Prevalence Of Chronic Obstructive Pulmonary Disease And Establishment Of Disease Risk Assessment Model In Liaoning Province,China

Objective:Chronic Obstructive Pulmonary Disease(COPD)is a disease characterized by incompletely reversible and persistent airflow limitation.Although the risk factors for COPD are generally clear,research shows that the prevalence of COPD varies greatly in different regions,and the risk factors are not completely the same.The reasons may be different economic levels,lifestyles,and population aging patterns.There is still a lack of large-scale COPD epidemiological surveys based on strict sampling in Liaoning,which is not conducive to the development of targeted COPD prevention and control.In addition,most COPD patients have no obvious symptoms in the early stage,leading to the progress of the disease to moderate to a severe disease before being diagnosed.Not only does the treatment effect fail,but it also causes a huge economic burden.Pulmonary function test is currently recognized as the gold standard for COPD diagnosis,but due to its time-consuming and costly pulmonary function tests,pulmonary function tests have not been widely carried out.Therefore,the development of efficient COPD risk assessment tools to screen high-risk groups and increase the rate of lung function tests to achieve early diagnosis and treatment is of great significance to the prevention and control of COPD.Bayesian Network(BN),as a probabilistic graph model,can use machine learning to construct a directed acyclic graph(DAG)to infer the probability of illness,thereby realizing risk prediction.BN can estimate uncertain knowledge based on known conditions.When BN is applied to disease state prediction,it can estimate the probability of the disease state according to the state of each relevant disease-related factor of the individual,to realize the individual disease risk assessment.This study intends to use the National COPD Surveillance Project as a platform to establish a COPD risk assessment model that is suitable for the characteristics of the disease of the residents in Liaoning Province and is convenient for practical application based on a rigorously sampled population flow adjustment.The main objectives of thisresearch:(1)To investigate the prevalence of COPD among people over 40 years old in Liaoning.(2)To explore the risk factors related to COPD and analyze the characteristics of risk factors related to COPD in Liaoning.(3)Establish a COPD risk assessment model to provide a scientific basis for screening high-risk populations and formulating COPD prevention and control strategies and measures in Liaoning.Methods:All permanent residents in Liaoning Province over the age of 40 were taken as the target population of the study,and residents who were determined by multi-stage cluster random sampling and met the selection criteria were included as the study objects.Research subjects selection criteria:The subjects of the survey are residents of Chinese nationality who have lived in the monitoring area for more than 6 months in the 12months before the survey,and are 40 years or older.First,select 4 COPD monitoring points according to the level of urbanization(high vs.low),economic level,and practical conditions.At each disease surveillance point,three townships/sub-districts were randomly chosen in the stage.Two villages/neighboring communities were then randomly selected within each township/sub-district.One group of villagers with at least100 households was randomly chosen within each community/village.Finally,100households within each group of villagers were randomly chosen,and one family member aged at least 40 years old was selected randomly from each household by a KISH selection table.Trained staff from local community clinics or health stations asked the participants questions based on a standardized questionnaire and Pulmonary function tests.Pulmonary function tests were then tested following recommendations by the American Thoracic Society.Pre-bronchodilator and post-bronchodilator forced expiratory volume in 1 s(FEV1),forced expiratory volume in 6 s(FEV6),peak expiratory flow(PEF),and forced vital capacity(FVC)were measured.In the current study,COPD was defined as a post-bronchodilator FEV1:FVC ratio less than 70%based on 2017 GOLD guidelines.The degree of obstruction was classified according to GOLD staging criteria.Statistical analysis:The standardized prevalence was estimated in the overall population.We did not impute missing data in this study.The basic situation of the population and the exposure level of risk factors are described by frequency and composition ratio.The difference in the prevalence of COPD among different groups was analyzed using the Rao-Scottc~2test.The potential risk factors were initially identified by multivariate logistic regression.The data set is divided into a training set and a test set at a ratio of 9:1,and a BN prediction model is constructed based on the training set.This study combined the logistic regression results,literature search,and expert experience to set up the blacklist and whitelist,then carried out structural learning and subsequent parameter learning and sorted out the high-risk conditions of COPD.BN is evaluated through internal verification(5-fold cross-validation)and external