Outbreaks of hand, foot and mouth disease (HFMD) have increased recently, as has the case fatality rate in severe cases. No scoring system currently exists to predict mortality risk for severe HFMD in previous study. We retrospectively collected laboratory parameters for 546 patients with severe HFMD (a derivation and a validation cohort) at Hunan Children's Hospitals between January 2012 and December 2014. We developed a mortality risk score comprising four laboratory parameters: blood glucose (GLU), white blood cells (WBC), lactate (LAC), and N-terminal-probrain natriuretic peptide (NT-proBNP). Using an "optimal" cutoff score of 4, the sensitivity, specificity, positive predictive value and negative predictive value was 88.00%, 96.14%, 62.86% and 99.08%, respectively, in the derivation cohort. Among severe HFMD patients with low- and high-risk scores in the validation cohort, case fatality rates were 1.49% and 74.07%, respectively. According to the "optimal" cut-off in the derivation cohort, the sensitivity, specificity, positive predictive value, and negative predictive value were 80.95%, 93.83%, 62.96% and 97.44%, respectively, in the derivation cohort. The mortality risk score demonstrated good discrimination (AUC > 0.9) and calibration (P > 0.05) in both cohorts. The mortality risk score, comprising WBC, GLU, LAC and NT-proBNP, has been demonstrated good discrimination and calibration in the both cohorts.