Objective: To investigate the association between physical activity and the risk of mortality, so as to provide the basis for guiding the elderly to maintain appropriate levels of physical activity. Methods: A retrospective cohort study was adopted. Basic information, weekly physical activity items and duration of the elderly aged 50-71 years old was collected from the National Institutes of Health-American Association of Retired Persons Diet and Health Study database. With all-cause mortality risk as the main outcome indicator, controlling for demographic, dietary and disease factors, the association between the duration, metabolic equivalent and intensity of physical activity and all-cause mortality risk was analyzed using restricted cubic spline and multivariable Cox proportional risk regression model. Results: A total of 266 072 participants were included, with an mean age of (70.11±5.36) years old. There were 155 244 males (58.35%) and 110 828 females (41.65%), with a total of 36 006 deaths by December 31, 2011. The median duration of physical activity was 14.00 (interquartile range, 14.00) h/week and the median metabolic equivalents was 53.00 (interquartile range, 54.71) MET-h/week. Restricted cubic spline analysis indicated that the risk of all-cause mortality declined rapidly within the physical activity range of 0 to 15.0 h/week or 50.0 MET-h/week, but with the continuing increase in physical activity, the decline in the risk of all-cause mortality slowed down (all P<0.05). Multivariable Cox proportional risk regression analysis showed that compared with participants with no physical activity, participants with the duration of 0.1-<15.0 h/week, 15.0-<29.5 h/week, ≥29.5 h/week (HR=0.502, 0.386 and 0.368), or the metabolic equivalent of 0.1-<50.0 MET-h/week, 50.0-<110.8 MET-h/week and ≥110.8 MET-h/week (HR=0.511, 0.379 and 0.354) were associated with a reduced risk of all-cause mortality. The combination of mild (0.1-<5.0 h/week), moderate (≥5.0 h/week) and vigorous (≥1.3 h/week) activities had a lower risk of all-cause mortality (HR=0.320). Conclusions Moderate physical activity is associated with a reduced risk of mortality, and it is recommended to do 15.0 h or 50.0 MET-h of physical activity per week in combination with different intensities.