OBJECTIVE: To evaluate caffeine prescribing practices in a tertiary neonatal unit, focusing on initiation, dose adjustment, discontinuation and recommencement, and to assess associations with gestational age and respiratory support. DESIGN: Retrospective observational study. SETTING: Neonatal unit, John Radcliffe Hospital, Oxford, United Kingdom. PATIENTS: Preterm infants born ≤32 weeks gestation and admitted between 1 February 2022 and 31 October 2023. Data extracted from paper patient records included daily caffeine dosing, initiation, discontinuation, recommencement, coadministration with doxapram, demographics and duration of respiratory support. Associations between caffeine administration and clinical factors such as gestational age were assessed using regression. RESULTS: 168 admissions were analysed from 163 infants. Caffeine was typically initiated with a loading dose of 20 mg/kg, and maintenance doses ranged from 5 mg/kg/day to 25 mg/kg/day. There were 1-8 dose adjustments per admission. Doxapram was administered to 19 infants. Caffeine was discontinued at a median (IQR) postmenstrual age of 34.0 (33.9-34.7) weeks and was recommenced in four infants. Gestational age at birth was negatively correlated with postmenstrual age at discontinuation (r(CI) -0.33 (-0.51 to -0.12), p=0.0029; R²=0.11) and infants born at lower gestational ages received higher doses. CONCLUSION: Caffeine therapy in this unit showed marked variability in dosing, discontinuation and recommencement, highlighting the individualised nature of bedside decision-making, which may reflect clinical response to therapy.