Figure 8

Between-patient relationships of clinical surveillance tools in core group. Outcome measures are represented by discs. Single variable associations, of strength p≤.05, are represented by lines ending at corresponding disc perimeters. The closer the discs, the more variance explained: line length is inversely proportional to variance in one outcome explained by the other. Relationships resting on influential points, or not holding in mixed-effects models, were dismissed. Absence of connecting lines means lack of association. Greatest variance represented by a connecting line between a pair of outcomes (78%) is for free-walking-speed and mean stride-length, least (13%) for reaction time and speed. Variance explained by multiple associates can, of course, be cumulative (e.g. variance in simian posture rating explained individually by flexor- and extensor-rigidity, stride and speed is 30, 53% 40% & 40%, respectively, but together they explain 68%). Principle component analysis of 13 items making up simian posture scale indicated a first component, average score for 9 items, with a greater within-scale consistency than the original (Chronbach’s alpha 0.8 cf 0.7), and retaining association with brady/hypokinesia and rigidity. Other components could not be interpreted simply in terms of remaining items: 3 compensatory strategies (chin projection, hands behind back at stance, shoulder retraction), and dropped jaw. These were not associated with brady/hypokinesia and rigidity, and showed no association inter se.