Quantitative Mobility Subtask Measures and Parkinsonian Gait (Stage 2).

This table shows the final step of a multistage process which was used to develop 5 quantitative mobility subtask measures from whole body sensor recordings and to examine their associations with parkinsonian gait score. On the left is a series of linear regressions to determine which of the 5 quantitative mobility subtask scores were associated with parkinsonian gait score. Each cell shows the β coefficients from the regression for the terms included, with (Standard Error, p-value) below. The Adj-R-sq is the adjusted R2 with the adjusted parkinsonian score, that is the fraction of variation explained relative to the variation of parkinsonian score not explained by demographic terms. By construction of the subtask scores, the single regression coefficients for models 1–5 are all equal to 1. In a second stage (right 2 columns), we employed backward elimination and started with a backward elimination regression model (Model 1) that included all 5 subtask scores which were all associated with parkinsonian gait score when considered alone in models A–E. Two subtasks did not show significant independent associations with parkinsonian gait; neither of these two subtasks were retained in the final model which showed that walking, turning and sit to stand accounted for 35% of the variance of adjusted parkinsonian gait.