Objectives: Severe and critical Hand-Foot-and-Mouth Disease (HFMD) patients have an acute onset and poor prognosis. This study intended to establish an appropriate risk prediction model by analyzing the blood biochemical indicators of patients. Methods: A total of 3,204 patients with HFMD were enrolled in this study, including 2,131 mild patients, 962 severe patients, and 111 critical patients. We first analyzed the data of each group through multivariate statistics based on SIMCA-P and screened out the variables that had important contributions to the discrimination of each group. Furthermore, the risk factors and predictors were screened out by comparison with the results of univariate statistical analysis. Finally, binary logistic regression analysis was used to establish a suitable prediction model. Results: With the aggravation of HFMD patients' conditions, the blood content and risk warning ability of seven indicators of SP, DP, NEUT%, TP, GLB, RBP, and Glu were significantly increased. We found for the first time that the more severe the HFMD patients, the lower the levels of Chr, %MRETIC, and %HRETIC in their blood. The average prediction accuracy of the established models for Mild/Severe, Severe/Critical, and Severe/Critical was 82.89, 96.16, and 89.37%, respectively, and the AUROC was 0.8722 (95%CI, 0.8583 - 0.8861), 0.9499 (95%CI, 0.9339 - 0.9659), and 0.7913 (95% CI, 0.7471 - 0.8356), respectively. Conclusions: Multivariate statistical analysis based on SIMCA-P could be used to analyze the clinical data of HFMD patients. Besides, SP, DP, NEUT%, TP, GLB, RBP, and Glu could be used as risk factors for severe and critical HFMD patients. The abnormal changes of Chr, %MRETIC, and %HRETIC reflected the possible damage to bone marrow hematopoietic function in HFMD patients. The predictive model established by us could be used for the differential diagnosis of Mild/Severe, Mild/Critical, and Severe/Critical.