Development and validation of a blood-based differential diagnosis model for pulmonary tuberculosis and community-acquired pneumonia
10.19405/j.cnki.issn1000–1492.2026.06.022
- VernacularTitle:肺结核与社区获得性肺炎血液鉴别诊断模型的构建与验证
- Author:
Xiaolan SU
1
;
Rui ZHANG
1
;
Zeqing YANG
1
;
Xinrui WANG
2
;
Ziyu BI
1
;
Nian LIU
1
;
Yunyan HAN
3
;
Qi REN
1
Author Information
1. School of Public Health， North China University of Science and Technology， Tangshan 063210
2. Department of Clinical Laboratory， Chang'an District Center for Disease Control and Prevention， Shijiazhuang 050011
3. Department of Clinical Laboratory， Tangshan Seventh Hospital， Tangshan 063021
- Publication Type:Journal Article
- Keywords:
pulmonary tuberculosis;
community-acquired pneumonia;
blood parameters;
diagnostic model;
external validation;
multivariate Logistic regression model
- From:
Acta Universitatis Medicinalis Anhui
2026;61(6):1143-1150
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo develop and validate a diagnostic prediction model for differentiating pulmonary tuberculosis from community-acquired pneumonia based on routine blood parameters. MethodsA total of 642 patients with pulmonary tuberculosis and 503 patients with community-acquired pneumonia were retrospectively enrolled from the Seventh Hospital of Tangshan. They were randomly divided into a training set and an internal validation set at a 7∶3 ratio. Additionally， 218 patients from the 981st Hospital were independently included as an external validation set. The Boruta algorithm and recursive feature elimination were employed to select predictors from 82 blood parameters， sex， and age. A multivariate Logistic regression model was established and evaluated using receiver operating characteristic （ROC） curves， calibration curves， and decision curve analysis. ResultsThe final model incorporated eight predictors， namely uric acid， urea nitrogen， alkaline phosphatase， monoamine oxidase， alanine aminotransferase， glutathione， neutrophil percentage， and age. The model achieved areas under the curve （AUCs） of 0.800 （95%CI： 0.769-0.830）， 0.787 （95%CI： 0.738-0.836）， and 0.736 （95%CI： 0.667-0.835） in the training， internal validation， and external validation sets， respectively. The model demonstrated good calibration， and decision curve analysis showed clinical net benefit within the threshold probability range of 10%-80%. ConclusionThe developed model exhibits good discriminative ability and clinical utility， serving as an effective early screening tool for primary healthcare institutions.
