Comparison of the accuracy of multiple regression analysis using four methods to predict the functional independence measure at discharge

Abstract

Tokunaga M, Yamanaga H. Comparison of the accuracy of multiple regression analysis using four methods to predict the functional independence measure at discharge. Jpn J Compr Rehabil Sci 2020; 11: 65-72.Objective: This study aims to compare the accuracy of four methods of multiple regression analysis in predicting the motor functional independence measure (mFIM) at discharge. Methods: The subjects of this study were 1,064 stroke patients who had been hospitalized in a convalescent rehabilitation hospital. Standard multiple regression analysis (S prediction) with mFIM at discharge as the objective variable, multiple regression analysis with reciprocal number of mFIM at admission as one of the explanatory variables (R prediction), prediction of the effectiveness of mFIM by multiple regression analysis, the conversion to mFIM at discharge (E prediction), and the creation of two multiple regression prediction formulas (S2 prediction) were performed. The absolute values of residuals (actual value minus predicted value) of mFIM at discharge were compared for the four methods. Results: The absolute value of the residuals was significantly smaller in the R prediction, E prediction, and S2 prediction than that in the S prediction. In addition, the absolute value was found to be significantly smaller in the E prediction and S2 prediction than that in the R prediction. Conclusion:In multiple regression analysis, the use of E prediction or S2 prediction is recommended because of their high prediction accuracies.

Explanatory variables to use in a multiple regression analysis to predict stroke patients' motor FIM score at discharge from convalescent rehabilitation wards: an investigation of patients with a motor FIM score of less than 40 points at admission

“…Therefore, it is necessary to stratify patients. Although methods have been reported to improve prediction accuracy in the high mFIM group [4][5][6], these methods were not effective in the low mFIM group [8]. In this study, we aimed to improve the prediction accuracy of the low mFIM group by using appropriate explanatory variables.…”

“…Explanatory variables to use in a multiple regression analysis to predict stroke patients' motor FIM score at discharge from convalescent rehabilitation wards: an investigation of patients with a motor FIM score of less than 40 points at admission at admission of 40-90 points) with the ceiling effect, but did not improve the prediction accuracy for the low mFIM group (patients with mFIM score at admission of 13-39 points) [8]. It is important to use the appropriate explanatory variables to improve prediction accuracy in the low mFIM group.…”

“…To improve the accuracy of predicting motor FIM (mFIM) improvement through a multiple regression analysis, three methods were reported: 1) conversion of mFIM at admission to a reciprocal number (1/mFIM at admission) and using it as one of the explanatory variables [4], 2) predicting mFIM improvement rate (mFIM effectiveness) through a multiple regression analysis and converting it to mFIM at discharge [5], and 3) creating two prediction formulas using the data of patients with low and high mFIM at admission [6]. However, these methods deal with the problem of the ceiling effect of mFIM gain (small mFIM gain in patients with high mFIM score at admission) [7,8]. These methods significantly improved prediction accuracy in the high mFIM group (patients with an mFIM score Japanese Journal of Comprehensive Rehabilitation Science (2020)…”