Analysis Of Influencing Factors Of Hearing Loss In Oil Workers And Establishment Of Risk Assessment Model

Objectives To investigate the current status of hearing loss in oil workers and analyze the influencing factors.Based on deep learning theory,the risk assessment models for hearing loss of oil workers were constructed and the optimal one was obtained by comparison.Methods Using the method of current situation research,the workers of an oil company who participated in occupational health examinations from 2018 to 2019 were selected as the research objects.The basic information about oil workers was collected through questionnaires,and a database was constructed in combination with the physical examination results collected by the hospital.According to the measurement results of pure tone audiometry to determine whether oil workers had hearing loss.Univariate,unconditional logistic regression analysis were used to screen the influencing factors of hearing loss in oil workers.The input variables of the risk assessment models in this study were determined based on the univariate analysis,multivariate analysis and related literature reviews of the hearing loss of oil workers.The random forest model,XG boost model and BP neural network model were constructed using Python to evaluate the risk of hearing loss in oil workers.At the same time,the performance of three models was comprehensively evaluated from the aspects of discrimination and calibration,and the optimal risk assessment model for hearing loss of oil workers was obtained by comparison.Finally,the risk assessment result of this study was visually displayed in the nomogram.Results 1 The detection rate of hearing loss among the workers of the oil company was27.06%.2 The results of the univariate analysis showed that the detection rate of hearing loss among oil workers of different age,sex,monthly household income,diabetes,labor intensity,physical exercise,ototoxic chemical toxicity exposure,sleep disorders,shift conditions,high temperature exposure,working age and cumulative noise exposure were statistically significant(P<0.05).3 The results of the multivariate logistic analysis showed that the hearing loss of oil workers aged over 50 years old,with diabetes and insomnia were 2.101(1.183～3.732)times,1.472(1.058～2.047)times,and 1.462(1.013～2.110)times as many as those 20～30 years old,who did not suffer from diabetes and had no sleep disorders,respectively.Hearing loss were 1.463(1.125～1.903)times higher in those exposed to ototoxic chemicals than in those without exposure.Hearing loss in those with≥30 years of service were 1.992(1.217～3.260)times that of those with <10 years of service.Hearing loss occurred 5.387(3.188～9.305)times and 6.589(3.914～11.091)times more frequently in those with cumulative noise exposure of 90～95 d B(A)year and ≥95d B(A)year than in those with cumulative noise exposure<80 d B(A)year,respectively.Workers with monthly household income over 11,000 yuan and moderate labor intensity had hearing loss,which were 0.530(0.300～0.938)times and 0.615(0.391～0.966)times that of workers with monthly household income below 5,000 yuan and low labor intensity,respectively.4 The goodness of fit test results of random forest,XG boost,and BP neural network models was 0.954,0.951 and 0.881,respectively.In terms of discrimination,the accuracy rate of the three models was 95.99%,95.22% and 88.62%,the sensitivity was97.69%,97.50% and 95.47%,and the specificity was 91.43%,89.09% and 70.13%,The ascent index was 0.89,0.87 and 0.66,the F1 score was 0.74,0.73 and 0.73,and the AUC was 0.95,0.93 and 0.83,respectively.In terms of calibration,Brier score was 0.04,0.04,and 0.11,observed-expected ratio was 1.02,1.04 and 1.21,and Calibration-in-the-large was 0.029,0.032 and 0.097,respectively.Conclusions 1 Advanced age,diabetes,ototoxic chemical exposure,insomnia,shift work,long working life,and excessive cumulative noise exposure were risk factors for hearing loss,and high monthly household income and moderate labor intensity were protective factors for hearing loss.2 The performance of the random forest model applied to the risk assessment of hearing loss in oil workers was better than the XG boost model and the BP neural network model,and it can more accurately assess the risk of hearing loss in oil workers.Figure8;Table15;Reference 153...