SummaryBackgroundIn resource-limited settings, pneumonia diagnosis and management are based on thresholds for respiratory rate (RR) and oxyhaemoglobin saturation (SpO2) recommended by WHO. However, as RR increases and SpO2 decreases with elevation, these thresholds might not be applicable at all altitudes. We sought to determine upper thresholds for RR and lower thresholds for SpO2 by age and altitude at four sites, with altitudes ranging from sea level to 4348 m.MethodsIn this cross-sectional study, we enrolled healthy children aged 0–23 months who lived within the study areas in India, Guatemala, Rwanda, and Peru. Participants were excluded if they had been born prematurely (<37 weeks gestation); had a congenital heart defect; had history in the past 2 weeks of overnight admission to a health facility, diagnosis of pneumonia, antibiotic use, or respiratory or gastrointestinal signs; history in the past 24 h of difficulty breathing, fast breathing, runny nose, or nasal congestion; and current runny nose, nasal congestion, fever, chest indrawing, or cyanosis. We measured RR either automatically with the Masimo Rad-97, manually, or both, and measured SpO2 with the Rad-97. Trained staff measured RR in duplicate and SpO2 in triplicate in children who had no respiratory symptoms or signs in the past 2 weeks. We estimated smooth percentiles for RR and SpO2 that varied by age and site using generalised additive models for location, shape, and scale. We compared these data with WHO RR and SpO2 thresholds for tachypnoea and hypoxaemia to determine agreement.FindingsBetween Nov 24, 2017, and Oct 10, 2018, we screened 2027 children for eligibility. 335 were ineligible, leaving 1692 eligible participants. 30 children were excluded because of missing values and 92 were excluded because of measurement or data entry errors, leaving 1570 children in the final analysis. 404 participants were from India (altitude 1–919 m), 389 were from Guatemala (1036–2017 m), 341 from Rwanda (1449–1644 m), and 436 from Peru (3827–4348 m). Mean age was 7·2 months (SD 7·2) and 796 (50·7%) of 1570 participants were female. Although average age was mostly similar between settings, the average participant age in Rwanda was noticeably younger, at 5·5 months (5·9). In the 1570 children included in the analysis, mean RR was 31·9 breaths per min (SD 7·1) in India, 41·5 breaths per min in Guatemala (8·4), 44·0 breaths per min in Rwanda (10·8), and 48·0 breaths per min in Peru (9·4). Mean SpO2 was 98·3% in India (SD 1·5), 97·3% in Guatemala (2·4), 96·2% in Rwanda (2·6), and 89·7% in Peru (3·5). Compared to India, mean RR was 9·6 breaths per min higher in Guatemala, 12·1 breaths per min higher in Rwanda, and 16·1 breaths per min higher in Peru (likelihood ratio test p<0·0001). Smooth percentiles for RR and SpO2 varied by site and age. When we compared age-specific and site-specific 95th percentiles for RR and 5th percentiles for SpO2 against the WHO cutoffs, we found that the proportion of false positives for tachypnoea increased with altitude: 0% in India (95% CI 0–0), 7·3% in Guatemala (4·1–10·4), 16·8% in Rwanda (12·9–21·1), and 28·9% in Peru (23·7–33·0). We also found a high proportion of false positives for hypoxaemia in Peru (11·6%, 95% CI 7·0–14·7).InterpretationWHO cutoffs for fast breathing and hypoxaemia overlap with RR and SpO2 values that are normal for children in different altitudes. Use of WHO definitions for fast breathing could result in misclassification of pneumonia in many children who live at moderate to high altitudes and show acute respiratory signs. The 5th percentile for SpO2 was in reasonable agreement with the WHO definition of hypoxaemia in all regions except for Peru (the highest altitude site). Misclassifications could result in inappropriate management of paediatric respiratory illness and misdirection of potentially scarce resources such as antibiotics and supplemental oxygen. Future studies at various altitudes are needed to validate our findings and recommend a revision to current guidelines. Substantiating research in sick children is still needed.FundingUS National Institutes of Health, Bill & Melinda Gates Foundation.