Examination of Matching Methods,Sparse Effects,and Limitations in a Nationwide Database Study on Alzheimer’s Disease

Abstract

Akada et al. conducted a nationwide database study on patients with Alzheimer’s disease, examining risk factors and outcomes over 3 years. A significant association emerged between decreased daily activities and hip fractures. However, the odds ratio was 1.95 (with p = 0.020) may be inaccurate in men, considering the wide 95% confidence interval (1.12–3.51). Possible influencing factors include an inappropriate outcome variable, sparse-data bias, collinear covariates, and comorbidities. Moreover, exact propensity-score matching would be more efficient than nested matching. Limitations include potential recall bias in measuring daily activities and limited applicability of cause-effect relationships in a national database study.