BACKGROUND:Childhood tuberculosis (TB) diagnoses often lack microbiologic confirmation and require empiric treatment. Barriers to empiric treatment include concern for poor outcomes and adverse effects. We thus determined the outcomes of empiric TB treatment from a retrospective cohort of children at a national referral hospital in Kampala, Uganda from 2010 to 2015. METHODS:Children were diagnosed clinically and followed through treatment. Demographics, clinical data, outcome and any adverse events were extracted from patient charts. A favorable outcome was defined as a child completing treatment with clinical improvement. We performed logistic regression to assess factors associated with loss to follow up and death. RESULTS:Of 516 children, median age was 36 months (IQR 15-73), 55% (95% CI 51-60%) were male, and HIV prevalence was 6% (95% CI 4-9%). The majority (n = 422, 82, 95% CI 78-85%) had a favorable outcome, with no adverse events that required treatment discontinuation. The most common unfavorable outcomes were loss to follow-up (57/94, 61%) and death (35/94, 37%; overall mortality 7%). In regression analysis, loss to follow up was associated with age 10-14 years (OR 2.38, 95% CI 1.15-4.93, p = 0.02), HIV positivity (OR 3.35, 95% CI 1.41-7.92, p = 0.01), hospitalization (OR 4.14, 95% CI 2.08-8.25, p < 0.001), and living outside of Kampala (OR 2.64, 95% CI 1.47-4.71, p = 0.001). Death was associated with hospitalization (OR 4.57, 95% CI 2.0-10.46, p < 0.001), severe malnutrition (OR 2.98, 95% CI 1.07-8.27, p = 0.04), baseline hepatomegaly (OR 4.11, 95% CI 2.09-8.09, p < 0.001), and living outside of Kampala (OR 2.41, 95% CI 1.17-4.96, p = 0.02). CONCLUSIONS:Empiric treatment of child TB was effective and safe, but treatment success remained below the 90% target. Addressing co-morbidities and improving retention in care may reduce unfavorable outcomes.