Background Little is known if the levels of physical activity required for the prevention of incident heart failure (HF) and other cardiovascular events vary in community-dwelling older adults. Methods We studied 5503 Cardiovascular Health Study (CHS) participants, age ≥ 65 years, free of baseline HF. Weekly metabolic equivalent task-minutes (MET-minutes), estimated using baseline total leisure-time energy expenditure, were used to categorize participants into four physical activity groups: inactive (0 MET-minutes; n = 489; reference), low (1-499; n = 1458), medium (500-999; n = 1086) and high (≥ 1000; n = 2470). Results Participants had a mean (± SD) age of 73 (± 6) years, 58% were women, and 15% African American. During 13 years of follow-up, centrally-adjudicated incident HF occurred in 26%, 23%, 20%, and 19% of participants with no, low, medium and high physical activity, respectively (trend p < 0.001). Compared with inactive older adults, age-sex-race-adjusted hazard ratios (95% confidence intervals) for incident HF associated with low, medium and high physical activity were 0.87 (0.71-1.06; p = 0.170), 0.68 (0.54-0.85; p = 0.001) and 0.60 (0.49-0.74; p < 0.001), respectively (trend p < 0.001). Only high physical activity had significant independent association with lower risk of incident HF (HR, 0.79; 95% CI, 0.64-0.97; p = 0.026). All levels of physical activity had significant independent association with lower risk of incident acute myocardial infarction (AMI), stroke and cardiovascular mortality. Conclusion In community-dwelling older adults, high level of physical activity was associated with lower risk of incident HF, but all levels of physical activity were associated with lower risk of incident AMI, stroke, and cardiovascular mortality. © 2013 Elsevier Ireland Ltd. All rights reserved.