Medical oxygen services are essential for the care of acutely unwell patients. We aimed to assess the effects of a multilevel, multicomponent health-system intervention on hypoxaemia detection, oxygen therapy, and mortality among neonates and children attending level IV health centres and hospitals in Uganda. For this before-after intervention study, we included children who attended paediatric or neonatal wards of 24 level IV health centres and seven general or regional referral hospitals in the Busoga and North Buganda regions of Uganda between June 1, 2020, and June 30, 2022. All neonates younger than 1 month and children aged 1 month to 14 years were eligible for inclusion. We excluded neonates who were not sick but stayed in the maternity ward for routine postnatal care. The intervention involved clinical training, mentorship, and supportive supervision; provision of pulse oximeters and cylinder-based oxygen sources; biomedical-capacity support; and support to develop and disseminate oxygen supply strategies, oxygen therapy guidelines, and lists of essential oxygen supplies. Trained research assistants extracted individual patient data from case notes using a standardised electronic data collection form. Data were collected on health-facility details, age, sex, clinical signs and symptoms, admission diagnoses, pulse oximetry readings, oxygen therapy details, and final patient outcome. The primary outcome was the proportion of admitted neonates and children with a pulse oximetry oxygen saturation reading documented in their patient case notes on day 1 of health-facility admission (ie, pulse oximetry coverage). We used mixed-effects logistic regression to evaluate the effect of the intervention. We obtained data on 71 997 eligible neonates and children admitted to 31 participating health facilities; the primary analysis included 10 001 patients in the pre-intervention period (ie, June 1 to Oct 30, 2020) and 51 329 patients in the post-intervention period (ie, March 1, 2021, to June 30, 2022). Because 1356 patients had missing data for sex, 4365 (46·7%) of 9347 in the pre-intervention group and 22 831 (46·2%) of 49 410 in the post-intervention group were female; 4982 (53·3%) in the pre-intervention group and 26 579 (53·8%) in the post-intervention group were male. The proportion of neonates and children with pulse oximetry at admission increased from 2365 (23·7%) of 10 001 in the pre-intervention period to 45 029 (87·7%) of 51 328 in the post-intervention period. Adjusted analysis indicated greater likelihood of a patient receiving pulse oximetry during the post-intervention period compared with the pre-intervention period (adjusted odds ratio 40·10, 95% CI 37·38-42·93; p<0·0001). Large-scale improvements in hospital oxygen services are achievable and have the potential to improve clinical outcomes. Governments should be encouraged to develop national oxygen plans and focus investment on interventions that have been shown to be effective, including the introduction of pulse oximetry into routine hospital care and clinical and biomedical mentoring and support. Bill & Melinda Gates Foundation and ELMA Philanthropies. For the Luganda and Lusoga translations of the abstract see Supplementary Materials section.