Factors associated with secondary lung injury in acute organophosphorus pesticide poisoning:based on random forest and decision tree models
[Objective]To investigate the related factors of secondary lung injury caused by acute organophosphorus pesticide poisoning(AOPP),and formulate reasonable measures to reduce the risk of lung injury.[Methods]A total of 102 patients with AOPP admitted to Nanyang First People's Hospital from March 2021 to March 2023 were selected as research objects.According to the incidence of lung injury within 12 h from onset to admission,they were divided into lung injury group and non-lung injury group.General data and laboratory indicators of the two groups were compared,and the influencing factors of secondary lung injury of AOPP were analyzed by random forest and decision tree model.Receiver operating characteristic(ROC)curve and area under curve(AUC)were plotted to analyze the prediction efficiency of the two models.[Results]The incidence of lung injury in 102 AOPP patients was 70.59％(72/105).Decision tree model showed that AOPP patients with dose ≥40 mg/kg+APACHEⅡ score ≥15 points had a high incidence of lung injury;AOPP patients with dose ≥40 mg/kg+ChE＜450 U/L+TGF-β1≥10 μg/mL+atropinization time ≥90 h had a high incidence of lung injury.Random forest model showed that the amount of poison had the highest effect on secondary lung injury in AOPP patients,followed by ChE at admission,TGF-β1 at admission and atropinization time.ROC curve showed that the accuracy and specificity of random forest model in predicting AOPP secondary lung injury(88.24％,90.91％)were higher than those of decision tree model(67.65％,59.10％)(P＜0.05).[Conclusion]The prediction model based on random forest model can accurately predict the secondary lung injury of AOPP,and its prediction ability is better than that of decision tree model,which can assist medical personnel to make clinical decision and reduce the risk of lung injury.
organophosphorus pesticide poisoninglung injuryrelevant factorsrandom forest modeldecision tree model
万方数据
维普
