Abstract
Background
Clinically diagnosed pulmonary tuberculosis (cPTB, PTB diagnosis despite negative bacteriological results) accounts for 36% of global PTB notifications, but factors influencing the diagnosis and its relationship with mortality are not well-characterised. We aimed to identify factors associated with cPTB and assess its association with mortality.Methods
Secondary data analysis of a prospective TB diagnostic cohort that evaluated the performance of artificial intelligence in detecting TB and COVID-19 using digital chest X-rays (dCXRs) among adults in Zambia. Logistic and Cox regression analyses were conducted.Results
Among 582 PTB patients, median age was 36 (IQR: 28-45) years, 70.6% were male and 50.3% had cPTB. Previous TB history (aOR 3.31, 95% CI 2.13-5.13), age ≥50 (aOR 2.36, 95% CI 1.31- 4.21), symptom duration ≥ 2 weeks (aOR 0.60, 95% CI 0.39-0.93) and dCXR interpretation of pneumonia (aOR 0.34, 95%CI 0.18-0.64) were associated with cPTB. Compared to bacteriologically confirmed PTB, cPTB was not associated with increased mortality hazard (aHR 0.73, 95% CI: 0.25-2.11).Conclusions
Half of PTB diagnoses were cPTB despite availability of molecular testing. While mortality was not increased among cPTB patients, there is an urgent need to improve TB confirmation to enhance diagnostic accuracy and ensure appropriate treatment.