2022 Volume 16 Issue 1 Pages 30-39
Objective: This study aimed to examine how to treat missing data of the Kaigo–Yobo checklist（KYCL）.
Methods: Participants constituted 6,187 independent older adults. We narrowed down the candidates’ score assigned to the missing data using the Cox proportional hazard analysis with the KYCL response status. Then, comparisons were made using the multiple imputation method.
Results: The average value of score of no answer for each item assigned based on the partial regression coefficient was 0.86 points and the median value was 0.63 points. We compared three types of complementation methods:（1）substitute 0.5 points for missing data and round down the total score;（2）substitute 0.5 points and round up the total score; and（3）substitute 1 point with multiple imputation methods. As a result, we found（1）to be the closest in terms of comparison with the results of the multiple imputation method in the score distribution and relationship between frailty and incidence of Long-term Care Insurance Certification.
Conclusions: It is appropriate to allocate 0.5 points to the missing data of KYCL and discard the total score.
