Abstract
BACKGROUND: In low- and middle-income countries (LMICs), child mortality rates remain substantially higher compared to high-income countries, with many deaths preventable through early recognition of deterioration. This systematic review and meta-analysis investigated predictive values of vital signs and anthropometric measurements for paediatric in-hospital mortality in LMICs. METHODS: A search of publicly available data in PubMed and OVID Embase was conducted in November 2021 and last updated in March 2025. Studies that reported on oxygen saturation; respiratory rate; heart rate; blood pressure; temperature; mid-upper arm circumference (MUAC); and/or weight-for-height z-score (WHZ), and paediatric in-hospital mortality were included. Neonatal and paediatric intensive care unit (PICU) studies were excluded. Data was extracted by two independent authors. Forest plots presented odds ratios (OR) using random effect models. Newcastle Ottawa Scale assessed risk of bias. FINDINGS: 104 out of 21,494 yielded studies were included in descriptive analysis and 75 in meta-analysis, encompassing 255,546 children. Associations with in-hospital mortality were observed in hypoxaemia (OR 5.53, 95% CI 4.18-7.30), tachypnoea (OR 1.65, 95% CI 1.16-2.34), tachycardia (OR 1.80, 95% CI 1.22-2.66), bradycardia (OR 3.29, 95% CI 1.38-7.83), hypotension (OR 4.42, 95% CI 2.54-7.70), hyperthermia (OR 1.31, 95% CI 1.04-1.66), hypothermia (OR 3.92, 95% CI 2.76-5.58), low MUAC (OR 3.22, 95% CI 2.12-4.91), and low WHZ (OR 3.19, 95% CI 2.47-4.11). INTERPRETATION: Several vital signs and anthropometric measurements are strongly associated with in-hospital mortality in children. Hypoxaemia demonstrated the highest odds of mortality, followed by hypotension, hypothermia, bradycardia and severe malnutrition. These findings highlight the need for early recognition and targeted interventions for children presenting with these high-risk signs, to improve outcomes in resource-limited settings and stress the need to monitor vital signs. FUNDING: None.