Sonja Mensch

98 Chapter 5 specified their judgement by indicating a position along a continuous line between two end-points, the left most side representing the lowest overall motor ability. This resulted in a score between 0-10 centimeters. Therapists then scored the items for the chosen situation. After Movakic completion, participants could not see or review the results of Movakic scores. After each 3-month interval, therapists were asked to note the occurrence of random events potentially positively or negatively influencing motor abilities. Events that could be registered were surgery, increase of contractures, pain, medication change, change in assistive devices, and other, unspecified events. Because of the frequent contacts of therapists with the children, their parents, and caring staff in the centres, therapists were well aware of such information. In summary, the following data were collected: at baseline child characteristics, expert judgment and Movakic score, and at T1 to T5 expert judgment and Movakic scores, as well as any events theorized to influence motor ability during the previous 3 months. Analyses To test construct validity, Pearson correlation between Movakic total scores and expert judgment scores was calculated. Validity was considered good when r >.40 (Cohen., 1992). Responsiveness was tested by comparing Movakic score changes during 3-month intervals with identified events and score changes during intervals without such events. For this purpose, intervals of all children were entered into a new database, with only two measurements, pre-test (T0) and post-test (T1). The new database had an event group and no-event group and could consist of the same children at different time points. An example is given in Table 1. With the new dataset we were able to investigate whether Movakic scores changed in the appropriate interval in thepresenceor absenceof events.We realized that complete data for all six intervals might not be available for all children. It was expected that participants with high numbers of measurements would be overrepresented in comparison with participants who were lost to follow up earlier. As a result, the number of collectedmeasurements of participants might mediate the change inMovakic scores, because children in whom less data were collected on events might be representative of a subgroup of children with characteristics that could also influence change in motor abilities. To determine whether this effect occurred, classification in the event/no event group was correlated with the number of measurements, using point-biserial Pearson’s correlation ( p <.05).