Social isolation presents a high risk for mortality. Evidence is limited, however, for older adults, especially in Japan, a country whose life expectancy at birth is the longest in the world. This retrospective longitudinal study included homebound Japanese citizens aged ≥75 years who underwent voluntary checkups conducted by Zentsuji City, Kagawa Prefecture, Japan from 2000 to 2024. The relationship between social isolation and mortality was assessed using the Cox proportional hazards model. Social isolation was defined as having no regular contact with family or friends, and survival information was confirmed by the city's database as of 1 July, 2024. Of the 3,366 participants (male: 42.4%), 3,024 (male: 42.6%) remained in the final cohort. At study entry, socially isolated participants tended to be older and have fewer exercising habits, worse self-rated health status, worse life satisfaction, and no family members in the same household with them. The mean follow-up time was 2.47 years, during which 9.1% of the participants died. Mortality risk was 2.37 times higher (95% confidence interval [CI]: 1.61–3.48) for socially isolated individuals than for those who were not socially isolated, after controlling for covariates such as physical status, self-rated health, lifestyle, life satisfaction, and sharing a household with family members. The results showed a consistent trend, even after excluding participants with short follow-up periods. Consequently, when evaluating mortality risks for older adults, it is important to consider whether they are socially isolated.