The Japan Gerontological Evaluation Study is a nationwide cohort study of individuals aged 65 years and older established in July 2010. Seven months later, one of the study field sites was directly in the line of the 2011 Great East Japan Earthquake and Tsunami. Despite the 1-hour warning interval between the earthquake and tsunami, many coastal residents lost their lives. We analyzed the risk of all-cause mortality on the day of the disaster as well as in the 38-month interval after the disaster. Among 860 participants, 33 (3.8%) died directly because of the tsunami and an additional 95 people died during the 38-month follow-up period. Individuals with depressive symptoms had elevated risk of mortality on the day of the disaster (odds ratio = 3.90 [95% CI: 1.13, 13.47]). More socially connected people also suffered increased risk of mortality, although these estimates were not statistically significant. In contrast, after the disaster, frequent social interactions reverted back to predicting improved survival (hazard ratio = 0.46 (95% CI: 0.26, 0.82)). Depressive symptoms and stronger social connectedness were associated with increased risk of mortality on the day of the disaster. After the disaster, social interactions were linked to improved survival.