OBJECTIVE: Although it is known that poor medication adherence is a pervasive problem, consequences for the US health care system are not well understood. Our objective was to generate quantitative estimates of the health and economic burden of poor medication adherence. METHODS: We used modeling techniques and data from the published literature to generate annual estimates of coronary heart disease (CHD) deaths and losses in net economic benefit attributable to poor adherence with chronic medications in the US. For CHD mortality, we quantified the extent to which poor adherence to antihypertensive, cholesterol-lowering, and hypoglycemic medications contributes to CHD deaths. We quantified losses in net economic benefit attributable to poor adherence in terms of forgone medical cost offsets and lost QALYs, taking into account also the counter influence of pharmacy costs avoided because poorly adherent patients fill fewer prescriptions. We assumed that 50% of patients are poorly adherent (either discontinuing therapy or not complying with recommended dosing), that the average therapy would have a cost-effectiveness ratio of $50,000/QALY under full adherence, and that QALYs are valued at $100,000 in the US (as assumed in a recent Institute of Medicine report on the cost of the uninsured). RESULTS: We estimate that 37,000 CHD deaths annually (approximately 8% of all CHD mortality in the US) are attributable to poor adherence with antihypertensive, cholesterol-lowering, and hypoglycemic medications. We further estimate that losses in net economic benefit attributable to poor adherence with all chronic medications amount to $138.5 billion annually (2003 USD), including $51.9 billion in forgone medical cost offsets and $242.3 billion in lost QALYs, counterbalanced by $155.8 billion in avoided costs of prescription drugs. These findings are sensitive to assumptions regarding the pervasiveness of poor medication adherence. CONCLUSIONS: The health and economic consequences of poor medication adherence in the US are substantial.