Objective   The prognosis of patients with paraquat poisoning was predicted, according to the commonly used and simple detected clinic indicators, in order to take timely and effective treatment and improve their survival rate.

  Methods   Totally 199 patients with paraquat poisoning treated in Guangzhou 12th people's Hospital during 2010 to 2019 were studied. The basic information of patients and their blood gas analysis indexes, coagulation indexes, blood routine indexes and blood biochemical indexes within 24 hours after admission were abstracted. SPSS 26 and R 4.0.3 software were used to sort out and statistically analyze the data. Lasso regression was used to screen out the influencing factors affecting the prognosis of patients with paraquat poisoning, and then lasso -logistic clinical prediction model was constructed by R software to verify it.

  Results   The fatality among 199 patients was 62.31%(124 / 199). Lasso regression screening showed that the intake dose of paraquat, bicarbonate ion concentration(HCO₃^-), thrombin time(TT), leukocyte count(WBC), blood glucose (Glu), urea nitrogen (BUN) and blood creatinine (SCR) were the potential prognostic factors of these patients. Further logistic regression analysis showed that the increase of oral intake dose, WBC and SCR were the risk factors of death in patients with paraquat poisoning (OR > 1, P < 0.05); the increase of HCO₃^- concentration was a protective factor for the death of patients with paraquat poisoning (OR = 0.811, P < 0.05). The C value of the nomogram model for paraquat poisoning prediction was > 0.9, and the ideal curve in the calibration curve highly coincided with the calibration curve. The analysis of decision curve showed that the prediction model had clinical effectiveness in the whole threshold range.

  Conclusions   The fatality of paraquat poisoning was high. The prediction model of death risk nomogram of paraquat poisoning based on lasso-logistic regression had certain accuracy and operability.