There are numerous choices to be made in the design of studies examining the impact of healthcare on patient-reported outcomes. We describe considerations in the design of the Veterans Health Study (VHS), a large-scale longitudinal observational study of healthcare in the Veterans Health Administration (VA). We also consider sampling issues, and discuss the broader theoretical and practical implications of our choices. The VHS was an observational study with a prospective longitudinal design. Subjects were recruited from a cross-sectional sample of the VA patient population, and identified when they came to ambulatory care clinics for a medical visit. Participating patients were contacted by telephone, and scheduled for an interview conducted at the clinic. Prior to the interview they completed a mailed questionnaire. The clinic interview included brief clinical assessments of selected study medical conditions, a medical history interview, limited health examination, and assessments of health status, health-related quality of life, process-of-care measures related to utilization of services, and other patient characteristics. Patients were empaneled and followed over time. Their health was monitored with brief mailed questionnaires completed at 3-month intervals, and with annual patient reassessments at 12 and 24 months. This design had several strengths. Its comprehensiveness and observational nature allowed for examination of a broad range of outcomes and processes of care as they occur in routine practice in the VA system. Study effects on outcomes should be minimal and the longitudinal design permitted the examination of changes in health status and evaluation of the extent to which changes in patients' illnesses and their treatments were associated with changes in outcomes. Many aspects of this study's design were innovative, reflecting careful consideration of design choices and lessons learned from previous outcomes research studies. Choices made in the design of the VHS can serve as models for future studies of the effects of healthcare on patient-reported outcomes.