verification(test set verification).In this study,Pad was used to input and store data,and SAS 9.2(SAS Institute Inc,Cary,NC,USA)was used for descriptive statistical analysis and logistic regression.Bayesian network construction and risk assessment using R software(version3.6.2,The Comprehensive R Archive Network,http://cran.r-project.org/)in"bnlearn"for analysis.All tests are two-sided tests,if P<0.05,the difference is considered to be statistically significant.Results:A total of 2400 people were surveyed in the area of Shenyang Xinmin,Dandong Donggang,Fuxin Haizhou,Benxi Mingshan,and 2397 valid questionnaires were collected.The number of people who had a lung function test of C or above was 2194(98.43%).The estimated standardized prevalence of COPD in the population over 40 in Liaoning was 21.23%.The prevalence rate of the rural population(24.59%)is higher than that of the urban population(13.92%,c~2=20.24,P<0.001);the prevalence rate of men(23.89%)is higher than that of women(18.88%,c~2=6.89,P=0.009);the prevalence of unmarried people(37.57%)is higher than that of married people(19.95%c~2=19.80,P<0.001).The prevalence of COPD is different in different age groups,the higher the age,the higher the prevalence(c~2=206.77,P<0.001).The prevalence of COPD among people with different education levels and different economic incomes is different(c~2=145.76,P<0.001;c~2=32.77,P<0.001).The awareness rate of COPD among residents?40 years old in Liaoning was only 7.34%,and the rate of lung function tests was 4.65%.Among patients with COPD,mild(GOLD1)patients accounted for 79.82%.Logistic regression analysis showed males(OR=1.473 95%CI:1.219?1.781),advanced age(51?60 OR=1.763 95%CI:1.284?2.421;61?70 OR=2.860 95%CI:1.817?4.503;?71 OR=5.789 95%CI:5.438?6.163),middle income(OR=1.105 95%CI:1.029?1.186),smoking(OR=1.437 95%CI:1.284?1.609),coronary heart disease(OR=1.393 95%CI:1.004?1.934),childhood lung infection(OR=1.781 95%CI:1.405?2.257),parents suffering from asthma(OR=3.482 95%CI:2.167?5.596),indoor fuel pollution(OR=1.287 95%CI:1.079?1.534),occupational hazardous gas/dust exposure(OR=1.620 95%CI:1.178?2.227)are the risk factors of COPD,high education level(6?9 OR=0.596 95%CI:0.533?0.665;<9 OR=0.405 95%CI:0.329?0.498),living with smokers before 14 years old(OR=0.737 95%CI:0.548?0.991)are protective factors of COPD.BN-based COPD risk assessment studies show that age,income,smoking,exposure to harmful dust,exposure to harmful gases,cough,and wheezing symptoms are directly related to COPD,and factors such as gender,region,and educational background can indirectly affect COPD through other factors.When a person is exposed to a single factor,the population with the highest risk of COPD is mainly those with frequent cough symptoms,P(COPD|cough)=0.374,withwheezingsymptoms,P(COPD|wheezing)=0.326,advanced age and smoking,P(COPD)|Old age)=0.293,P(COPD|Smoking)=0.253.As exposure factors increase,the risk of COPD gradually increases.When exposed to two factors at the same time,the combination with the highest risk probability of COPD is P(COPD|cough,wheeze)=0.553,P(COPD|cough,advanced age)=0.511;when exposed to three factors at the same time,the risk probability of COPD The highest combination is P(COPD|cough,wheeze,advanced age)=0.681 and P(COPD|cough,wheeze,coronary heart disease)=0.625;when the four factors are exposed at the same time,the highest risk of COPD is P(COPD|cough,Wheezing,smoking,advanced age)=0.738.The cross-validation results of this study show that the Bayesian network model is better than the logistic regression model.The average AUC of the BN model was 0.85(0.832?0.892)and the best accuracy was 0.87(0.846?0.893),which was higher than the logistic regression model(AUC 0.776?0.764,accuracy 0.825?0.840).Conclusions:(1)The prevalence of COPD in Liaoning Province is 21.23%,COPD disease awareness rate,and lung function test rate are both low.(2)As a whole,COPD patients have a higher proportion of patients with mild lung function(GOLD?).The highest proportion of respiratory symptoms in COPD patients are wheezing and dyspnea,and the main chronic diseases are hypertension and coronary heart disease.(3)This study found that men,advanced age,smoking,personal history of diseases(childhood respiratory infections and coronary heart disease),and parents with respiratory diseases(parents with asthma.)are all risk factors for COPD,high economic income and high education degree,living with smokers during childhood are protective factors for COPD;occupational exposure to harmful factors and indoor polluted fuels(biofuels and coal fuels)are both risk factors for COPD in Liaoning Province,which are different from the national survey conclusions.(4)The COPD risk assessment model constructed based on BN is performed better than the logistic regression model in predicting COPD risks.It suggests that the artificial intelligence model constructed by BN and other machine learning methods can be used as an effective tool for COPD risk assessment.(5)The BN model found that the high-risk populations in this region are in order of cough,wheezing,advanced age,and smoking.Residents with simple obesity are at a lower risk than those exposed to other factors,but when obesity is combined with older age and smoking,the risk of disease is higher than that of older age,smoking and other factors.