APSR2026-205

Yes
Late-Breaking
Submitted
Oral Presentation
AO21-001
 
 
https://storage.unitedwebnetwork.com/files/1285/1213453-46903-ASPR-AO21-001_440178.pptx
Format: PPTX Only
 
Format: PDF Only
06. Tuberculosis
1

No more than 15 authors can be listed.
Please ensure the authors are listed in the right order.
If there are multiple affiliations, please use the "+" button next to "Select Country/Region" to add additional affiliation fields.

Shanmuga Priya Kumar * Sri Lalithambigai medical college and hospital Respiratory Medicine Chennai India
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Presenting Author Details
Shanmuga Priya
Kumar
pshanmuga75@yahoo.com
+919677229629
31-40
India
Original Research
Machine Learning–Based Early Risk Stratification of Extrapulmonary Tuberculosis in Diabetic Pulmonary Tuberculosis Patients
Maximum 25 words
Abstract Content
Extrapulmonary tuberculosis contributes to delayed diagnosis and increased morbidity among patients with pulmonary tuberculosis, particularly in those with diabetes mellitus. Early identification of high-risk individuals remains challenging in TB-endemic settings. This study aimed to identify predictors of extrapulmonary tuberculosis and evaluate whether machine learning models improve risk stratification compared with conventional regression.
A retrospective cohort of 500 diabetic pulmonary tuberculosis patients was analysed over 1101.5 person-years. Seventeen patients (3.4%) developed extrapulmonary tuberculosis. Cox proportional hazards regression identified independent predictors. Predictive performance of Decision Tree, Random Forest, Extreme Gradient Boosting, and Cox regression models was assessed using area under the receiver operating characteristic curve, precision, and recall.
Independent predictors included age ≥52 years (AHR 4.8; p=0.03), elevated glycated haemoglobin (AHR 3.2; p=0.01), insulin use (AHR 2.5; p=0.04), and alcohol consumption (AHR 4.0; p=0.02). Normal body mass index (18.5–25 kg/m2) was protective (AHR 0.22; p=0.02). Extreme Gradient Boosting showed superior performance (AUC 0.91; precision 87.4%; recall 85.1%), outperforming Random Forest (0.89), Decision Tree (0.84), and Cox regression (0.82). Significant interactions were observed between age and glycated haemoglobin, and body mass index and alcohol use.
Metabolic and behavioural factors influence extrapulmonary tuberculosis risk among diabetic pulmonary tuberculosis patients. Machine learning models may enable early risk stratification in TB-endemic populations. Prospective multicentre validation is required before clinical implementation.
 
 
 
 
 
No
No
 
JPG or JPEG or PNG format ONLY
Extrapulmonary tuberculosis, Diabetes mellitus, Machine learning
maximum 3 keywords and no abbreviations, separated by commas
None
240