Assessment of the effects of factors in stroke rehabilitation using eight multiple regression analyses—An analysis of the Japan Rehabilitation Database—

Abstract

Objective:The objective of the present study was to determine via multiple regression analysis what types of patient groups demonstrate large effects for factors in stroke rehabilitation. Methods: The subjects were 1,465 stroke patients in Kaifukuki rehabilitation wards who were registered in the 2014 Japan Rehabilitation Database. The subjects were stratified into eight groups based on age, motor functional independence measure (FIM) score at hospital admission, and cognitive FIM score at admission; multiple regression analysis was then performed with motor FIM score at discharge as the dependent variable. Results: Among the eight groups, the following independent variables were significant: motor FIM score at admission in seven groups, Nichijo-seikatsukino-hyokahyo at admission in five groups, age and post-onset duration of hospitalization in four groups, cognitive FIM score at admission in three groups, and pre-onset modified Rankin Scale in one group. Conclusion:The creation of multiple predictive formulas in multiple regression analysis enables identification of the types of patient groups which demonstrate large effects for factors in stroke rehabilitation.

Predicting FIM gain in stroke patients by adding median FIM gain stratified by FIM score at hospital admission to the explanatory variables in multiple regression analysis—An analysis of the Japan Rehabilitation Database—

“…In model 3, which included median mFIM gain in the explanatory variables, R* 2 increased to 0.405, which is higher than the largest R 2 value (0.4) in the 20 studies that predicted FIM gain reviewed by Meyer et al [1]. Selecting appropriate explanatory variables [1] and creating multiple prediction formulas [7,8,[11][12][13] may increase the prediction accuracy of multiple regression analyses. In the present study, we found that median mFIM gain (median motor FIM gain stratified by motor FIM score at admission) was an explanatory variable that increased the accuracy of motor FIM gain predictions.…”

The stratification of motor FIM and cognitive FIM and the creation of four prediction formulas to enable higher prediction accuracy of multiple linear regression analysis with motor FIM gain as the objective variable―An analysis of the Japan Rehabilitation Database

“…Sonoda et al [3] showed the need for stratifying FIM score at admission into two groups (FIM score of 80 points or above, and below 80 points), and Hirano et al [4] stratified FIM at admission into three groups with three prediction formulas. Tokunaga et al [5] stratified the three factors of motor FIM score at admission, cognitive FIM score at admission and age, and created eight prediction formulas. Regarding multiple linear regression analysis predicting FIM gain, Wang et al [6] and Tokunaga et al [7] stratified FIM score at admission into two groups and created two prediction formulas, and Imada et al [8] stratified motor FIM score and cognitive FIM at admission, creating three prediction formulas.…”

“…Regarding multiple linear regression analysis predicting FIM gain, Wang et al [6] and Tokunaga et al [7] stratified FIM score at admission into two groups and created two prediction formulas, and Imada et al [8] stratified motor FIM score and cognitive FIM at admission, creating three prediction formulas. Nevertheless, in these reports [2][3][4][5][6][7][8] it was not clear to what degree the prediction accuracy increased with multiple prediction formulas as opposed to a single prediction formula.…”

Comparison of prediction accuracy of the total score of FIM motor items at discharge in post-stroke patients in a Kaifukuki rehabilitation ward

“…To improve prediction accuracy, in addition to the selection of variables to be included in regression models [3], various methods have been proposed. These methods include prior transformation of the variables used for prediction [4], using predicted FIM effectiveness, which is the ratio of the actual amount of improvement achieved to the maximum amount that can be improved [5,6], and construction of multiple prediction formulas within the same study population [7][8][9]. Although previous studies constructed prediction formulas by multiple regression analysis and compared their accuracy, the dependent variables and other conditions of analysis differed among the studies.…”