Risk Factors and Prognosis of Pneumoconiosis Combined With Bacterial Pneumonia:Application of a Random Forest Model
- VernacularTitle:尘肺合并细菌性肺炎的危险因素与预后:随机森林模型的应用
- Author:
Qiaolan WANG
1
;
Linshen XIE
;
Wen DU
;
Menglin CHEN
;
Rujia YOU
;
Qiaoling JIN
Author Information
- Keywords: Pneumoconiosis; Pneumonia,bacterial; Machine learning; Risk factors; Prognosis
- From: Journal of Sichuan University (Medical Sciences) 2025;56(4):1076-1082
- CountryChina
- Language:Chinese
- Abstract: Objective To apply a random forest model combined with logistic regression in the understudied area of pneumoconiosis complications,and to investigate the incidence and risk factors of pneumoconiosis complicated by bacterial pneumonia,and the effect of concomitant bacterial pneumonia on the survival and prognosis of patients with pneumoconiosis.Methods Pneumoconiosis patients admitted to the West China Fourth Hospital,Sichuan University,between January 2018 and April 2022 were enrolled and divided into a group of those with only pneumoconiosis and another group of those with pneumoconiosis complicated by bacterial pneumonia.Univariate analyses,including chi-squared test,t-test,or rank sum test,were conducted to examine the differences between the groups.A random forest model was used to screen the variables,and the risk factors of pneumoconiosis complicated by bacterial pneumonia were identified by stepwise forward logistic regression method.Cox regression was applied to the survival data to assess the effect of concomitant bacterial pneumonia on the survival and prognosis of pneumoconiosis patients.Results Among the 742 pneumoconiosis patients,536 cases(72.24%)had concomitant bacterial pneumonia.Among the 55 deaths,36 cases(65.45%)had concomitant bacterial pneumonia.Univariate analysis showed statistically significant differences in age,duration of disease,lung function,duration of exposure,lung lavage,pulmonary tuberculosis,and emphysema between the two groups(P<0.05).The variables were screened using the random forest model,and the risk factors were ranked in a descending order of their importance—the types of dust,duration of exposure,lung function,lung lavage,and pulmonary tuberculosis.After screening,multivariate logistic regression analysis showed that the types of dust(compared with silica dust,silicate dust:odd ratio[OR]=8.100,95%CI,1.386-47.331;carbon dust:OR=1.728,95%CI,1.034-2.887;artificial inorganic dust:OR=2.138,95%CI,1.146-3.988),impaired lung function(compared with undamaged lung function group,the group of patients with mild,moderate,and moderately severe damage:OR=2.292,95%CI,1.482-3.544),and pulmonary tuberculosis(OR=1.559,95%CI,1.071-2.271)were risk factors for pneumoconiosis complicated by bacterial pneumonia.The median follow-up was 30.0 months,ranging from 1.0 month to 64.0 months.Cox regression analysis showed that the mortality risk for pneumoconiosis patients with concomitant bacterial pneumonia was 2.369 times higher than that for patients without bacterial pneumonia(95%CI,1.286-4.367).Conclusion Pneumoconiosis patients are susceptible to bacterial pneumonia and are influenced by multiple risk factors.Concomitant bacterial pneumonia markedly affects the patient prognosis.
