[Application of generalized estimation equations to establish prediction equation for tuberculosis drug resistance in Zhejiang province].

Objective: Drug-resistant tuberculosis (TB) may be resistant to one or multiple anti-TB drugs. We used generalized estimation equations to analysis the risk factors of drug-resistant TB and provide information for the establishment of a warning model for these non-independent data. Methods: The drug susceptibility test and questionnaire survey were performed in sputum positive TB patients from 30 anti TB drug-resistance surveillance sites in Zhejiang province. The generalized estimation model was established by the GENMOD module of SAS, with resistance to 13 kinds of anti-TB drugs as dependent variables and possible influencing factors, such as age, having insurance, HBV infection status, and history of anti-TB drug intake, as independent variables. Results: In this study, the probability of drug resistance at baseline level was 20.26%. Age, insurance, whether being co-infected with HBV, and treatment history or treatment withdrawal were statistically significantly correlated with anti-TB drug resistance. The prediction equation was established according to the influence degree of the factors mentioned above on drug resistance. Conclusion: The generalized estimation equations can effectively and robustly analyze the correlated binary outcomes, and thus provide more comprehensive information for drug resistance risk factor evaluation and warning model establishment.