Recent reviews of burn resuscitation have included the suggestion that "fluid creep" may be influenced by practitioner error. Our center uses a nursing-driven resuscitation protocol that permits titration of fluid based on hourly urine output, including the addition of colloid when patients fail to respond appropriately. The purpose of this study was to examine protocol compliance. We reviewed 140 patients (26 children) with burns of ≥20% TBSA who received protocol-directed resuscitation from 2005 to 2010. We compared each patient's actual hourly fluid infusion with that predicted by the protocol. Sixty-seven patients (48%) completed resuscitation using crystalloid alone, whereas 73 patients required colloid supplementation. Groups did not differ in age, gender, weight, or time from injury to admission. Patients requiring colloid had larger median total burns (33.0 vs 23.5% TBSA) and full-thickness burns (15.5 vs 4.5% TBSA) and more inhalation injuries (60.3 vs 28.4%; P < .001) than those who resuscitated with crystalloid alone. Because we included basic maintenance fluids in their regimen, patients had median predicted requirements of 5.4 ml/kg/%TBSA. Crystalloid-only patients required fluid volumes close to Parkland predictions (4.7 ml/kg/%TBSA), whereas patients who received colloid required more fluid than the predicted volume (7.5 ml/kg/%TBSA). However, the hourly difference between the predicted and received fluids was a median of only 1.0% (interquartile range: -6.1 to 11.1%) and did not differ between groups. Pediatric patients had greater calculated differences than adults. Crystalloid patients exhibited higher urine outputs than colloid patients until colloid was started, suggesting that early over-resuscitation did not contribute to fluid creep. Adherence to our protocol for burn shock resuscitation was excellent overall. Fluid creep exhibited by more seriously injured patients was not due to nurses' failure to follow the protocol. This review has illuminated some opportunities for practice improvement, possibly using a computerized decision support system.