Because of the potential for confounding by indication (disease severity) in individual-level observational studies of intended treatment effects, a treatment designed to prevent an adverse event may appear to cause it. We use a hypothetical example to show that despite substantial variation in the frequency of treatment among patients residing in different geographic areas, a constant area-specific mortality rate can be observed, indicating the absence of confounding by indication at the ecologic level. The advantage of ecologic over individual-level observational studies in the assessment of intended treatment effects holds even if variations in disease severity, socioeconomic status, and other unmeasured factors are taken into account, as long as treatment utilization is influenced by practice style in the local medical community independently of disease severity. Ecologic studies can suggest the need for changes in practice, help resolve ethical issues, and indicate priorities for randomized trials.