BACKGROUND: Fluid resuscitation is the recommended management of shock, but increased mortality in febrile African children in the FEAST trial. We hypothesised that fluid bolus-induced deaths in FEAST would be associated with detectable changes in cardiovascular, neurological, or respiratory function, oxygen carrying capacity, and blood biochemistry. METHODS: We developed composite scores for respiratory, cardiovascular, and neurological function using vital sign data from the FEAST trial, and used them to compare participants from FEAST with those from four other cohorts and to identify differences between the bolus (n=2097) and no bolus (n=1044) groups of FEAST. We calculated the odds of adverse outcome for each ten-unit increase in baseline score using logistic regression for each cohort. Within FEAST participants, we also compared haemoglobin and plasma biochemistry between bolus and non-bolus patients, assessed the effects of these factors along with the vital sign scores on the contribution of bolus to mortality using Cox proportional hazard models, and used Bayesian clustering to identify subgroups that differed in response to bolus. The FEAST trial is registered with ISRCTN, number ISRCTN69856593. FINDINGS: Increasing respiratory (odds ratio 1·09, 95% CI 1·07-1·11), neurological (1·26, 1·21-1·31), and cardiovascular scores (1·09, 1·05-1·14) were associated with death in FEAST (all p<0·0001), and with adverse outcomes for specific scores in the four other cohorts. In FEAST, fluid bolus increased respiratory and neurological scores and decreased cardiovascular score at 1 h after commencement of the infusion. Fluid bolus recipients had mean 0·33 g/dL (95% CI 0·20-0·46) reduction in haemoglobin concentration after 8 h (p<0·0001), and at 24 h had a decrease of 1·41 mEq/L (95% CI 0·76-2·06; p=0·0002) in mean base excess and increase of 1·65 mmol/L (0·47-2·8; p=0·0070) in mean chloride, and a decrease of 0·96 mmol/L (0·45 to 1·47; p=0·0003) in bicarbonate. There were similar effects of fluid bolus in three patient subgroups, identified on the basis of their baseline characteristics. Hyperchloraemic acidosis and respiratory and neurological dysfunction induced by saline or albumin bolus explained the excess mortality due to bolus in Cox survival models. INTERPRETATION: In the resuscitation of febrile children, albumin and saline boluses can cause respiratory and neurological dysfunction, hyperchloraemic acidosis, and reduction in haemoglobin concentration. The findings support the notion that fluid resuscitation with unbuffered electrolyte solutions may cause harm and their use should be cautioned. The effects of lower volumes of buffered solutions should be evaluated further. FUNDING: Medical Research Council, Department for International Development, National Institute for Health Research, Imperial College Biomedical Research Centre.