Numeracy, the ability to understand and apply numerical information, has recently been the focus of much research on health and medical decision making. We report key findings from a systematic review of the empirical literature addressing: (a) subjective (self-rated skill) vs. objective (performance-based) measures of health-related numeracy; (b) correlates and sequelae of low numeracy; (c) relations to measures of health state utilities; (d) evidence-based theories of numerical processing; and (e) interventions, theory-based and not, that are designed to enhance understanding of health-relevant numerical information (e.g., health risks or treatment efficacy). These key findings include the following: National surveys show that numeracy varies significantly according to age, educational attainment, race/ethnicity, and socio-economic status, and low numeracy is apt to characterize the most vulnerable members of society with the highest disease burden. Subjective and objective measures of numeracy are correlated, but are best suited to different circumstances, and subjective measures may reflect self-estimation biases. Research also indicates that low numeracy constrains informed patient choice, reduces medication compliance, limits access to treatments, and affects medical outcomes. Although risk factors for low numeracy are correlated, numeracy explains unique variance in medical decision making, beyond that explained by such factors as education or intelligence, and low numeracy is pervasive. Theories of numerical processing explain some effects of information format (numeric, verbal, and various visual/graphical formats) and of such factors as affect (emotion) on medical decision making. Finally, we examine gaps in knowledge, such as implications of numeracy for patients' understanding of risk estimates derived from models such as the Gail model for breast cancer and challenges to patient-centered informed decision making.