Objective Assessment of abnormal vital signs in triage is a challenge in the paediatric emergency department (PED), since vital signs may reflect anxiety, fever or pain rather than the clinical deterioration of the child. We aimed to evaluate the efficacy of subjective 'down-triage' (change of the initially determined acuity levels) of Japanese Triage and Acuity Scale ( JTAS). Methods This is a retrospective cohort study of patients in PED up to 15 years of age at a tertiary paediatric medical centre in Japan during a 1-year period. At the end of every JTAS triage process, PED nurses were allowed to 'down-triage' acuity levels of well-appearing patients with abnormal HR or RR, which were presumably attributable to fever, crying or being upset. We compared predictive performance of the triage system before and after 'down-triage' using admission rate as the primary outcome. Results Among 37 961 PED visits during the study period, we analysed 37 219 records. A total of 17 089 patients (45.9%) were 'down-triaged' after their initial triage allocation upon arrival. Admission rates after 'down-triage' (83%, 33%, 7%, 1% and 3% for levels 1-5, respectively), compared with those of unmodified initial level (16%, 11%, 6%, 2% and 6% for levels 1-5, respectively), had a better apparent relevance with the anticipated admission rates of Canadian Triage and Acuity Scale. Conclusions Modification of JTAS through 'downtriage' by experienced staff improves prediction of disposition in a PED. Further research is needed to determine an objective protocol for 'down-triage' to ensure safe practice in a PED.