Purpose: The drug resistance pattern in tuberculosis (TB) is still under investigated. We analyzed the clinical data from the patients with smear positive TB and applied the model to predict the patients with smear-po...Purpose: The drug resistance pattern in tuberculosis (TB) is still under investigated. We analyzed the clinical data from the patients with smear positive TB and applied the model to predict the patients with smear-positive TB. Materials and Methods: Medical records information of 6977 cases was included from 11,950 inpatients from January 2009 to November 2013. The cases data were divided into a training set, test set and prediction set. Logistic regression analysis was applied to the training set data to establish a prediction classification model, the effect of which was then evaluated using the test set by receiver operating characteristic (ROC) analysis. The model was then applied to the prediction set to identify incidence of snMDR-TB. Results: Sixteen factors which correlate with MDR-TB-including frequency of hospitalization, province of origin, anti-TB drugs, and complications, were identified from the comparison between SP-TB and spMDR-TB. The area under the ROC curve (AUC) of the prediction model was 0.752 (sensitivity = 61.3%, specificity = 83.3%). The percentage of all inpatients with snMDR-TB (snMDR-TB/Total) was 28.7% ± 0.02%, while that of all SN-PTB with snMDR-TB (snMDR-TB/SN-PTB) was 26.5% ± 0.03%. The ratio of snMDR-TB to MDR-TB (snMDR-TB/MDR-TB) was 2.09 ± 0.33. Conclusion: snMDR-TB as an important source of MDR-TB is a significant hidden problem for MDR-TB control and can be identified by the prediction model. A kind of vicious circle with a certain delay effect exists between snMDR-TB and MDR-TB. To better control MDR-TB, it is necessary to pay greater attention to snMDR-TB, conduct further research and develop targeted therapeutic strategies.展开更多
文摘Purpose: The drug resistance pattern in tuberculosis (TB) is still under investigated. We analyzed the clinical data from the patients with smear positive TB and applied the model to predict the patients with smear-positive TB. Materials and Methods: Medical records information of 6977 cases was included from 11,950 inpatients from January 2009 to November 2013. The cases data were divided into a training set, test set and prediction set. Logistic regression analysis was applied to the training set data to establish a prediction classification model, the effect of which was then evaluated using the test set by receiver operating characteristic (ROC) analysis. The model was then applied to the prediction set to identify incidence of snMDR-TB. Results: Sixteen factors which correlate with MDR-TB-including frequency of hospitalization, province of origin, anti-TB drugs, and complications, were identified from the comparison between SP-TB and spMDR-TB. The area under the ROC curve (AUC) of the prediction model was 0.752 (sensitivity = 61.3%, specificity = 83.3%). The percentage of all inpatients with snMDR-TB (snMDR-TB/Total) was 28.7% ± 0.02%, while that of all SN-PTB with snMDR-TB (snMDR-TB/SN-PTB) was 26.5% ± 0.03%. The ratio of snMDR-TB to MDR-TB (snMDR-TB/MDR-TB) was 2.09 ± 0.33. Conclusion: snMDR-TB as an important source of MDR-TB is a significant hidden problem for MDR-TB control and can be identified by the prediction model. A kind of vicious circle with a certain delay effect exists between snMDR-TB and MDR-TB. To better control MDR-TB, it is necessary to pay greater attention to snMDR-TB, conduct further research and develop targeted therapeutic strategies.
