关键词: 结核病, 化疗依从性, 离散两水平模型

Abstract:

Objective To investigate the impact factors for chemotherapy compliance of tuberculosis patients in Beijing.
Methods Multilevel modeling techniques are commonly used in sociology or education, but seldom is it applied to the area of risk factor detection for tuberculosis. Tuberculosis patients’ registration information from 2005 to 2009 in Beijing was collected from Beijing Research Institute for Tuberculosis Control to explorer the impact factors for chemotherapy compliance of tuberculosis patients. In this study there are 21 222 confirmed tuberculosis patients for the multilevel binomial logistic model. The information included gender, age, occupation, patient registration categories, registered residence, diagnosis type, diagnostic results and reason of treatment termination. The authors identified there is a cluster tendency among participants from one district because of the similar economic level, lifestyle or medical conditions, so multilevel models were fitted to a twolevel hierarchy to identify impact factors affecting chemotherapy compliance of tuberculosis patients.
Results The cluster tendency was detected in the database(χ²=182.819, P<0.001). The risk factors for chemotherapy compliance of tuberculosis patients in Beijing included gender, age, occupation, patient registration categories, registered residence and diagnostic results. The results of the twolevel logistic regression analysis indicated that older than 40 years old(OR=1.357, 95% CI: 1.216~1.514), retreatment(OR=1.422, 95% CI: 1.320~1.532), other provinces(OR=1.501, 95% CI: 1.423~1.583) and smear-positive(OR=1.055, 95% CI: 1.005~1.108 for smear-negative) were risk factors for poor chemotherapy compliance, whereas female(OR=0.904, 95% CI: 0.861~0.949), workers and peasants(OR=0.830, 95% CI: 0.785~0.877 for retired staff, residents and unemployed people), government staff, health staff, teachers and students(OR=0.841, 95% CI: 0.783~0.904 for retired staff, residents and unemployed people) were protective factors.
Conclusion Multilevel binomial logistic models have higher accuracy than traditional multiple regression models. By establishing a multilevel model we can better describe the impacts of risk factors for chemotherapy compliance of tuberculosis patients, provide data for comparing the chemotherapy compliance rate with other regions, and helping the government to formulating targeted surveillance policies and prevention strategies. Strengthening management of poor chemotherapy compliance in tuberculosis patients is the key of improving tuberculosis cure rate and reducing transmission rate.

Key words: tuberculosis, chemotherapy compliance, multilevel binomial logistic model