Epidemiological understanding is incomplete if social system influences on statistical rates are not appropriately recognized. To the traditional epidemiological triumvirate of agent, host, and environment, a fourth influence must now be added--contributions from the social system (like government reimbursement policies, organizational priorities, and the variable behavior of providers). This paper has two sections: the first builds a progressively more complex model (from simple reification, through environmental and personal attributes, to social system influences) to depict the relative contribution of diverse influences to the social production of "facts." The second section considers how social system influences affect health statistics--most emphasis is given to the contribution of health provider behavior (using the widely accepted gender inequality in heart disease as an example). Data from a wide range of studies (employing seven different research methods) are consistent with an underlying thesis--that the gender difference in heart disease is as much a function of how providers and the system respond as it is of the biophysiology of patients and how they react to events. The incorporation of system influences permits the development of an epidemiological imagination, or the ability to link intimate health experiences with more remote social-structural influences.