Construction Of Causal Model Of Vitamin D And Drug-resistant Tuberculosis Based On Bayesian Network

Objective Based on the data of drug resistance and related influencing factors of TB patients in Urumqi,classic logistic regression was used to build a model to screen and analyze the relevant factors with statistical effects that may affect drug resistance in TB patients.A Bayesian network was used to construct a causal model to analyze the relationship between the important factors affecting tuberculosis resistance.By constructing a network model of drug-resistant tuberculosis with different levels of VitD,explore and analyze the causal effect of VitD levels and drug resistance in patients.To provide methodological reference and information basis for targeted prevention and intervention measures for TB-resistant high-risk groups.Methods A total of 774 medical records and blood samples were collected from patients with tuberculosis.Describe the basic information of the patients,and use the chi-square test to analyze the levels of VitD in patients with different characteristics and the correlation between VitD and tuberculosis resistance.Multivariate analysis of factors affecting drug resistance was performed using logistic regression.A Bayesian network model was constructed to analyze the causal effects between different VitD levels of tuberculosis resistance.By adjusting the nodes in the network to analyze the effect changes of each influencing factor,the effect of the factors on tuberculosis resistance was further clarified.Results Among the 176 patients monitored for resistance,the total drug resistance rate was 25.0%,the single drug resistance rate was 12.5%,the multidrug resistance rate was 7.4%,and the multidrug resistance rate was 4.5%.The average level of VitD in patients was(30.91±12.92)ng/ml.VitD deficiency patients accounted for 17.5% and VitD deficiency patients accounted for 37.5%.As the patient's VitD level increased,the overall resistance rate decreased.The resistance rates of isoniazid,rifampicin,and ofloxacin increased with increasing VitD levels,but the resistance rates of ethambutol and streptomycin decreased with increasing serum VitD levels.The logistic regression model showed that there were statistical effects between factors such as ethnicity,comorbidities,and drug resistance in patients with tuberculosis.Based on the causal Bayesian network model: when the probability of a positive smear at the end of the second month of treatment is changed to 100%,the drug resistance rate increases by 9%;when the probability of comorbidity is changed to 100%,the drug resistance rate increases significantly12.In addition,patient treatment classification,age,and baseline sputum smear can indirectly affect drug resistance.Causal models of TB patients with different vitamin D levels and TB outcomes are different.Conclusion Compared with classic regression modeling,a Bayesian network model based on causal effects reveals the direct and indirect factors that affect tuberculosis resistance in a combination of display graphics and conditional probability.Strengthening the management and supervision of those with comorbidities and positive smears at the end of the second month of treatment,and early detection of indications for poor treatment will help reduce the drug resistance rate of patients.The Bayesian network causal effect inference has certain application value for accurate analysis of important influencing factors related to key intervention diseases.