2011 Volume 21 Issue 3 Pages 265-281
Based on the empirical findings from the observational studies conducted by the Ministry of Health, Labour and Welfare’s specific study group (Tsuji et al., 2009), we examine the effectiveness of continuous participation in the “Exercise Program aimed at Improvement of Physical Capabilities of Potentially Frail Elderly Persons” (EXERCISE), which was introduced to prevent frail elderly people from requiring a long-term caretaker in his/her aging daily living, using Survival analyses and Heckman’s Difference in Differences (DID) estimation after the implementation of the Propensity Score Matching (PSM) method to avoid any harmful effect due to the selection bias.
As a result, we reconfirm the significant effect of continuous participation in EXERCISE with the result of survival analyses using an accelerated failure time model (AFT model). An estimated acceleration factor (AF), which describes “Stretching out” of the survival time when comparing the two groups (Participation and Non-Participation) and which is obtained by exponentiation of the estimated coefficient of TRT variable, giving an AF value of 1.6 for the “Person at High Risk of Frailty=PHRF (Tokuteikoureisha)” and an AF value of 3.1 for the “Person Requiring Support=PSR (Yoshiensha)” implies that the median survival time for the Participation group in PHRF is slightly less than double the survival time of the Non-Participation group and that in PSR is little more than triple, respectively. Moreover, we confirm that the estimated coefficients on predictor variables of the HDS-R Test score and on the dummy variables indicating the history of a certain disease obtained by the AFT parametric model indicate a significant effect on the results statistically.
In addition, DID estimations using the “Kihon-checklist” score indicate that the Kihon-checklist score for the Participation group of HRF increased by 2.0-2.5 units and by 0.45-0.5 units for PSR compared to that for the Non-Participation group of PHRF and that for PSR, respectively. Also, we found a significant effect on the score improvement of the Kihon-checklist in the both the estimation using samples for HRF and using for all samples.