Objective: The aim of this study was to measure the prevalence of social isolation in community-dwelling elderly and related factors based on household composition differences. Method: We used the six-item Lubben Social Network Scale to measure social isolation in 2,000 individuals. Multiple logistic regression analysis was performed to examine factors related to social isolation with household composition after adjusting for gender and age. Results: The prevalence of social isolation was 31.0% for elderly living alone and 24.1% for those living with family. For both, poor mental health and lack of social support from nonfamily members were associated with social isolation risk. For elderly living with family, low intellectual activities and poor health practice were associated with social isolation risk. Discussion: This study showed high prevalence of social isolation. For prevention, promoting mental health and encouraging them to make friends may be important. For elderly living with family, promoting intellectual activities and good health practice is recommended.