and responsiveness of the RMI, according to the classic psychometric indexes. However, some minor concerns arise regarding: (a) a floor effect of RMI in subacute rehabilitation stroke inpatients at admission and; (b) one item ('bathing') that seems sensitive to cultural and environmental factors. Moreover, even though RMI met the scaling criteria, the item hierarchy is not coincident with the one originally postulated. So, RMI should be considered only as a summated index with ordinal properties, and not a hierarchically ranked scale.