Abstract
Background: Mortality among children with presumptive tuberculosis (TB) empiric TB treatment can be high. We describe the predictors of death among children with presumptive TB, and the relation between treatment and mortality. Methods: A prospective cohort of children with presumptive TB who underwent clinical assessment, chest radiograph, tuberculin skin test and sputum bacterial tests for TB was followed up for 3 months. TB diagnosis was based on mycobacterial, clinical and radiologic findings. Predictors of deaths were determined using cox regression model. Results: Of 360 children included in the analysis, 31.4% were younger than 2 years; 31.6% were HIV infected and 11.3% were severely malnourished. One hundred forty (38.9%) were diagnosed with TB, 18 (13%) of whom were bacteriologically confirmed. At 3 months of follow up, 25 of 360 (6.9%) children had died: 15 of 140 (10.7%) were receiving TB treatment versus 10 of 220 (4.5%) were not receiving treatment (P = 0.025). Severely malnourished children [adjusted hazard ratio (aHR), 9.86; 95% confidence interval (CI): 3.11-31.23] and those with chest radiographs suggestive of TB (aHR, 4.20; 95% CI: 0.93-19.01) were more likely to die. Children receiving empiric TB treatment had an increased risk of death (aHR, 2.37; 95% CI: 1.01-5.55) compared with children without treatment after adjustment for age, sex, HIV status and Bacillus Calmette-Guerin (BCG) vaccination. Conclusions: The high mortality in children receiving empirically TB treatment highlights the difficulty in diagnosing childhood TB, the increased likelihood of starting treatment in critically ill children and in children with chronic disease, and the possibility of misdiagnosis. It strengthens the need to invest further in early TB detection and diagnosing nonsevere illness.