Abstract Background Diagnosing tuberculosis (TB), the leading cause of death in people with HIV, remains a challenge in resource-limited countries. We assessed TB diagnosis using a strategy that included systematic urine lipoarabinomannan (LAM) testing for all HIV patients hospitalized in medical wards and 6-month mortality according to LAM results. Methods This prospective, observational study included adult HIV patients hospitalized in the medical wards of a public district hospital in Malawi regardless of their TB symptoms or CD4 count. Each patient had a clinical examination, and Alere Determine TB-LAM, sputum microscopy, sputum GeneXpert MTB/RIF (Xpert), chest x-ray, and CD4 count were systematically requested. Results Among 387 inpatients, 54% had a CD4 <200 cells/µL, 64% had presumptive TB, and 90% had ≥1 TB symptom recorded in their medical file. LAM results were available for 99.0% of patients, microscopy for 62.8%, and Xpert for 60.7%. In total, 26.1% (100/383) had LAM-positive results, 48% (48/100) of which were grades 2–4. Any TB laboratory test result was positive in 30.8% (119/387). Among patients with no Xpert result, 28.5% (43/151) were LAM-positive. Cumulative 6-month mortality was 40.1% (151/377): 50.5% (49/97) in LAM-positives and 36.2% (100/276) in LAM-negatives (P = .013). In multivariable regression analyses, LAM-positive patients had a higher risk of mortality than LAM-negatives (adjusted odds ratio, 2.5; 95% CI, 1.1–5.8; P = .037). Conclusions In resource-limited hospital medical wards with high TB prevalence, a diagnostic strategy including systematic urine LAM testing for all HIV patients is an easily implementable strategy that identifies a large proportion of patients with TB at risk of death.