Abstract
To clarify the factors related to the improvement of the Action Research Arm Test (ARAT) after the stratification
for the baseline ARAT score. One hundred eighty-four stroke patients were stratified into three groups based on
the total ARAT score before the NEURO treatment. Multivariate logistic regression analysis was performed to
assess the relationship between the improvement of ARAT and other factors. In the severe group, the improvements
in FMA-A and C items were significantly associated with those of ARAT. None of the factors in the moderate and
mild groups had a significant relationship. In severe group, the treatment designed to improve
“shoulder/elbow/forearm” and “finger” function was considered as useful. On the other hand, there would be a
divergence between the improvement of motor function and activity level of upper extremities in the moderate and
mild group.