Aims Previous prospective studies of the association between mortality and physical activity have generally not fully accounted for the interplay between movement behaviours. A compositional data modelling approach accounts for relative scale and co-dependency in time-use data across physical activity behaviours of the 24-hour day. Methods A prospective analysis of the National Health and Nutrition Examination Survey 2005-2006 on N = 1468 adults (d = 135 deaths) in ages 50-79 years was undertaken using compositional Cox regression analysis. Daily time spent in sedentary behaviour, light intensity (LIPA) and moderate-to-vigorous physical activity (MVPA) was determined from waist-mounted accelerometer data (Actigraph 7164) and supplemented with self-reported sleep data to determine the daily time-use composition. Results The composition of time spent in sedentary behaviour, LIPA, MVPA and sleep was associated with mortality rate after allowing for age and sex effects (p < 0.001), and remained significant when other lifestyle factors were added (p < 0.001). This was driven primarily by the preponderance of MVPA; however, significant changes are attributable to LIPA relative to sedentary behaviour and sleep, and sedentary behaviour relative to sleep. The final ratio ceased to be statistically significant after incorporating lifestyle factors. The preponderance of MVPA ceased to be statistically significant after incorporating health at outset and physical limitations on movement. Conclusions An association is inferred between survival rate and the physical activity composition of the day. The MVPA time share is important, but time spent in LIPA relative to sedentary behaviour and sleep is also a significant factor. Increased preponderance of MVPA may have detrimental associations at higher levels of MVPA.