This study evaluated the effects of three different kinds of user fee on the quality of prescribing in rural Nepal. Using data from 33 public health facilities, we performed a controlled before-and-after study, comparing a fee per prescription (covering all drugs in whatever amounts) against one- and two-band fees per drug item (covering a full course of treatment for each item). With the one-band item fee, each item incurred the same fee; with the two-band item fee, more expensive items incurred a higher fee and cheaper ones a lower fee. Thirteen indicators of prescribing quality were evaluated based on an average of 400 prescribing episodes per facility per year. The percentage of prescriptions conforming to standard treatment guidelines was 12% (95% confidence interval [CI] 3% to 21%) and 15% (95% CI 6% to 24%) greater with the one- and two-band item fees, respectively, than with the fee per prescription. Prescribing quality improved through a reduction in the number of unnecessary, but not necessary, drug items prescribed per patient. Item-based fees are associated with significantly better prescribing quality than a fee per prescription; therefore, item-based fees are preferred over a fee per prescription when considering methods of cost recovery.