Objective: This secondary analysis of data of a randomized controlled trial (RCT) retrospectively investigated the performance of pulse oximetry in identifying children with severe illnesses, with and without respiratory signs/symptoms, in a cohort of children followed for morbid episodes in an intervention trial assessing the efficacy of Intermittent Preventive Treatment for malaria in infants (IPTi) in Papua New Guinea (PNG) from June 2006 to May 2010. Setting: The IPTi study was conducted in a paediatric population visiting two health centres on the north coast of PNG in the Mugil area of the Sumkar District. Participants: A total of 669 children visited the clinic and a total of 1921 illness episodes were recorded. Inclusion criteria were: age between 3 and 27 months, full clinical record (signs/symptoms) and pulse oximetry used systematically to assess sick children at all visits. Children were excluded if they visited the clinic in the previous 14 days. Outcomes: The outcome measures were severe illness, severe pneumonia, pneumonia, defined by the Integrated Management of Childhood Illness (IMCI) definitions, and hospitalization. Results: Out of 1921 illness episodes, 1663 fulfilled the inclusion criteria. A total of 139 severe illnesses were identified, of which 93 were severe pneumonia. The ROC curves of pulse oximetry (continuous variable) showed an AUC of 0.63, 0.68 and 0.65 for prediction of severe illness, severe pneumonia and hospitalization, respectively. Pulse oximetry allowed better discrimination between severe and non-severe illness, severe and non-severe pneumonia, admitted and non-admitted patients, in children ≤12-months of age relative to older patients. For the threshold of peripheral arterial oxygen saturation ≤ 94% measured by pulse oximetry (SpO2), unadjusted odds ratios for severe illness, severe pneumonia and hospitalization were 6.1 (95% Confidence Interval (CI) 3.9–9.8), 8.5 (4.9–14.6) and 5.9 (3.4–10.3), respectively. Conclusion: Pulse oximetry was helpful in identifying children with severe illness in outpatient facilities in PNG. A SpO2 of 94% seems the most discriminative threshold. Considering its affordability and ease of use, pulse oximetry could be a valuable additional tool assisting the decision to admit for treatment.