This paper investigates whether patient-level factors, in particular cost considerations, affect the physicians' prescribing decisions. In the context of a natural experiment, we examine the effect of the first US commercial free-antibiotics program on retail antibiotic sales. We find an overall increase in antibiotic prescriptions under the program and substitutions to covered antibiotics from not-covered antibiotics. The shift away from not-covered antibiotics, particularly from those without covered equivalents, indicates a change in the physicians' prescribing decisions. We locate stronger program effects in low-income areas. Our findings, robust to a variety of specifications, are in contrast with previous literature. Copyright © 2013 John Wiley & Sons, Ltd.