When documentation changes but behaviour persists: de-implementation in a neonatal intensive care unit

De-implementation of low-value clinical practices requires changes in both documentation and clinician behaviour, yet documentation may not accurately reflect bedside practice. We conducted a multifaceted quality improvement de-implementation initiative in a level III neonatal intensive care unit that combined education, electronic medical record workflow modification and reinforcement of symptom-based feeding assessment to eliminate routine gastric residual (GR) checks in infants with birth weight <1500 g. Following implementation, documented GR checks declined to zero, with no increase in necrotising enterocolitis or feeding intolerance and no improvement in time to full enteral feeds.To evaluate behavioural adoption, we performed a secondary analysis of an anonymous nursing survey. Of 56 nurses, 24 responded to the item assessing ongoing practice; 7 of 24 (29%) reported continuing GR checks without documentation. While most nurses supported the protocol change, some reported anxiety and perceived increases in emesis, abdominal imaging and feeding interruptions despite stable clinical outcomes. Less-experienced nurses were more likely to report concern about potential disciplinary consequences (OR 7.33; 95% CI 1.2 to 45; p<0.05). In sensitivity analysis, nurses who did not report continued checks demonstrated higher anxiety (OR 152; 95% CI 3.0 to 7600; p<0.001) although this post hoc finding should be interpreted cautiously.These findings demonstrate discordance between documented and actual practice following de-implementation. Documentation alone may overestimate implementation fidelity, particularly when discontinuing practices perceived as protective, highlighting the need for measurement strategies that capture true bedside behaviour.